Essays on Macroeconomics and Development

Georgios N. Manalis

Thesis submitted for assessment with a view to obtaining the degree of Doctor of Economics of the European University Institute

Florence, 28 May 2021

European University Institute Department of Economics

Essays on Macroeconomics and Development

Georgios N. Manalis

Thesis submitted for assessment with a view to obtaining the degree of Doctor of Economics of the European University Institute

Examining Board Prof. Evi Pappa, Universidad Carlos III Madrid, Supervisor Prof. Axelle Ferrière, Paris School of Economics, Co-supervisor Prof. Tasso Adamopoulos, York University Prof. Katheline Schubert, Paris School of Economics

Researcher declaration to accompany the submission of written work Department Economics - Doctoral Programme

I Georgios Manalis certify that I am the author of the work 'Essays on Macroeconomics and Development' I have presented for examination for the Ph.D. at the European University Institute. I also certify that this is solely my own original work, other than where I have clearly indicated, in this declaration and in the thesis, that it is the work of others.

I warrant that I have obtained all the permissions required for using any material from other copyrighted publications.

I certify that this work complies with the Code of Ethics in Academic Research issued by the European University Institute (IUE 332/2/10 (CA 297).

The copyright of this work rests with its author. Quotation from it is permitted, provided that full acknowledgement is made. This work may not be reproduced without my prior written consent. This authorisation does not, to the best of my knowledge, infringe the rights of any third party.

I declare that this work consists of 28,378 words.

Statement of inclusion of previous work: I confirm that chapter 'Mutual Insurance and land security in rural Ghana' was jointly co-authored with Mr. Karol Mazur and I contributed 50% of the work.

I confirm that chapter 'Contagion as a dealmaker? The effect of financial spillovers on regional lending programs' was jointly co-authored with Mrs. Alica Ida Bonk and Mrs. Alexandra Fotiou and I contributed 33.3% of the work.

Signature and date:

Abstract

Land Rights and risk sharing in rural West Africa: Despite arduous efforts of advancing land rights in Africa, most of the continent experiences low levels of formally recognized property. I propose a novel contextualisation of formal land titling that motivates a theoretical model to account for land reforms’ effects when implemented in weak institutional environments with high risk. Village communities have developed informal mechanisms of risk-sharing to provide households with a safety net, while land allocation is centrally decided by traditional leaders. Therefore, when a land reform, aiming at granting individual property rights, takes place, it operates in a highly antagonistic way to the established customary rules. I build a model of risk-sharing with limited commitment to explain the competing forces developed between statutory land reform and informal mutual insurance at the community level. The model shows that a land reform increases the share of surplus that a villager can extract from a risk-sharing contract among community members and decreases the size of the pie available to the community. Additionally, it shows a non-monotonic relation between land allocation and productivity revealing a trade-off between output efficiency and size of risk-sharing. Subsequently, I use data from Burkina Faso to validate the theoretical predictions.

Mutual insurance and land security in rural Ghana: We study the impact of land rights’ formalization on functioning of informal insurance and land re-allocations in Ghana’s rural communities. First, we provide empirical evidence suggesting that communities holding more of formal land titles enjoy higher land security, as measured by number of disputes due to multiple claims over land. Second, we ﬁnd that land reallocations are more intense in those places, leading to increases in agricultural productivity and level of average consumption. Third, we show that communities with higher formality of land rights enjoy improved risk-sharing against idiosyncratic shocks. Motivated by this evidence, we develop a dynamic model of land and risk sharing subject to limited commitment constraints, where the equilibrium degree of co-operation is determined by the degree of formal land rights chosen. We show that the model can rationalize our empirical ﬁndings and can serve as a useful quantitative laboratory. Most interestingly,

i Essays on Macroeconomics and Development we ﬁnd that although positive in the data, the eﬀects of increasing land rights may be highly non-linear as at some point they may lead to a complete unraveling of informal co-operation in rural economies.

Contagion as a dealmaker? The effect of financial spillovers on regional lending programs: The recent European sovereign debt crisis highlighted the critical role of regional lending arrangements. For the first time, European mechanisms were called to design financing programmes for member countries in trouble. This paper analyses how the risk of contagion, an essential characteristic of interlinked economies, shapes borrowing conditions. We focus on the role of spillovers as a channel of bargaining power that a country might have when asking for financial support from regional lending institutions. We build and present a new database that records both the dates on which official meetings took place, relevant statements were released and the timing of the announcements regarding loan disbursements. This database allows us to assess the defining role that announcements of future actions have in mitigating spillover costs. In addition, we study the design of lending arrangements within a recursive contract between a lender and a sovereign country. When accounting for spillover costs, arising from the borrower to the creditor, we find that it is in the lender’s best interest to back-load consumption by giving more weight to future transfers in order to reduce contagion cost. Subsequently, we test and validate our theoretical predictions by assessing the effect of spillovers on loan disbursements to programme-countries and by juxtaposing lending conditions imposed by the IMF and the European mechanisms.

ii Acknowledgements

I am indebted to Evi and Axelle who are bright examples of supervisors. Always available for meetings, guided my research, advised my steps and were always willing to give their opinion on my research attempts. I was lucky enough to meet more people that greatly assisted my research. Arpad Abraham, even though not my supervisor, plenty of times gave me precious advice, and last but not least, the two academics in my thesis committee, Tasso Adamopoulos and Katheline Schubert. Even though not knowing my work, they were more than willing to read through it and provide helpful insight that will push this research further. I would also like to thank Dimitris Xefteris, an academic I met during the late stages of my graduate studies, but who supported me substantially.

Special thanks are owed to my co-authors, Alica and Karol who not only contributed in building what constitutes this thesis, but became lasting friends on the way. Alexandra holds a special place, as a friend that was always there in need and never stopped advising me on what is best for me.

I am indebted to Gelly, my lifetime companion, who put up with me for all this time, rough and lovely as she has always been, during diﬃcult times, she found the strength to support both of us. Amalia and Nicos never challenged my choices - always challenged my methods and I am deeply grateful for that. Pantelis, my older brother who never stopped taking care of me since day one and keeps on doing so. A bunch of good friends, whining for not visiting Firenze once during my studies, but always justifying my absence over holidays and never stopped believing in me.

During my time in Florence, I was surrounded by a beautiful cohort and managed to make long-lasting friends. I have shared great memories with every single one of them, Essi, Chiara, Alica, Agnes, Oliko, Chiara, Ana, Carolina, Rafa, Gergo, Simon, Matteo, Brian, Adamos, Christoph, Giannis and Alejandro, and I am expecting to share many more. Outside my cohort, Alessandro, the remaining half of the dynamic duet, should be thankful for our meeting - and I am grateful that I crossed land borders for the ﬁrst time in his car and for all the academic insight provided on the way.

iii Essays on Macroeconomics and Development

Lastly, I need to give my special thanks to the greek community in Firenze with a special reference to two prominent figures, Georgios - the man who at our very first meeting was counting how many greeks have failed the first year at the economics department and Stavros who was always there to have a drink after him scoring in Coppa. This list does not do justice to the people that in one way or another helped me go through those years.

iv Contents

Introduction5

1 Land rights and risk sharing in rural West Africa6 1.1 Introduction...... 7 1.2 Literature Review...... 10 1.3 Background on the land reform in Burkina Faso...... 12 1.3.1 Loi 034/2009...... 12 1.3.2 Rural Land Certiﬁcate of Possession (APFR)...... 13 1.3.3 Assessment of the results of the RLG...... 13 1.4 One-Sided Limited Commitment with land re-allocation...... 15 1.5 Data from Burkina Faso...... 22 1.5.1 Rural Land Governance Project...... 22 1.5.2 Monitoring the progress of RLG project...... 23 1.5.3 Empirical Regularities in Burkina Faso...... 23 1.6 Evidence from the RLG programme in Burkina Faso...... 25 1.7 Collateralization eﬀect...... 30 1.8 Conclusion...... 33 References...... 35 1.9 Appendix...... 38 A Proofs...... 38 B Figures...... 55 C Tables...... 61 D Naive productivity measure...... 67

2 Mutual insurance and land security in rural Ghana 70 2.1 Introduction...... 71 2.2 Literature Review...... 73 2.3 Statutory and customary land institutions in Ghana...... 75 2.4 Empirical analysis...... 77

v Essays on Macroeconomics and Development

A Data...... 77 B Suggestive empirical observations...... 80 C Regression analysis...... 81 2.5 Quantitative model...... 86 A Environment...... 86 B Outside Option...... 86 C First best...... 88 D Land and risk sharing with limited commitment...... 88 E Preliminary quantitative results...... 90 F Outlook on quantitative analysis...... 91 2.6 Conclusion...... 93 References...... 95 2.7 Appendix...... 98 A Regression analysis...... 98 B Endogeneity of selling rights...... 99 C Suggestive empirical observations...... 104

3 Contagion as a dealmaker? The eﬀect of ﬁnancial spillovers on regional lending programs 109 3.1 Introduction...... 110 3.2 A new dataset on loan announcements and disbursements during the Euro crisis...... 115 3.3 Empirical analysis...... 117 A The Spillover Index - Vector Autoregressive Model...... 118 B Bivariate-GARCH Dynamic Conditional Correlation Model..... 119 C Spillovers and lending during the Euro crisis - A linear regression analysis...... 120 3.4 A recursive contract model with spillover costs...... 121 A Model...... 121 3.5 Discussion of model predictions...... 125 A Empirical Validation of Model Predictions...... 126 3.6 Conclusion...... 128 References...... 129 3.7 Appendix...... 131 A Financial linkages across the euro area...... 131 B Theoretical appendix...... 138 C Descriptive statistics from database on loan announcements and disbursements...... 141 vi Essays on Macroeconomics and Development

D Empirical analysis...... 145 E Spillovers and loan conditions...... 150 F The eﬀect of announcements on spillovers - Summary tables.... 152 G The eﬀect of announcements on spillovers - Country details..... 153

vii

Introduction

This thesis is comprised of three independent chapters. In the first chapter titled "Land rights and risk-sharing in rural West Africa", I study the interaction between informal risk-sharing arrangements and land reforms that aim to establish firm individual land rights. The study is motivated by a statutory land reform that took place in Burkina Faso in 2009. The newly introduced legal framework allowed rural population of Burkina Faso to obtain a certificate of land ownership upon community’s approval, proving its concern about customary land tenure regimes across rural population. However, data from the World Bank’s living standards measurement study in 2014 exhibit an extremely low level of certified ownership rights across Burkina Faso.

Firstly, I tackle the problem theoretically. I build a model of one-sided limited commitment which includes consumption transfers and periodical land re-allocation among agents. This structure captures the main components of an informal risk-sharing arrangement in agricultural communities that share consumption and land. Such informal contracts rely on voluntary participation thus they are aﬄicted by limited commitment frictions. Any member of the community can break social norms and deviate to autarky without participating to any production or land sharing. Consequently, a land reform that provides the so called “assurance eﬀect” of property rights, meaning that it only ensures the land holder of perpetual ownership, increases individual’s incentives to deviate from collective arrangements.

The ﬁrst result of the theoretical analysis is that the introduction of the land reform increases the value of the outside option for participants in mutual insurance. Now autarky appears more attractive due to land ownership. This impacts heavily the functioning of the informal contract. It increases the surplus that a participant can extract from community’s pool of resources at the cost of risk-sharing.

The second result relies on the endogeneity of the outside option. The community responds to the increased incentives of the participants by misallocating land. The theoretical result delineates a non-monotonic and concave relationship between productivity and

1 Essays on Macroeconomics and Development land size. Thus, in the presence of a land reform, the community allocates more land to relatively less productive individuals and less land to the more productive ones. In this way, the contract overcomes the limited commitment friction by lowering the incentives to deviate for the individuals that are most likely to do so. Essentially, the model predicts a trade-off between output efficiency and size of risk-sharing. The introduction of a land reform induces the community to sustain members’ participation at the cost of inefficient output production.

The work proceeds by validating the misallocation result empirically. I employ survey data provided from the Millennium Challenge Corporation that closely assisted the implementation of the reform cover agricultural activities before and after the new land law (2009, 2012). I build a productivity measure based on the market value of produced output divided by the size of the cultivated parcel, using crop and unit speciﬁc prices from local markets. I use this measure to estimate changes in land size before and after the reform. Indeed, individuals with initial productivity lower than the village’s median acquire more land relative to the more productive ones. I furtherly perform robustness checks of this result by utilizing a non-linear speciﬁcation by estimating the probability of receiving more land after the reform. The result also points toward that less productive individuals were more likely to obtain more land in the presence of the land reform.

Lastly, I extend the theoretical analysis by studying a normative argument. I do so by examining what would happen in this theoretical context if the land reform was not confined to providing the “assurance effect” but could expand to implement the so-called “collateralization effect”. The collateralization effect refers to the ability of land as an asset to act as collateral thus allowing access to credit markets. In this case, the individual would not only be able to perpetually own land but could also borrow from the credit markets. This addition would result to a race between mutual insurance provided by the informal risk-sharing arrangement and self-insurance provided by the provisions of the land reform.

The corresponding theoretical result suggests that for certain levels of issued debt the self-insurance option is superior to the mutual insurance. The intuition behind this result lies on a characteristic of the informal contract. It does not allow agents to borrow against future income. The participants share risk by allocating consumption units from productive agents to less productive ones in a recursive manner. The option of borrowing is provided by granting access to the credit market allowing for the self-insurance option to substitute eﬀectively for the mutual insurance.

This work provides a reasoning behind the low participation rates in Burkina Faso’s land reform of 2009 as well as it is in line with the well documented empirical fact of land

2 Essays on Macroeconomics and Development misallocation in small size farms in several African countries. Most importantly, it oﬀers a context to study potential interactions between customary institutions and state’s land interventions.

In the second chapter, titled "Land market and mutual insurance institutions in rural Ghana", joint with Karol Mazur we adhere to the same vein of thought but now complement the analysis with (i) varying level of land rights based on intial land holdings and (ii) market-based outcomes that might impact informal risk-sharing arrangements at the community level. In particular, we explore how land security aﬀects mutual insurance and land re-allocations in rural Ghana.

Ghana also abides to the norm of sub-Saharan countries where the institution of local traditional chiefs-leaders who are in charge of resolving disputes, organizing cooperation over public goods and allocating land is still powerful. In absence of formal land markets, land is mostly changing hands within the limits of the village and among members of extended families. In addition, the country exhibits vastly disparate levels of formal land markets.

For the empirical component of the work, we use the household-level Ghana Socioeconomic Panel Survey (EGC and ISSER) containing detailed data on agricultural activities of country’s rural population. To estimate the level of land security in different areas of Ghana, we build a proxy as the share of plots in a given enumeration area that their user holds the right to sell them. Subsequently we use this measure to explore two aspects. Firstly, how it affects risk-sharing through consumption transfers and secondly how it affects land re-allocation within the community.

In the first front, we perform the standard in literature Townsend regressions which examine the elasticity of individual consumption on idiosyncratic income shocks and we find that areas with higher level of formal land markets achieve higher level of risk-sharing. As an additional test, we provide evidence that those communities are able to smooth consumption more effectively. In the second front, we build a land fluidity index which is community specific and is based on the size-difference of land holdings at the household level between the two survey waves available. Using these measures, we find that in areas with higher level of land security, land changes hands more easily. We complement our analysis, by showing a positive correlation between land security and village consumption and productivity.

To further discipline our study, we build a dynamic model rationalizing our empirical ﬁndings. Our village economy features co-operation over insurance transfers and land re-allocations in the presence of voluntary participation (or limited commitment) con-

3 Essays on Macroeconomics and Development straints. We assume that the level of land rights alters the amount of land securely owned by the household. Consequently, this assumption alters the households’ outside option attainable had they chosen to break out of the co-operation with the rest of the community. This component is crucial to understand the nature of informal co-operation patterns under scrutiny as it effectively endogenizes the functioning of informal institutions. For sufficiently low values of land right, the dynamics generated by the model are entirely in line with the empirical evidence presented. As land security increases, land allocations become more efficient, average income and consumption increase, and consumption smoothing against idiosyncratic productivity shocks is improved. However, there exists a level of land rights above which the co-operation over risk and land sharing unravels, pushing the whole society into a bad equilibrium where they can still trade land, but have to rely on self-insurance only. Thus, our model uncovers potentially non-linear effects of land formalization efforts and as such provides a candidate explanation for strong persistence of informal institutions in rural areas of developing countries.

In the third chapter titled “Contagion as a dealmaker? The effect of financial spillovers on regional lending programs” joint with Alica Ida Bonk and Alexandra Fotiou, we study how the degree of financial linkages among sovereign countries affects the conditionality of lending programs.

This work is motivated by the recent European sovereign debt crisis, which gave rise to the critical role of regional lending arrangements in providing financial assistance. A distinct feature of the designed rescue packages for the countries under financial distress was the effective collaboration between international institutions such as the International Monetary Fund (IMF) and regional lending mechanisms such as the European Commis- sion (EC), the European Central Bank (ECB) and the European Stability Mechanism (ESM). A key difference of the lending programs built by the IMF compared to the ones built by the European mechanisms was the difference in their primary objective and the conditionality of the lending agreement. While IMF’s primary objective is to ensure stability and growth to the borrowing country, the Europeans’ primary objective was also complemented by the containment of the crisis, so that it does not spread to other country members. The core research question of the project is to identify whether the risk of financial spillovers gives countries in trouble bargaining power that results in more favorable borrowing conditions.

We first employ empirical methods to identify the extend of financial linkages between borrowing countries and the remaining Eurozone countries. To do so we build a spillover index among Eurozone countries relying on dynamic conditional correlation models and using a range of macroeconomic and financial data. We subsequently build a novel dataset

4 Essays on Macroeconomics and Development by recording press releases, memos and other official statements of EC the ESM and the IMF. Combining those two elements we are able to identify empirically whether announcements on program details affect financial and macroeconomic interconnectedness among the countries involved. Our results suggest that in the case of the European institutions, announcements are negatively correlated with the level of inter-countries financial spillovers, while in the case of IMF announcements, no significant impact is observed. This result validates that the objective of the European institutions was to assist countries in distress but also contain the spread of the crisis to other member countries.

We proceed by building a theoretical model of lending between a creditor and a borrowing sovereign. The novel feature of the model is the addition of spillover costs incurred by the lender every time the borrower credibly threatens to default. In this way we are able to account for spillovers as a mechanism that can provide increased bargaining power to the borrowing country. We then provide a comparison between this lending program to a benchmark model which does not include spillover costs. This comparison captures the difference between loans stemming from European mechanisms and loans that stem from the IMF. When accounting for spillover costs, arising from the borrower to the lender, we find that it is in the lender’s best interest to back- load consumption by giving more weight to future transfers in order to reduce contagion cost. This outcome suggests different loan terms in lending agreements within a union with strong links between the borrower and the lender, compared to lending agreements with an external institution not exposed to the borrower.

The theoretical result is in line with the observation in the literature suggesting that European institutions provided loans with longer maturities compared to those provided by the IMF. However, we also test the model’s predictions by constructing a measure of current consumption by calculating the ratio of cumulative disbursements and the total loan amount promised by the European lending mechanisms. Relating this measure to the spillover index we infer that ﬁnancial linkages are negatively correlated to current transfers received by the borrower and consequently positively correlated to future transfers, a result consistent with the back-loading of consumption predicted by the theory.

5 Chapter 1

Land rights and risk sharing in rural West Africa

6 Essays on Macroeconomics and Development

1.1 Introduction

Despite harsh climatic conditions of the West African Sahelian zone, agriculture in Burk- ina Faso constitutes one of the major economic activities. It contributes almost one third of the Buriknabé GDP while it employs more than 90% of the country’s labour force (FAO., 2012). The agricultural sector in Burkina Faso is predominantly small-scale and organized towards subsistence farming. As such, harvest ﬂuctuates due to weather conditions, pests but also various idiosyncratic risks. Village communities in rural West Africa adapting to this high risk environment, have devised mechanisms of mutual insurance. Those informal arrangements are mostly comprised of borrowing and lending and gift- giving in consumption units as to provide a safety net around the community members (Platteau, 1991).

At the same time, land tenure in Burkina Faso is mostly governed by ethnic, cultural and customary rules. Traditional leaders’ influence over land is deeply rooted in pre-colonial institutions of Burkina Faso and those figures maintain extensive power despite multiple governmental interventions. Traditional norms of land management depart from the no- tion of private property. Land is largely considered to be inherited from tribal ancestors and every member of the tribe, clan, lineage is entitled to use it. Land management is coordinated by a land chief, a religious figure who ensures righteous land allocation. Thus, access to land is reassured in the context of the community, but not formally.1 Household farms perceive their land as their own even without an official certificate of land ownership issued by a state authority. Since agricultural activity in rural West Africa is usually confined within the boarders of the village with production aimed mostly for own consumption, land ownership is adequately recognised at the community level.

Economic literature has long emphasized the critical role of strong property rights in economic growth. The main benefits from individual ownership can be summarized into three broad categories (Bambio and Agha, 2018; Brasselle et al., 2002). The “assurance” effect would provide the necessary incentives to productively invest to land, since land as well as investments’ returns are secured by the rights’ holder. The “transferability” effect would allow more efficient land allocation through rentals and purchases (Alchian and Demsetz, 1973; Bambio and Agha, 2018).2 The “collateralization” effect would allow the owner to pledge the plot as collateral and thus gain access to credit (Feder and Nishio,

1This is also mirrored in the the vast absence of official ownership documents in Burkina Faso (figure B.1) 2A subtly different but linked to the “transferability” effect is the “realisability” effect of property rights Brasselle et al.(2002). This refers to the ease with which property can be converted into liquid assets through sales.

Chapter 1 7 Essays on Macroeconomics and Development

1998). However, existing theory overlooks potential interactions with traditional land management remotely linked to the western canon of individual ownership. Pre-colonial institutions at an ethnic level strongly live up until today in rural Africa, creating a conﬂict between institutional, statutory interventions and customary norms.

In this paper I argue that land reforms introduced by governmental authorities constitute an antagonistic mechanism to traditional land arrangements in rural communities. I build on the premise that reforms attempting to render land as privately owned, interact with local mutual insurance networks developed among neighbours, relatives and extended families. Those networks utilise land not only as the main (and often the sole) production factor but also as a major mean of risk-sharing. As a result, when reforms solely focus on the “assurance” eﬀect without granting access to land and/or credit markets, while their application locus is village economies with low level of mechanisation and scarce investment opportunities, state interventions severely disrupt their functioning.

To theoretically account for mutual insurance and land reform as competing forces, I employ a model of constrained optimal risk-sharing under limited commitment. A principal, head of the community and an agent, a farming household, engage into reciprocal state contingent transfers of consumption and land allocation. While the principal is fully committed, the agent can renege the contractual agreement at any point in time. Within this framework, I model the agent as a small agricultural household, which uses land as a production factor. In order to accurately trace the practices of African communities, I allow the fraction of land allocated to the household to be decided by the principal. This feature is consistent with the practice of periodic land redistribution observed in rural Burkinabé villages and renders land as an additional insurance mechanism co-existing with consumption transfers inside the contract.

Agent’s outside option is the interaction channel between government’s land intervention and community’s risk-sharing. In the presence of a land reform and voluntary participation, the outside option of the household is to register the fraction of land that was lastly allocated within the contract. This distorts the incentives of the principal to allo- cate land according to idiosyncratic productivity in order to sustain the contract. The antagonistic force stemming from the existence of a land reform entails eﬃciency costs on the functioning of the informal risk-sharing mechanism.

The case of Burkina Faso constitutes an illustrative example of a state that implemented a land reform to establish individual property. The reform was introduced with the enactment of a truly innovative and inclusive rural land law (034/2009), allowing individuals to register their plots and obtain a certiﬁcate of ownership. The legislation was followed by an extended eﬀort to disseminate information about the formal procedure to

8 Chapter 1 Essays on Macroeconomics and Development be followed. Government’s plan to reform land management was closely assisted by the Millennium Challenge Corporation, which was actively engaged in all stages of implementation. Regarding the evaluation of the success of the Burkinabé plan, the results were not as expected. The number of approved land registrations and the number of agricultural households receiving certiﬁcates of ownership recognition were far below the set targets. Indeed, according to the United States Agency for International Development (USAID, 2017), land management almost a decade after the enactment of the law keeps on being under customary norms and community control.

To empirically validate the theoretical analysis, I employ survey panel data which were collected and used by the Millennium Challenge Corporation to assess the progress of the rural land law in 2009. The dataset gives a complete picture of agricultural activities in rural Burkina Faso at the plot, individual and household level and provides detailed information on land use. I am able to track plots across time and test the land allocation mechanism with respect to farmers’ productivity. Indeed, in treated villages, where the Millenium Challenge Corporation eﬀectively informed rural population about the up- coming legal initiative, land is ineﬃciently allocated compared to control villages where information dissemmination did not take place.

I extend the analysis by studying what would happen if the land reform could expand to provide the so-called “collateralization eﬀect”, by allowing land to act as a guarantee to take up a loan, thus granting access to the credit market. In this case, a race between mutual insurance and self-insurance through the credit markets would emerge. The corresponding theoretical result suggests that for certain levels of issued debt the self-insurance option is superior to the mutual insurance. The community insures against risk by allocating consumption units from productive to less productive agents in a recursive manner, however, borrowing against future income is not possible. Thus the self-insurance option can potentially substitute and crowd out mutual insurance.

The paper unfolds as follows. In section 1.2, the related literature is presented. In section 1.3, the background of the 2009 land reform in Burkina Faso motivates the study. In section 1.4, the theoretical model of a second generation optimal contract with limited commitment is presented, formalising the main mechanism. Section 1.5 presents empirical regularities related to the main pillars of the theoretical model and overviews the dataset in hand, while section 1.6 performs the empirical analysis to test the theoretical predictions. Section 1.7 extends the analysis to the “collateralization” eﬀect illustrating the change in the model’s dynamics and lastly section 1.8 concludes.

Chapter 1 9 Essays on Macroeconomics and Development

1.2 Literature Review

A large strand of literature advocates the importance of property rights in economic development. De Soto(2001) explicitly stresses the importance of property rights in alleviating poverty. He considers secured property rights as the means to higher investment, easier access to credit and higher surplus value creation. Deininger et al.(2003) highlights the key role of land policies for the developing world where large population increases and thus scarcity of land demands for higher land security. Besley and Ghatak(2010) build a sound theory of the multiple channels through which property rights aﬀect economic activity.

However, the clear positive stance towards property rights comes into question in the case of developing countries. Two characteristics inhibit formal individual property from unfolding its benefits. First is the weak institutional framework. Weak institutions regarding the enforcement of property rights (Bellemare, 2010) or the marekts’ functioning (Besley et al., 2012) can render formal land titling moot.3 The second reason is customary arrangements that can effectively substitute for the lack of formal ownership (Chari et al., 2017; Gollin and Udry, 2019; Besley, 1995).4 Bruce(2012) argues that within a bundle of customary behavioural codes, land ownership can be fairly well established. At the same time, those customary land tenure regimes exhibit an impressive degree of flexibility and adaptation (Bruce et al., 1994; Platteau, 1996). On the other hand, ambiguous property rights might disproportionally benefit people high in local hierarchy (Goldstein and Udry, 2008) or fail to grant tenants and small scale farmers with the appropriate bargaining power (Banerjee et al., 2002).

This work contributes to this literature by modelling the interaction between informal risk-sharing arrangements as those observed in rural areas of developing countries and a state intervention which aims to establish individual property. Thus, it oﬀers a detailed description of the channel in which land reforms can aﬀect customary institutions of land allocation.

The present study is also related to the literature of resource misallocation in the developping world. Institutional pluralism regarding land rights undoubtedly exerts a negative

3Bellemare(2010) juxtaposes the effects of de jure and de facto property rights on agricultural productivity in Madagascar and finds null results for the former category due to poor enforcing institutions. Besley et al.(2012) challenging the magic bullet nature of the de soto arguement assert that the positive effect of property rights of granting access to credit depends largely on competition in this market 4This view of de facto ownership emerging through the functioning of informal institutions is consistent with the idea of endogenous property rights, an idea dating back to Demsetz(1974). Feder and Feeny(1991) present a detailed overview of the evolution of land rights by outlining the emergence of property in parallel with the development stage of the agricultural sector.

10 Chapter 1 Essays on Macroeconomics and Development impact on the functioning of land markets. Restuccia and Santaeulalia-Llopis(2017) ﬁnd extensive factor misallocation in Malawi and attribute it to customary institutions governing land allocation. Adamopoulos and Restuccia(2019) study a reform of land redistribution in Philippines that prohibited land market transactions causing a significant decrease in agricultural productivity. Chari et al.(2017) and Chen et al.(2017) document signiﬁcant gains in land productivity from land reforms allowing land leases in China and Ethiopia respectively. This work presents land misallocation as the response of informal institutions of land management to implemented land reforms. It shows that misallocation can emerge in the transition from communal informal land governance to formal individual land ownership.

The present paper links the concept of property rights to the distinct characteristic of risk- sharing in rural communities. The pattern of village economies engaging into transfers of consumption units to tackle adverse shocks has been well documented in the literature. Townsend(1994) explores the magnitude or risk sharing in Indian villages. He finds a substantial flexibility from the side of community to adapt to adverse shocks, concluding that the assumption of collective insurance in village communities is not absurd. The form of transfers among the members of the community is studied by Platteau and Abra- ham(1987) and Udry(1994) who find that loans can actively serve as a risk-insurance mechanism.

The literature has also evolved on the theoretical front by building models of mutual insurance applied in village economies, under diﬀerent types of frictions. Ligon et al.(2002) and Attanasio and Rıos-Rull(2000) build on the model of risk-sharing with limited commitment in order to explore the imperfect insurance observed in village economies as well as the eﬀects of voluntary participation on the resulting allocation. Models of collective insurance have been employed to study various social phenomena in the developping world. Morten(2019) and Munshi and Rosenzweig(2016) study temporary migration of population with respect to risk-sharing in rural villages.

The paper in hand provides a novel contextualisation of property rights and risk insurance mechanisms in small agricultural communities as competing mechanisms. It traces the transmission of land reform’s eﬀects on risk-sharing contracts through the increase of the outside option. This results in land reform jeopardising mutual insurance by increasing the bargaining power of the individual within the community. In order to preserve risk- sharing, the community responds by misallocating land as to decrease deviation incentives from the informal contract.

Chapter 1 11 Essays on Macroeconomics and Development

1.3 Background on the land reform in Burkina Faso

A motivating example for the present study is the case of Burkina Faso, a landlocked country in the Western Africa’s Sahelian zone. The economy of Burkina Faso is mostly based on agriculture (ﬁgure B.2), with a recent increase of mining activities due to a gold mining boom in 2009-2010. The vast majority of working population is engaging to rural activities (90%). The predominant form of agricultural production is small-scale farming, managed by members belonging to the same lineage or family (USAID, 2017).

Land use in Burkina Faso faces considerable problems mainly due to rapid increase of population fuelling competition for available land, high internal migration and climate change. While those threatening factors are in place, land tenure security scores are at a record low relative to other African countries (ﬁgure B.4). After independence in 1960, management of Burkinabé land was following entirely customary norms with the government only managing protected areas (Ouédraogo, 2002). The concept of private property over land appears in 1984 with the introduction of Réorganisation Agraire Fon- cière (RAF). This legislation granted all land to the state in an attempt to disrupt the control of traditional chiefs over land, and allowed rural population to gain access to land following government’s rules (Hughes, 2014). Amendments of this law (1991, 1996) introduced a type of private ownership through granting user-rights over plots of land.

1.3.1 Loi 034/2009

Much legislative progress has been achieved since the 1980s regarding land tenure. In 2009 Burkina Faso adopted an inclusive piece of rural land tenure legislation (Loi 034/2009 ). The law’s application locus was rural areas and aimed at equitable access to land, en- hancing productivity, sustainable management and social peace (Article 1 Loi 034/2009 ). The legislative procedure was preceded by the establishment of the National Committee for Secure Land Tenure (CNSFMR) under the Ministry of Agriculture aiming to coordi- nating rural land policy reform. The plan’s most striking characteristic was inclusiveness, in terms of reconciliation between statutory land management based on national laws and customary land tenure referring to local norms. Rather than alienating all informal land practices, it integrated them in a formal national legislation.

In the attempt of introducing, implementing and monitoring the new legislation the Burk- inabé government was assisted by the Millennium Challenge Corporation (MCC). This partnership led to a 5-year compact plan (2009-2014) of $58 million under the title Rural Land Governance Project (RLG) (section 1.5). Three activities took place under the Ru- ral Land Governance plan. The ﬁrst activity comprised of legal and procedural changes

12 Chapter 1 Essays on Macroeconomics and Development and dissemination of the details on the new legislation to rural communities. Activity 2 focused on developing the necessary institutional changes and capacity building, while activity 3 performed site-speciﬁc land tenure interventions (IMPAQ, 2015).

1.3.2 Rural Land Certiﬁcate of Possession (APFR)

The aforementioned inclusive character of the 2009’s land reform was reflected in the capacity provided to individual farmers of issuing the so called Rural Land Certificate of Possession (Attestation de Possession Foncière Rurale, APFR). Articles 36-50 of the 039/2009 law outlines the procedures to be followed for the issuance of the APFR. The predominant characteristic of the APFR is that the community in which the individual, requesting the certificate, belongs to is strongly engaged in the procedure and has the capacity to veto it.

The APFR can be issued to either individuals or collective associations. The issuing period is 75 days conditional on no objections being raised by the community. Essentially, the community has to approve the request of the certiﬁcate before it is granted. The cross checking that the referred parcel does not belong to another individual is made with the direct involvement of the customary and traditional authorities (Hughes, 2014).

The APFR differs from full land title on the capacity that grants to the holder regarding sale of the allocated parcel. Productive use of land which can lead to profiting out of it is allowed, however, sale of the parcel to a third party is forbidden. Transfer of the certificate to members of the same family is allowed with no additional cost (Article 47, Loi 034/2009). Moreover, APFRs may be used to obtain bank loans depending on the bank’s requirements (Hughes, 2014).

1.3.3 Assessment of the results of the RLG

However inclusive and innovative the land tenure legislation was, its results concerning grant of private ownership were not as expected. The Millennium Challenge Corporation (MCC), the organisation responsible also for the monitoring and the implementation of the new legislation in close collaboration with the Burkinabé government, issued reports on the progress of the programme. *0

In table 1.1 the results after the end of the 5-year plan are presented regarding the issuances of the APFRs. The diﬀerence between the actually achieved numbers and the targets set by the MCC is substantial. A little more than one third of the set target of

0Even though by the end of the compact the target of 6000 APFRs approved by local authorities was not met, the MCC asserts that the project resulted to 13,447 ﬁled applications for APFRs

Chapter 1 13 Essays on Macroeconomics and Development

Actually achieved Indicators Target (%) (July 2014) Number of APFRs approved by the 2167 6000 36.1% local government Number of HHs 403 3000 13.4% receiving APFRs

Table 1.1: Results from Land Reform in Burkina Faso. Source: Millenium Challenge Corporation - Measuring Results of Rural Governance Programme (Millenium-Challenge-Corporation, 2016)

APFRs were approved by the local government, while the number of households actually receiving APFRs is a little more than one tenth of the target. Along these lines, the United States Agency for International Development (USAID) in its report on Burkina Faso in 2017, explicitly states:

Although the 2009 Rural Land Law and the 2012 RAF provide the mandate and mechanisms to formalize and secure a variety of tenure types in rural Burkina Faso, most rural land continues to be governed according to customary, informal rules. (USAID, 2017)

In order to examine further the result of the land tenure reform in the region, I use survey data from the World Bank and in particular the Burkina Faso Enquête Multisectorielle Continue 2014 which belongs to the collection Living Standards Measurement Surveys (LSMS).5 The study was conducted between 2014 - 2015 (5 years after the introduction of the reform) and it is nationally representative. Among many survey units there is the module referring to parcels which includes questions on the cultivating land each household holds. In figure 1.1, the responses to the method of land security are presented. It is striking that the option "Land Title" which would correspond to an APFR is only answered by 177 respondents. From figure 1.1, it is apparent that the predominant land tenure regime is the “Possesseur Terrien”, which represents all native people that have inherited land from their family (Ouédraogo, 2002). The second most answered option is “None” indicating a complete absence of any official document certifying ownership.

Economic theory has long advocated the benefits from establishing strong individual property rights. However, in the case of Burkina Faso, a puzzling phenomenon is observed. People are offered the opportunity to officially register their land plots, however they

5Institut National de la Statistique et de la Démographie. Enquête Multisectorielle Continue (EMC) 2014. Ref. BFA-2014-EMC-v01-M. Downloaded from https://microdata.worldbank.org/index.php/ catalog/2538/get-microdata

14 Chapter 1 Essays on Macroeconomics and Development

Figure 1.1: Burkina Faso Enquête Multisectorielle Continue 2014 - LSMS - World Bank choose not to or they are prevented by local authorities. The reasoning behind this observation lies on the core of the present study. The premise which the theory builds upon is that land reforms introduced by the state act as a competing mechanism to the risk-sharing network developed in a community level.

1.4 One-Sided Limited Commitment with land re- allocation

The theoretical part of the present study models the functioning of risk-sharing informal contracts among members of rural communities and their interaction with land reforms when customary land re-allocation is in place. To motivate the assumptions of the model I need to deﬁne certain customary aspects of the social structure in rural communities of Western Africa.

Customary land management in Burkina Faso is generally considered homogeneous. A predominant social ﬁgure at a community level is that of the land chief (chef de terre).6

6The predominance of the land chief can be seen in ﬁgure B.3

Chapter 1 15 Essays on Macroeconomics and Development

The land chief is a religious ﬁgure with legal power and has the complete control over land on behalf of the community (Ouédraogo, 2002).7 One of the main duties of the land chief is the periodic redistribution of land. This land re-allocation takes place among the members of the same community/village but also to foreigners in case they arrive. This practice aims at preventing the creation of monopolies in land-use or underuse of land plots. The periodic redistribution of land is decided upon the needs of the members of the community.

In the theoretical model presented in this section, the land chief is the principal of the risk-sharing contract. The informal arrangement not only prescribes consumption units allocation among community members but also allocates land among the members. The ultimate target of the model is to trace the interaction between those two components and its welfare implications.

The theoretical framework presented here attempts to shed light on the diverse views expressed regarding the land regime policy that should be followed in the African continent. Illustrative of the diversity of the land policy in Africa is the position that the World Bank has held. During the mid-1970s the World Bank was advocating a firm regime of strong individual property rights in Africa. It was persuaded by most of the literature’s theoretical arguments relating land tenure security and agricultural productivity (Udry, 2011). However, this stance evolved over time, resulting to the adoption of a more favourable view towards customary land tenure systems. The flexibility and efficient adaptation of indigenous land sysytems were appreciated (Migot-Adholla et al., 1991).

The contract prescribes the pooling of all households’ resources in the hands of the principal who allocates consumption back to them. The principal after allocating consumption, 1 invests the remainder outside the village at a risk free rate R = β , where β is the common to all discount factor. The principal is the only one that can borrow and lend resources outside the community, the households rely only on the risk-sharing mechanism. The environment builds on Ljungqvist and Sargent(2000).

The community is consisting of a large number of villagers with the preferences over consumption. ∞ X t E−1 β u(ct) t=0 where u(c) is increasing and strictly concave and β is the common discount factor (β ∈ ∞ (0, 1)). Each villager receives a stochastic idiosyncratic productivity each period ({zt}t=0).

Idiosyncratic productivity is iid with P rob(zt = zs) = Πs, with s ∈ {1, 2, ..., S} satisfying

7The land chief is considered to be descended from lineage of the group of the ﬁrst occupants of the earth.

16 Chapter 1 Essays on Macroeconomics and Development

the property, zs < zs+1.

The villager is considered as a small agricultural household which produces output using a fraction of land as the primary production function. The technology is modeled as follows:

ys = zsf(κs¯l) where zs is the idiosyncratic productivity, ¯l denotes land, which is in ﬁxed supply normalized to 1 and κs is the variable of interest. It is the fraction of land that each period the principal decides for the villager to productively use it (κs ∈ [0, 1]). κs eﬀectively captures land re-allocation as a mechanism of risk sharing. Technology f(.) is increasing 0 00 in the fraction of land, κs (f (.) > 0), strictly concave (f (.) < 0) and I assume that with no land there is no produced output f(0) = 0.

Participation of the household to the community risk sharing mechanism entails transfers towards and from the community. The budget constraint of each individual household is:

cs = ys + τs, ∀i ∈ N

If τs > 0 then the household is receiving transfer from the community which adds up to the produced output, while if τs < 0, the household is rendering part of its output to be granted as transfers to other members of the community.

The land chief (principal) maximizes her stream of profits, which consists of the contem- poraneous difference between the pooled output and the consumption allocation, and the discounted future profits stream. In a recursive form, the objective function is

S   X P (v) = max Πs(ys − cs) + βP (ws) {c ,κ ,w } s s s s=1 or equivalently substituting the villager’s budget constraint : S   X P (v) = max Πs(−τs) + βP (ws) {τ ,κ ,w } s s s s=1 where v is the expected discounted future utility previously promised to the villager and ws is the promised value with which the agent will enter next period, given that zt = zs.

In the absence of commitment frictions the economy reaches its ﬁrst best.

Proposition 1: Given a promised utility v, the ﬁrst best allocation satisﬁes the following properties. Consumption and promised utility are constant and equal to the levels cfb(v) and wfb(v), while κfb is constant at its maximum level.

Chapter 1 17 Essays on Macroeconomics and Development

Proof: See Appendix

In the case of a commitment friction, while the head of the community is committed to the agreement, the villager is not. However, what fundamentally changes is the outside option of the villager. The primary channel of interaction between the land reform that aims to establishing strong individual property rights and the contractual agreement among community members emerges through the workings of the outside option. Assumption 1 deﬁnes the rationale behind its formulation.

Assumption 1: The land reform allows the agent-villager to register the fraction of land she was last allocated with, inside the contract.

Assumption 1 determines the form of the outside option of the agent-villager.

aut u(zsf(κs)) + βv (κs)

First, notice that the fraction of land allocated to productive use is endogenous and it is determined within the contract. Second, due to the limited commitment friction, the agent-villager can leave the contract at any state. If she does so, due to the existence of a land reform, she can register the last allocated fraction of land (from within the contract) as individual property.

The continuation value of autarky takes the following form:

∞ S aut X t X v (κs) = β Πru(zrf(κs)) t=0 r=1

Notice that the level of fraction of land is constant and equal to what was last decided within the contract.

The participation constraint of the contract takes the form:

aut u(cs) + βws ≥ u(zsf(κs)) + βv (κs)[PC]

The head of the community is choosing consumption allocated to the agent-villager, fraction of land and promised utility, in order to maximize her stream of proﬁts.

X P (v) = max Πs (zsf(κs) − cs) + βP (ws) {c ,κ ,w } s s s s∈S where v is the promised utility that agent-villager enters the current period with and carries all past histories, in order to recursify the problem.

18 Chapter 1 Essays on Macroeconomics and Development

The maximization problem of the principal takes the following form:

S X P (v) = max Πs (zsf(κs) − cs) + βP (ws) {c ,κ ,w } s s s s=1 S X Πs{u(cs) + βws} ≥ v [PKC] s=1 aut u(cs) + βws ≥ u(zsf(κs)) + βv (κs)[PC] ∀s  κs ≥ 0 κs ∈ [0, 1] =  1 − κs ≥ 0

aut ws ∈ [v , v¯]

Proposition 2: For a given promised utility v, when the participation constraint is non- binding, the consumption and promised utility allocations are constant and equal to cs = g1(v) and ws = v, while the fraction of land reaches the ﬁrst best (κs = κmax). When the participation constraint binds then consumption, promised utility and fraction of land satisfy equations 1.1, 1.2 and 1.3.

0 u (cs)[θ + φs] = 1 (1.1)

0 P (ws) = −(θ + φs) (1.2)

0 1 1 β 0 u (zsf(κs)) = − Eru (zrf(κs))zr (1.3) φs zs 1 − β

Proof: See Appendix

From relations (1.1) and (1.2) the following is derived:

0 1 u (cs) = 0 P (ws) which states that the villager’s marginal rate of substitution between current consumption and promised utility should be equal to the land chief’s marginal rate of transformation.

The eﬀects of the land reform can be seen in relation (1.3). With certain manipulations, equation (1.3) reads:

0 0 0 ∂vaut(κs) zsf (κs) = u (zsf(κs))zsf (κs)φs + φsβ | {z } | {z } ∂κs Marginal Increase in Agent’s Marginal Beneﬁt | {z } Land Chief’s Revenues from deviation Land Reform’s Intertemporal eﬀect

Chapter 1 19 Essays on Macroeconomics and Development

Which shows that the principal when optimally choosing the fraction of land to be allocated to the agent, she equates her marginal benefit to the villager’s marginal benefit. The principal’s marginal benefit constitutes the marginal increase in her revenues. On the right hand side of the equation the marginal benefit of the villager appears as the sum of the intratemporal marginal benefit from deviating at the current period plus the intertemporal effect of the land reform due to the assurance effect. In particular, the derivative of the continuation autarky value that the farmer attains if she deviates from the contract with fraction of land lastly allocated to her (κs) takes the following form: aut ∂v 1 0 0 = f (κs) Eru (zrf(κs))zr ∂κs 1 − β | {z } ω>0 It is worth noticing that in the expression of the continuation value in the outside option, idiosyncratic productivity (zr) and fraction of land (κs) have different subscripts denoting that once reneging the contract, land ceases adjusting to produtivity shocks, turning the ω term dependent only on the probability distribution of idiosyncratic productivity. This stems directly from the functioning of the assurance effect. The land reform ensures ownership of land but nothing more.

The model as delineated above presents an interesting trade-oﬀ which encompasses the core interaction between land rights and risk sharing when seen as competing mechanisms.

Notice that the level of fraction of land (κs) has two opposing eﬀects on the model. First it raises the revenues of the community. This can be seen from the objective function of the principal-head of the community. A higher level of κs will increase the produced output for a given realisation of zs and consequently the size of the pie to be allocated among consumption to households and proﬁts for the principal. At the same time, κs is on the right hand side of the participation constraint. A higher fraction of land allocated to the villager makes the outside option more attractive, increasing deviation incentives.

In order to characterize the nature of the land tenure system under the contract in the presence of a land reform as an outside option, I deﬁne the following possible land regimes:

Deﬁnition: A land regime is productive if it adjusts fraction of land positively to id- ∂κs > rigid iosyncratic productivity ( ∂zs 0). It is if it does not adjust fraction of land to ∂κs counter productive changes in idiosyncratic productivity ( ∂zs = 0) and it is when it ∂κs < adjusts fraction of land opposite to idiosyncratic productivity ( ∂zs 0).

By manipulating the optimality condition with respect to fraction of land, I can obtain an optimal response of the κs to realisations of idiosyncratic productivity. 0 0 0 ∂vaut(κs) H(zs, κs) = zsf (κs) − φs u (zsf(κs))zsf (κs) + β ∂κs

20 Chapter 1 Essays on Macroeconomics and Development

Applying the implicit function theorem on function H(zs, κs), I can derive a relation between fraction of land and idiosyncratic productivity which leads to proposition 3.

∂H(zs,κs) ∂κs = − ∂zs ∂H(zs,κs) ∂zs ∂κs

Proposition 3: For a given v and for each s ∈ S that leads to a binding participation ∗ constraint, there exists threshold zs which determines the nature of the land regime under the contract.

Land Regime zs 0 ∗ ∂κs ∗ 1−φsu (zs f(κs)) Productive > 0 zt > z = 00 ∗ ∂zs s φsu (zs f(κs))f(κs) 0 ∗ ∂κs ∗ 1−φsu (zs f(κs)) Rigid = 0 zt = z = 00 ∗ ∂zs s φsu (zs f(κs))f(κs) 0 ∗ ∂κs ∗ 1−φsu (zs f(κs)) Counter Productive < 0 zt < z = 00 ∗ ∂zs s φsu (zs f(κs))f(κs)

Using the following functional forms for utility and technology that satisfy the conditions on monotonicity and concavity,

(1−α) cs u(cs) = 1 − α 1−γ ys = zsf(κs) = zsκs the above proposition takes the following form:

Land Regime zs ∂κ ∗ 1 −(1−γ) Productive s > z < z − α φ α κ ∂zs 0 t s = [(1 ) s] s ∂κ ∗ 1 −(1−γ) Rigid s z z − α φ α κ ∂zs = 0 t = s = [(1 ) s] s ∂κ ∗ 1 −(1−γ) Counter Productive s < z > z − α φ α κ ∂zs 0 t s = [(1 ) s] s

Proof: See Appendix

The result from proposition 3 illustrates the variability of the customary land tenure regime. In the presence of a land reform as an outside option, the principal responds strategically to the allocation of land to the agent such that to keep the contract sustainable at all times. This means that given an allocation of consumption, promised utility and a realisation of idiosyncratic productivity, the contract might optimally adjust fraction of land downwards, upwards or not at all. This is due to the strategic way of the principal to enforce contract participation. The land chief when proceeding to redistribution of land weighs those two opposing eﬀects. How much allocated land, increases

Chapter 1 21 Essays on Macroeconomics and Development the size of the pie (her revenues) and how much the incentives of the villager to deviate. This essentially depends on how close to a realisation of productivity that would lead to a binding participation constraint the current idiosyncratic productivity is. This is when the threat of reneging the contract from the side of the villager becomes credible.

This strategic behavior regarding allocation of land, entails efficiency costs. In the absence of the limited commitment friction, the incentives of the principal would be in line with a flexible land tenure regime. A flexible land tenure regime would increase principals revenues and would increase the size of the pie to be distributed among the members of the community. A land reform distorts those incentives, and induces a strategic allocation of land, which might lead to productive villagers being allocated smaller fraction of land, due to the threat of deviating from the contract.

1.5 Data from Burkina Faso

1.5.1 Rural Land Governance Project

The empirical analysis is employing the Millennium Challenge Corporation (MCC) compact with the government of Burkina Faso. The ultimate aim of this project was alle- viation of poverty by boosting economic growth. This 5-year plan, agreed in July 2008, consisted of four distinct projects aiming at diﬀerent targets. The rural land governance (RLG) project, the agricultural development project, the roads project and the Burkinabé response improvement of girls’ chances to succeed to schools projects (BRIGHT II).

The present work focuses on the ﬁrst project, the rural land governance. The motivation of the project was the pervasiveness of land conﬂicts due to scarcity of land resources and tension between statutory laws and customary norms regarding land tenure. Its primary target was to establish a legal framework through which rural population could gain easier access to local land governance and administration.

The RLG consisted of three main activities implemented in a sequential manner. The ﬁrst activity focused on the legal and procedural change and communication. The second addresses the institutional development and capacity building and the third attempted site-speciﬁc land tenure interventions (table 1.11).

The time span of the compact was 5 years, from 2009 to 2014. The project was divided in two phases in which the prescribed activities took place sequentially. Phase I of the programme lasted from 2009-2012. This phase focused on 17 pilot communes, where it implemented activity 1’s plan and started implementing the actions described in activity 2 and 3. In Phase II the implementation of the plan was extended to 30 additional

22 Chapter 1 Essays on Macroeconomics and Development communes, reaching a total of 47 communes for which the MCC implemented the RLG project.

1.5.2 Monitoring the progress of RLG project

The MCC assigned the evaluation of the project to an independent organization, IMPAQ. The evaluation consists of collection of survey data from the 34 communes, 17 of which participated as treated or ans 17 as control areas during phase I (ﬁgure B.5). The survey is divided in baseline and interim, which refer to pre-reform and post-reform time periods respectively. However, the interim survey is conducted at the end year of phase I, so it does not capture the eﬀects of phase II activities. As a result, only the legal initiation of the reform, the dissemination of information regarding this legal option to rural population and some early option of APFR (ownership) issuance is evaluated.

The baseline and interim survey consist of four questionnaires focusing on diﬀerent levels. Household, individual, parcel and production are the topics covered in the questionnaires. The size of the sample is 3,352 households from all 34 communes, accounting for more than 10,000 individuals and more than 6,000 land plots used for cultivation.

1.5.3 Empirical Regularities in Burkina Faso

• Role of traditional leaders

Burkina Faso is characterised by a pluralistic ethnic and religious environment. Apart from Mossi who constitute the major ethnicity of over 50% of the population, there are numerous others (Touaregs, Peuls, Lobi, Gourmantché etc). The existence of strong ethnic and religious norms exerts great inﬂuence on land rights and grants traditional leaders with extensive power over them.

Despite the ethnic and religious heterogeneity, customary land management in Burkina Faso is generally considered to be homogeneous. A predominant social ﬁgure in rural communities is that of the land chief (chef de terre or “Tengsoaba” in Mossi). The land chief is a religious ﬁgure with legal power that has the complete control over land on behalf of the community (Ouédraogo, 2002).8 The land chief is considered to be the intermediator between the living and their ancestors ensuring righteous land management as land is considered to be owned by the ethnicity, clan, extended family but not individually. One of the main duties of the land chief is the periodic redistribution of land. This land re- allocation takes place among the members of the same community/village but also to

8The land chief is considered to be descended from lineage of the group of the ﬁrst occupants of the earth.

Chapter 1 23 Essays on Macroeconomics and Development foreigners in case they arrive. This practice aims at preventing the creation of monopolies in land-use or underuse of land plots. The periodic redistribution of land is decided upon the needs of the members of the community.

Round 4 of the Afrobarometer, a pan-African survey on economic, social and political attitudes, took place in 2008, coinciding with the introduction of the land legislation under examination. The effect of traditional leaders in rural communities’ issues is shown in figure B.6. In 8 out of the 13 administrative regions in Burkina Faso, traditional leaders exert sizeable influence to the community’s functioning.

The Afrobarometer provides more details on the issues that the traditional leaders have a word on. In particular, the survey poses the question of who is primarily responsible for allocating land. In almost all regions there is a non-negligible percentage of people who answer that traditional leaders are responsible for land management (in all regions but Centre-Est, Plateau Central an Est, this percentage exceeds 20%). In six out of the thirteen administrative regions, most people answer as the ﬁrst choice that traditional leaders are the primarily responsible for allocating land (ﬁgure B.7).

• Land

Agriculture in Burkina Faso is predominantly organized at a small scale. The majority of the individuals engaged in farming operate a parcel of at most 2 hectares (ﬁgure B.9) with the average plot size being of 1.5 hectare (table 1.7). Even though households are large in terms of household members, the average household operates approximately 2.5 plots (table 1.8) with the average household’s land holdings covering an area of approximately 4 hectares (table 1.9).

As cultural norms govern the allocation of land within the community, the size of the parcel that an individual cultivates is not written in stone, but changes according to not only the needs of the household but also the needs of the community. This leads to signiﬁcant land reshuﬄing among the members of the same community.

The panel dataset allows tracking of speciﬁc parcels reported in both waves. As a result, it was possible to compare the reported size of the corresponding parcel between the two waves. Figure B.11 exhibits the recorded plot size in the baseline and interim survey. Across all regions of Burkina Faso, very few plots did not experience a change in their size. Most of the plots are either lying closely above or below the 45 degree line signifying a small but non zero size variation.

In addition, land holdings’ variation does not cancel out at the household level. Aggre- gating land size at the household level in ﬁgure B.10 evinces large changes in households’

24 Chapter 1 Essays on Macroeconomics and Development land holdings. This provides an insight on the nature of land allocation. The size of households land holdings is not ﬁxed and land parcels are re-allocated among household members but rather there is land re-allocation across households.

As far as the intensive margin of the plot size variation is concerned, a 10% of the plots in the sample experienced no change. 50% of the respondents reported variation lying in the range from a 18% decrease a percentage increase of 10% . It is worth noticing that the distribution of land size variation does not change when plot size is aggregated at the individual and household level (table 1.10).

1.6 Evidence from the RLG programme in Burkina Faso

The present section attempts to validate the theoretical prediction exploiting the panel data from the RLG project.

The case of Burkina Faso constitutes an ideal application locus for the theory developed in this work. Firstly, the panel dataset allows the comparison between communes that were exposed to the treatment and others that were not (ﬁgure B.5). Secondly, the time gap between the two survey waves is not large allowing us to accurately capture changes in both land and productivity without omitting important events in between 2009/10 and 2011/12. Lastly, the survey design was built with the aim of assessing the impact of the 2009 land reform in its ﬁrst stages, therefore allowing to test the implications of the introduction of a land reform.

The central theoretical result refers to the non-monotonic relation between land allocation and productivity levels in the presence of a land reform. In particular, the testing hypothesis reads as follows:

Testing Hypothesis: In treated areas, individuals exhibiting low level of agricultural productivity are increasing their share of land, while high productivity individuals are decreasing their share of land. In control areas, low productivity individuals are decreasing their share of land, while high productivity individuals are increasing their share of land, since the absence of the outside option leads to an eﬃcient allocation

The testing hypothesis addresses the eﬀect of a land reform on the allocation of land. As already shown, agricultural plots both at the individual and the household level experience signiﬁcant change in their size between the two waves. I exploit this variation for the

Chapter 1 25 Essays on Macroeconomics and Development empirical analysis.

For each household h, the land holdings are deﬁned as:

X hh.landh,t = κp,t p∈h

κp is the size of plot p operated by a member of household h in time t and the total land holdings are the sum of all plots operated by members of the household in each wave.

According to the theoretical part, land allocation is determined by the level of productivity. I build a measure of productivity for each household in the sample based on plot productivity. To allow for eﬀective comparison between diﬀerent crop types, I use the monetary value of sold harvest.

qc,p ∗ pc,i agricultural productivityp = κp where qc,p is the quantity of crops produced in plot p expressed in the local measurement unit. pc,i is the price per local measurement unit for the speciﬁc crops grown in plot p and κp is the size of the plot.

In case the respondent has sold part of the production to the market then the pc,i is her selling price per local measurement unit. However, the majority of the farmers do not sell their harvest but keep it for household-consumption. In cases, in which I do not have actual realised price data at the individual level, pc,i ≡ pc,v where pc,v is the price of the speciﬁc crops calculated as the mean price of all other farmers in the village that have sold the same crop. I generalise geographically in case there is no crop-unit speciﬁc price data by using community (pc,i ≡ pc,com) and country prices (pc,i ≡ pc,bf ). The numerator is expressed in $ while the denominator is expressed in hectares. Hence agricultural productivity is measured in $/ha.

Even with the aforementioned geographical generalizations, imputed prices correspond to a low proportion of plots that reported production data, signifying a very low participation to the goods markets, rendering the corresponding crop-price-unit matching diﬃcult, as shown in table 1.2.

I proceed next in deriving the household productivity. This requires the aggregation of plot productivity at the household level, since the average household operates more than one plots. Therefore, I derive the weighted mean of agricultural productivity for all parcels under household’s use.

For each household h:

κp household productivityh = Σp∈h agricultural productivityp ∗ Σp∈iκp

26 Chapter 1 Essays on Macroeconomics and Development

Production Imputed prices Season (# parcels with (# parcels with available data) imputed data) Dry 2008/09 396 160 40% Rainy 2008/09 9,035 1,814 20% Dry 2009/10 161 100 62% Rainy 2009/10 10,154 2,141 21% Dry 2010/11 629 372 59% Rainy 2010/11 14,748 2,878 19.5% Dry 2011/12 406 294 72.4% Rainy 2011/12 15,521 3,128 20%

Table 1.2: On the second column is the number of parcels for which production data are available, in the third is the number of parcels for which prices per local unit of measure are imputed. The low percentages are due to the diﬀerent local units of measurement used and the low participation to goods market.

Where the weights used in the speciﬁcation are determined by the relative size of the particular plot with respect to the total size of household land holdings.

In the theoretical result, productivity is not set in absolute terms but rather relative. In fact, the corresponding proposition predicts that in the presence of a land reform, less productive household in the community would be allocated more land, while more productive less. As such, the ranking of high and low productivity takes place in the context of the community. In order to capture this empirically, I compute the median village productivity for all villages in the sample. Subsequently, I record whether household productivityh is below the median at the village level through the following dummy variable:

 1 for household productivityh < median village productivityv below village medianh = ∀v, t  0 for household productivityh ≥ median village productivityv

The following speciﬁcation is testing the household land holdings based on household productivity ranking within the community.

hh.landh,t = αh,t + β0 ∗ household productivityh,t

+ β1 ∗ below-village-medianh,t (1.4) + β2 ∗ Xh,t

+ β3 ∗ ζh + β4ζv,t + h,t where Xh,t is a vector of household and land controls. In particular, Xh,t controls for house-

Chapter 1 27 Essays on Macroeconomics and Development holds’ land regime, rights upon land, years of use, irrigation investments and other.9 More- over, ζh controls for unobserved heterogeneity across households, while ζv,t is a dummy variable controling for village and time fixed effects, controling for unobserved heterogeneity across villages. Lastly, in both specifications, standard errors were clustered at the commune level consistently with the sample design of the survey.

Table 1.3: Results from LSDV on regression (1.4)

Dependent variable: Dependent variable:

Household Land Holdings Household Land Holdings Treated Areas Control Areas

(1) (2) below.median 0.699∗ -0.394 (0.415) (0.604) Constant 1.224 -2.890 (5.210) (5.950)

Household Controls Yes Yes Household FE Yes Yes Village+Time FE Yes Yes Commune Clustered SE Yes Yes Observations 1,318 959 R2 0.976 0.984 Adjusted R2 0.381 0.340 Residual Std. Error 2.151 (df = 51) 2.801 (df = 23) F Statistic 1.640∗∗ (df = 1266; 51) 1.527 (df = 935; 23)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Column (1) corresponds to the 17 treated communes and columns (2) corresponds to the 17 control communes. Household productivity has been winsorized at the 2% level. Household productivity is measured in dollars/sq.m

Table 1.3 shows a clear diﬀerence between households residing in treated areas compared

9   share of land in which the hh lives inh,t    share of land within the villageh,t       share of inherited landh,t     share of land given by the chief   h,t     number of years hh uses the landh,t    share of land hh has right to plant trees   h,t Xh,t =    share of land hh has right to lendh,t       share of irrigated landh,t     share of land with conﬂicth,t       hh’s number of plotsh,t     hh’s average age   h,t  hh’s sizeh,t Note that for all variables that needed to be aggregated from the parcel to the household level, aggregation used the weighted mean of all plots under household’s operation, with relative parcel land over total household’s land used as weights.

28 Chapter 1 Essays on Macroeconomics and Development

to those residing in control areas. The βˆ1 estimated coefficient is positive and statistically significant for the low productivity individuals in treated areas (column 1), signifying that in communes where the option of the land reform was disseminated effectively, low productivity individuals acquired more land relative to the high productivity individuals. In particular for a low level productivity individual a 1$/sq.m increase in productivity, corresponds to larger land size by 0.699 hectares compared to the high productivity individuals. On the other hand, in control areas the corresponding β1 coefficient is statistically insignificant, indicating a difference between high and low productivity individuals to be indistinguishable from 0 and if anything has a negative sign.

To further support the empirical results, I estimate a modified specification of equation (1.4). Firstly, I use a fixed effect within estimator and secondly I add the interaction term:

below village medianh,t ∗ household productivityh,t in order to examine how the household’s relative position aﬀects the correlation between acquired land and household’s productivity. Consequently, the speciﬁcation reads:

hh.landh,t = αh,t + β0 ∗ household productivityh,t

+ β1 ∗ below-village medianh,t ∗ household productivityh,t (1.5) + β3 ∗ below village medianh,t

+ β2 ∗ Xh,t + h,t where the control vector remains the same as in speciﬁcation (1.4) and the within estimator takes care of unobserved household and time heterogeneity. To account for aggregate shocks that might have an eﬀect on the relation between land allocation and household’s productivity I add a control variable that records the production of the remaining households residing in the village.

Table 1.4 presents the results from the estimation of speciﬁcation (1.5). Consistently with the previous estimates, the low productivity households in treated areas acquire more land than the high productivity ones. Moreover, a household which is below the median productivity at the village level and experiences an unit increase in productivity acquires more land compared to the high productivity household in the same village. On the other hand, the coeﬃcients of interest in control areas are indistinguishable from zero in control areas.

Chapter 1 29 Essays on Macroeconomics and Development

Table 1.4: Resutls from regression (1.5)

Dependent variable: Dependent variable:

Household Land Holdings Household Land Holdings Treated Areas Control Areas

(1) (2) hh.productivity*below.median 2.793∗∗ 21.350 (1.334) (14.485)

Household Controls Yes Yes Commune Clustered SE Yes Yes Observations 1,318 959 R2 0.429 0.344 Adjusted R2 -4.491 -6.669 F Statistic 7.347∗∗∗ (df = 14; 137) 3.066∗∗∗ (df = 14; 82)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Column (1) corresponds to the 17 communes that were treated areas and column (2) corresponds to 17 communes that were control areas. Diﬀerence in individual land holdings is winsorized at 1% level. Individual productivity is measured in dollars/sq.m

1.7 Collateralization eﬀect

The analysis so far has adopted a positive stance on the study of the interaction between informal risk sharing and statutory land reforms. It examined the response of informal insurance contracts to a land reform that increases the outside option of the household by granting it the simplest and most direct of land rights beneﬁts, the assurance eﬀect.

Already from assumption 1, the effect of the land reform on the contract is that the household is rest assured that the last allocated fraction of land, will remain under its ownership in the future. This effect increased the outside option of the household by giving rise to the limited commitment friction. However, the contract in order to survive responds to this change by misallocating land. In this way, it effectively ensures households’ voluntary participation at a production efficiency cost.

In this section, the analysis proceeds to a normative policy prescription that can inform the design of land reform policies. The main advantage of informal risk-sharing is that it provides partial insurance to its participants. Therefore, the land reform on top of the assurance effect should also guarantee the collateralization effect. The collateralization effect of land rights can grant the affected areas with access to credit, which in turn can provide the formal self-insurance required to substitute for the informal mututal- insurance.

30 Chapter 1 Essays on Macroeconomics and Development

The aforementioned collateralization effect was weakly defined in the 2009 land reform in Burkina Faso, since the APFR could operate as an official document allowing land to be pledged as collateral, but this depended on banks’ requirements (Hughes, 2014).

The reason why the collateralization eﬀect could eﬀectively provide the self-insurance to substitute for the mutual-insurace is based on the functioning of the one-sided limited commitment informal contract. An insightful exposition of the one-sided limited commitment contract’s functioning is provided by Zhang(2013) and summarized below.

ys ys

3 3 c˜ 2 2 c˜ 1 1

Time Time 1 2 3 4 1 2 3 4

(a) Case 1 (b) Case 2

Figure 1.2: Cases of consumption smoothing under one-sided limited commitment. source: Zhang(2013)

In figure 1.2, Zhang 2013 exhibits two cases that reveal the limitations of the one-sided limited commitment contact operations. The economy lasts for four periods. Endowment is not stochastic and takes the following values, y3 in period 1, y1 in period 2, and y2 in periods 3 and 4, with y1 < y2 < y3. The first question is whether consumption smoothing can be achieved between period 1 and period 2. The answer is yes, since the principal can transfer wealth from period 1 (when the maximum endowment is realised) to period 2 (when the minimum endowment is realised) in order to smooth consumption. Note that c˜ could be anywhere between y3 and y1 and would still satisfy the participation constraint in period 2 - therefore making consumption smoothing possible. The following question is whether consumption smoothing can be achieved between period 2 and period 3. The answer depends on the level of c˜ allocated between periods 1 and 2. In case 1 (figure 1.2a) c˜ is higher than y2, the endowment to be realised in period 3. Therefore there is space for transfer of wealth between period 2 and period 3 that still satisfies the participation constraint in period 3. Note however that if c˜ between periods 1 and 2 was set below y2 as shown in figure 1.2b, further consumption smoothing would require transfer of wealth from period 3 to period 2. However, this is not feasible, since the allocated consumption would not satisfy the participation constraint in period 3.

Zhang’s analysis reveals a critical characteristic of the one-sided limited commitment contract. Due to the presence of limited commitment, the contract allows for transfers of

Chapter 1 31 Essays on Macroeconomics and Development wealth from present high endowment states to future lower endowment states, but cannot achieve transfers of wealth from future high endowment states to present low endowment states. In other words, the agent cannot borrow against her future high income.

The claim of the present section is that a land reform that grants access to credit markets, promoting the collateralization eﬀect of land rights, constitutes a superior option for the risk-averse household. The reason is twofold. First, compared to the outside option with only the assurance eﬀect (analysed in section 1.4) it can additionally provide self- insurance options for the household to attain the desired consumption smoothing. Second, with respect to the allocation provided by the contract, issuing debt through access to the credit markets can perform the operation that was not feasible under one-sided limited commitment, namely to allow households to borrow against their future endowments.

To formalize the claim, I depart from the production economy of section 1.4 and focus on stochastic endowment economy in favour of tractability. I compare two versions of this economy, one operating under the one-sided limited commitment informal contract and the second allowing the household to issue debt.

In particular, the two economies under examination are the following:10

OSLC S   X Debt Economy P (v) = max Πs(¯ys − cs) + βP (ws) ∞ {cs,ws} X t s=1 max E0 β u(ct) {c ,d } S t t t=0 X Πs u(cs) + βws ≥ v [PKC] ct + (1 + r)dt−1 = yt + dt s=1 dt+j u(cs) + βws ≥ u(¯ys) + βνaut ∀s [PC] limj→∞ = 0 (1 + r)j cs ∈ [cmin,cmax] ws ∈ [νaut, v¯]

The OSLC economy is the principal-agent dynamic contract analysed in section 1.4 but in simpler form with stochastic endowment and not stochastic productivity. The debt economy represents the household maximizing expected utility under its budget constraint where (1+r)dt−1 denotes debt obligations maturing at time t and dt denotes the acquisition of debt maturing at time t + 1. Moreover the household is subjected to a transversality condition, preventing it from dying holding debt. Preferences u() and endowment yt maintain the same properties as under the contract. So the probability distribution of the stochastic component y is identical under both economies.

10the OSLC economy is expressed in a recursive form to facilitate the implementation of its solution algorithm.

32 Chapter 1 Essays on Macroeconomics and Development

r Pj Proposition 4: For debt levels that satisfy dj ≥ 1+r κ=1[E(yj) − yκ] ∀j ∈ S the debt economy provides a consumption allocation that second order stochastically dominates the consumption allocation provided by the OSLC economy, and therefore is strictly preferred by the risk-averse household.

Proof: See Appendix

Based on proposition 4, land reform policy can be sufficiently informed in order to serve its purpose. The advantages of a land reform can be materialized and supersede those of mutual insurance only by exploiting both the assurance and the collateralization effect. Those two effects combined can provide a superior outside option to the informal contract and achieve the desired results of strong individual property rights.

1.8 Conclusion

The study of the interaction between land reforms and customary risk-sharing mechanisms as illustrated in section 1.4 provides valuable lessons regarding policy design of land reforms in weak institutional frameworks.

Attempts for reforming land rights should take into serious consideration the pre-existence of customary safety networks. This is critical in cases of ethnic minorities, or vulnerable groups of people that have to rely solely to the community for tackling risk. Those customary norms prescribe transfers of consumption units and land re-allocation as ways to insure their members against risk. When these two mechanisms constitute the predominant means of risk-sharing in the aﬀected communities, then a land reform can distort the functioning of the customary contract.

As shown in section 1.4 the land reform’s effect on the outside option can cause efficiency costs. It creates a clear trade-off between the amount of risk-sharing and production efficiency. In order for communities to maintain the existence of their informal contracts they can manipulate land allocation in a counter-productive way. In this case, a land reform can lead to misallocation of land, an inefficiency that would have been avoided, were the community was unaffected by land reforms.

Lastly, the present study provides a potential theoretical justification of the World Bank’s stance on land rights in Africa. The international organization, since the early 1990s has adopted a more inclusive and integrating policy stance regarding the functioning of local communities regarding land management. Based on section 1.4 it is explicit that the land tenure regime under the informal contract can achieve a certain flexibility of adjustment to productivity leading to a more efficient allocation of land.

Chapter 1 33 Essays on Macroeconomics and Development

To conclude, the implementation of a land reform aiming at granting private property should be preceded by a careful documentation and examination of the way local communities operate. The effect of a reform on the rural population might be beneficial if it strengthens the bargaining position of the villager, but also could bring detrimental effects regarding output efficiency.

34 Chapter 1 Essays on Macroeconomics and Development

References

(2019). Prindex. https://www.prindex.net/data.

Adamopoulos, T. and D. Restuccia (2019). Land reform and productivity: A quantitative analysis with micro data. Technical report, National Bureau of Economic Research.

Alchian, A. A. and H. Demsetz (1973). The property right paradigm. The journal of economic history 33 (1), 16–27.

Attanasio, O. and J.-V. Rıos-Rull (2000). Consumption smoothing in island economies: Can public insurance reduce welfare? European economic review 44 (7), 1225–1258.

Bambio, Y. and S. B. Agha (2018). Land tenure security and investment: Does strength of land right really matter in rural burkina faso? World Development 111, 130–147.

Banerjee, A. V., P. J. Gertler, and M. Ghatak (2002). Empowerment and eﬃciency: Tenancy reform in west bengal. Journal of political economy 110 (2), 239–280.

Bellemare, M. F. (2010). The productivity impacts of de jure and de facto land rights.

Besley, T. (1995). Property rights and investment incentives: Theory and evidence from ghana. Journal of Political Economy 103 (5), 903–937.

Besley, T., K. B. Burchardi, and M. Ghatak (2012). Incentives and the de soto eﬀect. The Quarterly Journal of Economics 127 (1), 237–282.

Besley, T. and M. Ghatak (2010). Property rights and economic development. In Handbook of development economics, Volume 5, pp. 4525–4595. Elsevier.

Brasselle, A.-S., F. Gaspart, and J.-P. Platteau (2002). Land tenure security and investment incentives: puzzling evidence from burkina faso. Journal of Development Economics 67 (2), 373–418.

Bruce, J. W. (2012). Simple solutions to complex problems: land formalisation as a “silver bullet”. Fair Land Governance: How to Legalise Land Rights in Rural Development; Otto, JM, Hoekema, AJ, Eds, 31–56.

Bruce, J. W., S. E. Migot-Adholla, F. Place, M. Roth, P. Hazell, P. Matlon, B. Blarel, G. Benneh, S. Atsu, W. Oluoch-Kosura, et al. (1994). Searching for land tenure security in Africa. Number 28043. The World Bank.

Chari, A., E. M. Liu, S.-Y. Wang, and Y. Wang (2017). Property rights, land misallocation and agricultural eﬃciency in china. Technical report, National Bureau of Economic Research.

Chapter 1 35 Essays on Macroeconomics and Development

Chen, C., D. Restuccia, and R. Santaeulàlia-Llopis (2017). The eﬀects of land markets on resource allocation and agricultural productivity. Technical report, National Bureau of Economic Research.

De Soto, H. (2001). The mystery of capital. Finance and Development 38 (1), 29–33.

Deininger, K. W. et al. (2003). Land policies for growth and poverty reduction. World Bank Publications.

Demsetz, H. (1974). Toward a theory of property rights. In Classic papers in natural resource economics, pp. 163–177. Springer.

FAO. (2012). State of Food and Agriculture 2012: Investing in Agriculture for a Better Future. FAO.

Feder, G. and D. Feeny (1991). Land tenure and property rights: Theory and implications for development policy. The World Bank Economic Review 5 (1), 135–153.

Feder, G. and A. Nishio (1998). The beneﬁts of land registration and titling: economic and social perspectives. Land use policy 15 (1), 25–43.

Goldstein, M. and C. Udry (2008). The proﬁts of power: Land rights and agricultural investment in ghana. Journal of political Economy 116 (6), 981–1022.

Gollin, D. and C. Udry (2019). Heterogeneity, measurement error and misallocation: Evidence from african agriculture. NBER Working Paper (w25440).

Hughes, O. (2014). Literature review of land tenure in niger, burkina faso, and mali: Context and opportunities. Baltimore, MD, USA: Catholic Relief Services.

IMPAQ (2015). Burkina faso rural land governance project impact evaluation.

Ligon, E., J. P. Thomas, and T. Worrall (2002). Informal insurance arrangements with limited commitment: Theory and evidence from village economies. The Review of Economic Studies 69 (1), 209–244.

Ljungqvist, L. and T. J. Sargent (2000). Recursive macroeconomics.

Migot-Adholla, S., P. Hazell, B. Blarel, and F. Place (1991). Indigenous land rights systems in sub-saharan africa: a constraint on productivity? The World Bank Economic Review 5 (1), 155–175.

Millenium-Challenge-Corporation (2016). Measuring results of the burkina faso rural land governance project. https://www.mcc.gov/resources/doc-pdf/summary-measuring- results-of-the-burkina-faso-rural-land-governance-project.

36 Chapter 1 Essays on Macroeconomics and Development

Morten, M. (2019). Temporary migration and endogenous risk sharing in village india. Journal of Political Economy 127 (1), 000–000.

Munshi, K. and M. Rosenzweig (2016). Networks and misallocation: Insurance, migration, and the rural-urban wage gap. American Economic Review 106 (1), 46–98.

Ouédraogo, M. (2002). Land tenure and rural development in burkina faso. Drylands Issue Pap 112, 1–24.

Platteau, J.-P. (1991). Traditional systems of social security and hunger insurance: Past achievements and modern challenges.

Platteau, J.-P. (1996). The evolutionary theory of land rights as applied to sub-saharan africa: a critical assessment. Development and change 27 (1), 29–86.

Platteau, J.-P. and A. Abraham (1987). An inquiry into quasi-credit contracts: The role of reciprocal credit and interlinked deals in small-scale ﬁshing communities. The Journal of Development Studies 23 (4), 461–490.

Restuccia, D. and R. Santaeulalia-Llopis (2017). Land misallocation and productivity. Technical report, National Bureau of Economic Research.

Townsend, R. M. (1994). Risk and insurance in village india. Econometrica 62, 539–539.

Udry, C. (1994). Risk and insurance in a rural credit market: An empirical investigation in northern nigeria. The Review of Economic Studies 61 (3), 495–526.

Udry, C. (2011). Land tenure. The Oxford Companion to the Economics of Africa 1.

USAID (2017). Property rights and resource governance. https://www.land-links.org/wp-

content/uploads/2010/08/USAIDLandT enureBurkinaF asoC ountryP rofile.pdf.

Zhang, Y. (2013). Characterization of a risk sharing contract with one-sided commitment. Journal of Economic Dynamics and Control 37 (4), 794–809.

Chapter 1 37 Essays on Macroeconomics and Development

1.9 Appendix

A Proofs

Proposition 1 :

Proof: Under the ﬁrst best, commitment friction is absent, hence in the optimization problem, the principal-head of the community does not take into account the participation constraint of the agent-villager. Hence the problem becomes:

X P (v) = max Πs (zsf(κs) − cs) + βP (ws) {c ,κ ,w } s s s s∈S X Πs{u(cs) + βws} ≥ v [PKC]:( θ) s∈S  κs ≥ 0 :(Π sν1s) κs ∈ [0, 1] =  1 − κs ≥ 0 :(Π sν2s)

aut ws ∈ [v , v¯]

Assigning the designated lagrange multipliers above, the lagrangian becomes:

X L = Πs (zsf(κs) − cs) + βP (ws) + s X +θ Πs[u(cs) + βws] − v s∈S

+Πsν1sκs + Πsν2s(1 − κs)

Deriving optimality conditions with respect to the choice variables:

∂L 0 0 1 = 0 → −1 + θu (cs) = 0 → u (cs) = [constant] ∂cs θ

∂L 0 0 = 0 → zsf (κs) + (ν1s − ν2s) = 0 → zsf (κs) = ν2s − ν1s ∂κs

since f(0) = 0 then ν1s = 0

Below I take cases regarding the zs ∈ [0, 1] constraint:

(a) κs = 0, ν1s > 0, ν2s = 0. The optimality condition reads:

∂L 0 = 0 → zsf (κs)|κs=0 = −ν1s ∂κs From technology f(·), f 0(0) > 0, hence the left hand side of the equation is positive, while the right hand side is negative. I can rule out this case.

38 Chapter 1 Essays on Macroeconomics and Development

(b) 1 − κs = 0, ν1s = 0, ν2s > 0. The optimality condition reads:

∂L 0 = 0 → zsf (κs)|κs=1 = ν2s ∂κs From technology f(·), the LHS is positive and the RHS is positive, this case can not be ruled out.

∂L 0 = 0 → zsf κs = 0 ∂κs From technology f(·), the LHS is positive and the RHS is zero. This case can be ruled out.

Cases (a), (b) and (c) partition the whole parameter space, so they prove that at the ﬁrst best, the κs = 1 is optimal.

∂L 0 0 = 0 → βP (ws) + θβ = 0 → P (ws) = −θ [constant] ∂ws

Proposition 2:

Proof: The maximization problem in the presence of the commitment friction takes the following form: S X P (v) = max Πs (zsf(κs) − cs) + βP (ws) {c ,κ ,w } s s s s=1 S X Πs{u(cs) + βws} ≥ v [PKC]:( θ) s=1 aut u(cs) + βws ≥ u(zsf(κs)) + βv (κs)[PC] ∀s :(Π sφs)  κs ≥ 0 :(Π sν1s) κs ∈ [0, 1] =  1 − κs ≥ 0 :(Π sν2s)

aut ws ∈ [v , v¯]

Assigning the lagrange mutlipliers as above, the Lagrangian reads: X L = (zsf(κs) − cs) + βP (ws) + s∈S X +θ Πs[u(cs) + βws] − v s∈S aut +Πsφs u(cs) + βws − u(zsf(κs) − βv )(κs) +

+Πsν1sκs + Πsν2s(1 − κs)

Chapter 1 39 Essays on Macroeconomics and Development

Before deriving the optimality conditions, I derive the following:

aut ∂v 1 X 0 0 = Πru (zrf(κs))zrf (κs) ∂κs 1 − β r aut ∂v 1 0 X 0 = f (κs) Πru (zrf(κs))zr ∂κs 1 − β r aut ∂v 1 0 0 = f (κs) Eru (zrf(κs))zr ∂κs 1 − β | {z } ω>0 aut ∂v 1 0 = f (κs)ω ∂κs 1 − β Deriving the focs:

∂L 0 0 0 :Πs(−1) + θΠsu (cs) + φsΠsu (cs) = 0 → u (cs)[θ + φs] = 1 (1.6) ∂cs

∂L 0 0 :ΠsβP (ws) + θΠsβ + φsΠsβ = 0 → P (ws) = −(θ + φs) (1.7) ∂ws

 aut  ∂L 0 0 0 ∂v :Πszsf (κs) − Πsφsu (zsf(κs))zsf (κs) + β  + Πsν1s − Πsν2s = 0 ∂κs ∂κs   0 0 0 1 0 X 0 Πszsf (κs) − Πsφsu (zsf(κs))zsf (κs) + β f (κs) Πru (zrf(κs))zr + Πsν1s − Πsν2s = 0 1 − β r 0 0 β 0 zsf (κs) 1 − φsu (zsf(κs)) − φs f (κs)ω + [ν1s − ν2s] = 0 1 − β

[Assume that constraints on κs are slack - corner solutions excluded ν1s, ν2s = 0] 0 β 0 zs 1 − φsu (zsf(κs)) = φs f (κs)ω 1 − β 0 φs β 1 − φsu (zsf(κs)) = ω zs 1 − β

0 φs β φsu (zsf(κs)) = 1 − ω zs 1 − β

0 1 1 β u (zsf(κs)) = − ω φs zs 1 − β (1.8)

Proposition 3:

Proof: Let the following functional forms:

(1−α) cs u(cs) = 1 − α

40 Chapter 1 Essays on Macroeconomics and Development

and 1−γ y = zsf(κs) = zsκs

∂vaut 1. First I derive the ∂κs under those functional forms.

∞ S ∂vaut X t X 0 0 = β Πru (zrf(κs))zrf (κs) ∂κs t=0 r=1 S ∂vaut 1 0 X 0 = f (κs) Πru (zrf(κs))zr ∂κs 1 − β r=1 S ∂vaut 1 0 X 0 = f (κs) Πru (zrf(κs))zr ∂κs 1 − β r=1 | {z } ω>0 Using the functional forms to get ω

X 0 ω = Πru (zrf(κs))zr = Eruc(zrf(κs))zr r Plugging the functional forms of u, f X −α −α ω = Πrzr f(κs) zr r X −α 1−γ −α ω = Πrzr (κs ) zr r −α(1−γ) X −α ω = κs Πrzr r −α(1−γ) X −α ω = κs Πrzr r | {z } =ξ>0: constant

Now plug this expression to the foc wrt κs:

1 1 β uc(zsf(κs)) = − ω φs zs 1 − β −α 1 1 β −α(1−γ) zsf(κs) = − κs ξ φs zs 1 − β −α −α 1−γ 1 1 β −α(1−γ) zs κs = − κs ξ φs zs 1 − β

−α −α(1−γ) 1 1 β −α(1−γ) zs κs = − κs ξ φs zs 1 − β

Now I want to derive a relationship between κs and zs from the above relationship which is the optimal rule for setting the fraction of land.

Step 1 Multiply by zs:

1−α −α(1−γ) zs β −α(1−γ) zs κs = − κs ξ φs 1 − β

Step 2 Multiply by φs:

Chapter 1 41 Essays on Macroeconomics and Development

1−α −α(1−γ) β −α(1−γ) φsz κ = zs − φs κ ξ s s 1 − β s

Step 3 Transfer everything to the RHS and name it H(κs, zs) on which you apply the IFT

1−α −α(1−γ) β −α(1−γ) H(κs, zs) = zs − φszs κs − φs 1−β κs ξ = 0 From the IFT i know the following:

∂H(κs,zs) ∂κs = − ∂zs ∂H(κs,zs) ∂zs ∂κs

where ∂H(κs, zs) −α(1−γ) −α = 1 − (1 − α)φsκs zs ∂zs and

∂H(κs, zs) 1−α −α(1−γ)−1 β −α(1−γ)−1 = −(−α(1 − γ)φszs κs ) − (−α(1 − γ)φs κs ξ) → ∂κs 1 − β

∂H(κs, zs) 1−α −α(1−γ)−1 β −α(1−γ)−1 → = α(1 − γ)φszs κs + α(1 − γ)φs κs ξ → ∂κs 1 − β   ∂H(κs, zs) −α(1−γ)−1 1−α β → = α(1 − γ)φsκs zs + ξ ∂κs 1 − β

Hence the IFT becomes as follows:

∂H(κs,zs) −α(1−γ) −α ∂κs 1 − (1 − α)φsκ z = − ∂zs = − s s ∂H(κs,zs)   ∂zs ∂κs −α(1−γ)−1 1−α β α(1 − γ)φsκs zs + 1−β ξ

−α(1−γ) −α ∂κs (1 − α)φsκs zs − 1 =   ∂zs −α(1−γ)−1 1−α β α(1 − γ)φsκs zs + 1−β ξ

Note that the sign of the relationship between κs and zs depends on the sign of the nominator:

Flexible Land Regime ∂κs > 1. : ∂zs 0

−α(1−γ) −α (1 − α)φsκs zs − 1 > 0 → −α(1−γ) −α →(1 − α)φsκs zs > 1 →

−α(1−γ) 1 →(1 − α)φsκs > −α → zs α −α(1−γ) →zs < (1 − α)φsκs → 1 α −(1−γ) →zs < [(1 − α)φs] κs

42 Chapter 1 Essays on Macroeconomics and Development

Rigid Land Regime ∂κs 2. : ∂zs = 0

1 α −(1−γ) zs = [(1 − α)φs] κs

Counter Productive Land Regime ∂κs < 3. : ∂zs 0

1 α −(1−γ) zs > [(1 − α)φs] κs

To summarise the above result, the risk-sharing contract within the community might end up with a land allocation regime that falls within one or more of the following categories, depending on the relation between idiosyncratic productivity and fraction of land allocated to the villager at the time of the land reform implementation.

Notes on Propostions 2 & 3:

The results stemming from the two propositions hold under the case in which corner solutions at the level of κs are ruled out. Below I address the case in which corner solutions are included in the analysis. I will only work on the optimality condition with respect to κs as the remaining choice variables do not change. wrt κs:

∂L = 0 ∂κs 0 0 0 ∂vaut(κs) Πszsf (κs) − Πsφs u (zsf(κs))zsf (κs) + β + Πsv1s − Πsv2s = 0 ∂κs   0 0 0 β 0 zsf (κs) − φsu (zsf(κs)zsf (κs) + f (κs)ω(κs) + ν1s − ν2s = 0 1 − β

Below I take cases that partition the parameter space κs ∈ [0, 1]:

(a) κs ≥ 0 is binding, κs = 0, which implies v1s > 0 and v2s = 0 The optimality condition reads: 0 0 0 β 0 zsf (κs) − φs u (zsf(κs))zsf (κs) + f (κs)ω(κs) + v1s = 0 1 − β I can rule out this case by examining the participation constraint under this case,

which must hold with equality since φs > 0.

u(cs) + βws = u(zsf(κs)) + βvaut(κs)

Chapter 1 43 Essays on Macroeconomics and Development

The right hand side of the PC is strictly positive, since from foc wrt cs:

0 1 θ,φs>0 0−1 1 u (cs) = −−−−→ cs = u ( ) > 0 θ + φs θ + φs

The left hand side for a binding non-negative constraint on κs becomes:

u(zsf(0)) + βvaut(0) = 0

where the ﬁrst term using f(0) = 0 from the properties of the production function P∞ t PS is 0 and thus u(0) = 0 and the second term vaut(0) = t=1 β r=1 Πru(zrf(0)) = 0.

As a result this case cannot hold with a binding non-negative constraint on κs since the PC will not bind.

(b) 1 − κs ≥ 0 is binding, κs = 1 which implies v1s = 0 and v2s > 0

0 0 0 β 0 zsf (κs) − φsu (zsf(κs))zsf (κs) − φs f (κs)ω(κs) = 0 1 − β 0 0 0 β 0 zsf (κs) − φs u (zsf(κs))zsf (κs) − f (κs)ω(κs) = 0 ∀κs ∈ (0, 1) 1 − β

Note: From cases (a), (b) and (c) we can rule out only κs = 0. The optimality condition is consistent with κs ∈ (0, 1].

In fact the following additional analysis that the optimality condition can accommodate both cases (ie κs = 1 and κs ∈ (0, 1)) reads:

• From the optimality condition in case (b), I get:   0 0 0 β 0 zsf (1) − φsu (zsf(1))zsf (κs) + f (1)ω(1) = ν2s > 0 1 − β

and from the optimality condition in case (c), I get that:

0 0 0 β 0 zsf (κs) − φs u (zsf(κs))zsf (κs) − f (κs)ω(κs) = 0 ∀κs ∈ (0, 1) 1 − β

• Combining the two I infer that the RHS for κs = 1 should be larger than the RHS

for κs ∈ (0, 1) and I will show that.

0 0 • zsf (1) < zsf (κs) ∀κs ∈ (0, 1) since f is stricly convave.

44 Chapter 1 Essays on Macroeconomics and Development

• Based on the above, a necessary condition for the inequality to hold is that

0 0 β 0 φs u (zsf(κs))|κ =1zsf (κs)|κ =1 + f (κs)|κ =1ω(κs)|κ =1 < s s 1 − β s s 0 0 β 0 φs u (zsf(κs))zsf (κs) + f (κs)ω(κs) 1 − β

which holds since

0 0 0 0 – u (zsf(1))zsf (1) < u (zsκs)zsf (κs) ∀κs ∈ (0, 1) since u and f are strictly concave.

0 0 – and f (1)ω(1) < f (κs)ω(κs) ∀κs ∈ (0, 1) again due to concavity of u and f.

– Since there is no contradiction so far I proceed to the suﬃcient condition for

the inequality to hold, ie for v2s > 0: 0 0 zs f (κs) − f (1)

<     0 0 β 0 0 0 β 0 u (zsf(κs))zsf (κs) + f (κs)ω(κs) − u (zsf(1))zsf (1) + f (1)ω(1) → 1 − β 1 − β 0 0 β 0 0 0 β f (κs) zsu (zsf(κs)) + ω(κs) − zs −f (1) zsu (zsf (1)) + ω(1) − zs > 0 1 − β 1 − β

which holds since the right term is larger than the left term. Note that inside

the brackets, we are subtracting the same zs.

The diference derived above can give an expression for ν2s.

Since both cases ((b) and (c)) can occur under a binding PC, then I should examine all possible contingencies - by combining the sole case under a slack PC and the two possible cases under a binding PC.

1. Case Slack PC: φs = 0

(a) κs = 1

2. Case Binding PC: φs > 0

(a) κs = 1

(b) κs ∈ (0, 1)

First take only two states: sL and sH .

Then for sL and based on the above there can only be 3 cases possible:

• Case 1a: φsL = 0 and κsL = 1

Chapter 1 45 Essays on Macroeconomics and Development

• Case 1b: φsL > 0 and κsL = 1

• Case 1c: φsL > 0 and κsL ∈ (0, 1)

Then for sH and based on the above there can only be 3 cases possible:

• Case 2a: φsH = 0 and κsH = 1

• Case 2b: φsH > 0 and κsH = 1

• Case 2c: φsH > 0 and κsH ∈ (0, 1)

As a result, you need to compute the allocation for all 3 × 3 cases - rule out any if they cannot hold and apply maximization to the remaining. This means that you derive all allocations which are {cL, cH , wL, wH , κL, κH } for each of the 9 cases.

9 cases

1. Case 1a, Case 2a 0 0 Slack PC-low: u (cl) = 1/θ, P (wl) = −θ 0 0 Slack PC-high: u (ch) = 1/θ, P (wh) = −θ

Hence, cl = ch and wl = wh

Slack PC-low: u(cl) + βwl > u(zlf(1)) + βva(1)

Slack PC - high: u(ch) + βwh > u(zhf(1)) + βva(1)

Condition: u(ch) + βwh = u(cl) + βwl > u(zhf(1)) + βva(1)

2. Case 1a, Case 2b 0 0 Slack PC-low: u (cl) = 1/θ, P (wl) = −θ 0 0 Binding PC-high: u (ch) = 1/(θ + φh), P (wh) = −(θ + φh)

Hence, ch > cl and wh > wl

From above: u(ch) + βwh > u(cl) + βwl

Slack PC-low: u(cl) + βwl > u(zlf(1)) + βva(1)

Binding PC - high: u(ch) + βwh = u(zhf(1)) + βva(1)

From above rhs: u(zhf(1)) + βva(1) > u(zlf(1)) + βva(1)

→ u(zhf(1)) > u(zlf(1)) which holds.

3. Case 1a, Case 2c 0 0 Slack PC-low: u (cl) = 1/θ, P (wl) = −θ 0 0 Binding PC-high: u (ch) = 1/(θ + φh), P (wh) = −(θ + φh)

Hence, ch > cl and wh > wl

From above: u(ch) + βwh > u(cl) + βwl

Slack PC-low: u(cl) + βwl > u(zlf(1)) + βva(1)

Binding PC - high: u(ch) + βwh = u(zhf(κs)) + βva(κs)

46 Chapter 1 Essays on Macroeconomics and Development

→ u(zhf(κs)) + βva(κs) > u(zlf(1)) + βva(1)

Conditions: zhf(κs) > zlf(1) [necessary] u(zhf(κs))−u(zlf(1)) > β and va(κs)−va(1) [suﬃcient] 4. Case 1b, Case 2a 0 0 Binding PC-low: u (cl) = 1/(θ + φl), P (wl) = −(θ + φl) 0 0 Slack PC-high: u (ch) = 1/θ, P (wh) = −θ

From above: cl > ch and wl > wh

From above: u(cl) + βwl > u(ch) + βwh

sub lhs & rhs: u(zlf(1)) + βva(1) > u(zhf(1)) + βva(1)

→ u(zlf(1)) > u(zhf(1)) which cannot hold -RULED OUT

5. Case 1b, Case 2b 0 0 Binding PC-low: u (cl) = 1/(θ + φl), P (wl) = −(θ + φl) 0 0 Binding PC-high: u (ch) = 1/(θ + φh), P (wh) = −(θ + φh)

Binding PC-low: u(cl) + βwl = u(zlf(1)) + βva(1)

Binding PC - high: u(ch) + βwh = u(zhf(1)) + βva(1)

Compare rhs: u(zhf(1)) + βva(1) > u(zlf(1)) + βva(1)

From above: u(ch) + βwh > u(cl) + βwl which can hold if:

(I) u(ch) − u(cl) > β(wl − wh) but this would give diﬀerent predictions for phih and

φl when taking w and c...

(II) β(wh − wl) > u(cl) − u(ch) but this would give diﬀerent predictions for phih and

φl when taking w and c...

(III) u(ch) > u(cl) and wh > wl

So only (III) can hold which imples φh > φl

0 0 6. Case 1b, Case 2c Binding PC-low: u (cl) = 1/(θ + φl), P (wl) = −(θ + φl) 0 0 Binding PC-high: u (ch) = 1/(θ + φh), P (wh) = −(θ + φh)

Let φh > φl then ch > cl and wh > wl

From above: u(ch) + βwh > u(cl) + βwl

From binding PCs: u(zhf(κs)) + βva(κs) > u(zlf(1)) + βva(1)

Condition: zhf(κs) > zlf(1) [necessary] u(zhf(κs))−u(zlf(1)) > β and va(κs)−va(1) [suﬃcient]

0 0 7. Case 1c, Case 2a Binding PC-low: u (cl) = 1/(θ + φl), P (wl) = −(θ + φl) 0 0 Slack PC-high: u (ch) = 1/θ, P (wh) = −θ

From above: cl > ch and wl > wh

From above: u(cl) + βwl > u(ch) + βwh

From above with Binding PClow & Slack PChigh: u(zlf(κs))+βva(κs) > u(zhf(1))+

Chapter 1 47 Essays on Macroeconomics and Development

βva(1) which cannot hold as all terms on the left lower than all terms on the right. - RULED OUT

0 0 8. Case 1c, Case 2b Binding PC-low: u (cl) = 1/(θ + φl), P (wl) = −(θ + φl) 0 0 Binding PC-high: u (ch) = 1/(θ + φh), P (wh) = −(θ + φh)

Compare RHS of pc: u(zhf(1)) + βva(1) > u(zlf(κs)) + βva(κs)

Both pcs are binding: u(ch) + βwh > u(cl) + βwl Which holds under the three cases of (1b, 2b) but we only keep the third which implies

φh > φl

0 0 9. Case 1c, Case 2c Binding PC-low: u (cl) = 1/(θ + φl), P (wl) = −(θ + φl) 0 0 Binding PC-high: u (ch) = 1/(θ + φh), P (wh) = −(θ + φh)

Let φh > φl then ch > cl and wh > wl

From above: u(ch) + βwh > u(cl) + βwl h h l l from binding PCs: u(zhf(κs )) + βva(κs ) > u(zlf(κs)) + βva(κs) h l The above holds for κs > κs but also h l for κs > κs iﬀ: h l u(zhf(κs ))−u(zlf(κs)) l h > β va(κs)−va(κs )

Proposition 4:

Proof: Representing the two consumption allocations as two lotteries, ie probability distributions with diﬀerent returns for each realisation, for proposition 4 to hold I just need to show that the lottery corresponding to the debt economy consumption allocation second order stochastically dominates the lottery corresponding to the OSLC economy. Then it follows that a risk-averse household would prefer the s.o.s dominant lottery.

Assume that C = R+ is the set of feasible consumption allocations under the two regimes I examine. I represent the two regimes as lotteries with corresponding cumulative distributions O for the oslc economy and D for the debt economy. The corresponding density functions are o and d. The maximal support of the two distributions is [ymin, ymax] and

C ∈ [ymin, ymax].

For D to second order stochastically dominate O I need to prove the two necessary and suﬃcient conditions:

1. E(cdebt) ≥ E(coslc)

2. min{cdebt} ≥ min{coslc}

48 Chapter 1 Essays on Macroeconomics and Development

I start from proving condition 2, then condition 1 and then proceed to proving that a risk-averse agent strictly prefers the second order stochastically dominant allocation.

The one-sided limited commitment economy corresponds to the textbook model from Ljungqvist and Sargent(2000) chapter 20. Therefore I am using the equation 20.3.25 which provides a closed form solution for consumption:

j X u(cj) = u(yj) − β [u(yj) − u(yk)] ∀j ∈ S (20.3.25) κ=1

This relation informs about the optimal setting of the choice variable cj under the contract. Note that for j = 1 ie the lowest possible realisation of endowment, the following holds:

1 X u(c1) = u(y1) − β Πk[u(y1) − u(yk)] κ=1

u(c1) = u(y1) − βΠ1[u(y1) − u(y1)] u0()>0 u(c1) = u(y1) −−−→ c1 = y1 which denotes that the lowest value of consumption allocation under the oslc is equal to the lowest endowment realisation:

oslc min{c } = ymin = y1 (1.9)

Note that in the debt economy the household can borrow in bad times and pay back in good times. As a result the mincdebt cannot be lower than the minimum level of endowment realisation. Therefore, condition 2 holds. From (20.3.25) I can ﬁnd an upper bound for consumption. Due to the household being risk averse: j X u(cj) = u(yj) − β [u(yj) − u(yk)] κ=1 j j X X u(cj) ≥ u yj − yjβ Πκ + β yk κ=1 κ=1 j j X X cj ≤ yj − yjβ Πκ + β Πkyk κ=1 κ=1 where the RHS of the above inequality constitutes the upper bound for values of the consumption allocation under the oslc.

Assume that the weak inequality holds with equality and consumption acquires its upper bound. j j X X cj = yj − yjβ Πκ + β Πkyk κ=1 κ=1 j j (1.10) X X E(cj) = E(yj) − E(yj) Πκ + β Πkyk κ=1 κ=1

Chapter 1 49 Essays on Macroeconomics and Development and j j X X V ar(cj) = V ar(yj(1 − β Πk + β Πκyκ)) κ=1 κ=1 j (1.11) X 2 V ar(cj) = (1 − β Πκ) V ar(yj) κ=1 (9) and (10) gives the ﬁrst and second moment of the consumption allocation under the oslc.

Debt Economy The Lagrangian for the problem above takes the following form:

∞ X t L = β {u(ct) + λt yt + dt − ct − (1 + r)dt−1 } t=0

Deriving the ﬁrst order conditions with respect to consumption and debt obligation maturing at the next period I have:

0 u (ct) = λt wrt [ct]

λt = β(1 + r)Etλt+1 wrt [dt]

I make the assumption that β(1 + r) = 1. Hence the ﬁrst order condition with respect to debt becomes:

0 0 11 λt = Etλt+1 → u (ct) = Etu (ct+1) → ct = Etct+1

I turn from the sequential budget constraint of the agent under the PR regime to the intertemporal budget constraint, in order to derive an expression for the lifetime consumption path prescribed by the problem.

solving for (1+r)dt−1 ct + (1 + r)dt−1 = yt + dt −−−−−−−−−−−−→ (1 + r)dt−1 = yt + dt − ct [at t]

yt+1 − ct+1 dt+1 (1 + r)dt = yt+1 + dt+1 − ct+1 → dt = + [at t + 1] 1 + r 1 + r ...

yt+j − ct+j dt+j (1 + r)dt+j−1 = + [at t + j] 1 + r 1 + r

11at the last step I am assuming that the marginal utility of consumption is linear, so that the expectation can pass through. Moreover, note that ct = Etct−1 stems from assuming quadratic preferences, which is appealing because you can have a closed form solution for consumption.

50 Chapter 1 Essays on Macroeconomics and Development

By substituting backwards, I obtain:

yt+j−1 − ct+j−1 dt+j−1 (1 + r)dt+j−2 = + 1 + r 1 + r ... ∞ X yt+j − ct+j dt+j (1 + r)dt−1 j + j j=0 (1 + r) (1 + r)

Here, I use the no-Ponzi scheme condition:

dt+j lim = 0 j→∞ (1 + r)j

Hence the expression for dt−1 becomes:

∞ X yt+j − ct+j (1 + r)dt−1 = j j=0 (1 + r)

Solving the above for consumption I have:

∞ ∞ X ct+j X yt+j j = j − (1 + r)dt−1 j=0 (1 + r) j=0 (1 + r) where the LHS of the above expression can be written as follows:

∞ X ct+j ct+1 ct+2 j = ct + + 2 + ... j=0 (1 + r) (1 + r) (1 + r)

The agent optimally sets:

ct = Et(ct+1) ∀t

Hence, ∞ ∞ X ct+j X 1 1 + r j = ct j = ct j=0 (1 + r) j=0 (1 + r) r Hence, the intertemporal budget constraint is given by:

∞ 1 + r X yt+j ct = Et j − (1 + r)dt−1 r j=0 (1 + r) ∞ r X yt+j r ct = Et j − (1 + r)dt−1 1 + r j=0 (1 + r) 1 + r ∞ r X yt+j ct = Et j − rdt−1 [I] 1 + r j=0 (1 + r)

I use [I] to derive the ﬁrst and second moment for consumption under the PR regime. pr (For expositional reasons I substitute ct with ct to denote that consumption corresponds

Chapter 1 51 Essays on Macroeconomics and Development to the PR regime). Taking expectation in [I]:

 ∞  debt r X yt+j E(ct ) = E j − rdt−1 1 + r j=0 (1 + r)  ∞  debt r X yt+j E(ct ) = E j  − rdt−1 1 + r j=0 (1 + r)   debt r yt yt+1 yt+2 E(c ) = E + + + ... − rdt−1 t 1 + r (1 + r)0 (1 + r)1 (1 + r)2   debt r E(yt) E(yt+1) E(yt+2) E(c ) =  + + + ... − rdt−1 t 1 + r (1 + r)0 (1 + r)1 (1 + r)2   (1.12) debt r 1 1 1 E(c ) = E(yt) + + + ... − rdt−1 t 1 + r (1 + r)0 (1 + r)1 (1 + r)2 ∞ debt r X 1 E(ct ) = E(yt) j − rdt−1 1 + r j=0 (1 + r)

debt r 1 E(ct ) = E(yt) 1 − rdt−1 1 + r 1 − 1+r debt r 1 + r E(c ) = E(yt) − rdt−1 t 1 + r r debt E(ct ) = E(yt) − rdt−1

correspondingly the variance of the consumption allocation under the debt economy is

52 Chapter 1 Essays on Macroeconomics and Development the following: Now take the Variance of [I]:

 ∞  debt r X yt+j V ar(ct ) = V ar j − rdt−1 1 + r j=0 (1 + r) 2    ∞  debt r X yt+j V ar(ct ) =   V ar j  1 + r j=0 (1 + r)  2   debt r yt yt+1 yt+2 yt+3 V ar(c ) =   V ar + + + + ... yt is iid t 1 + r (1 + r)0 (1 + r)1 (1 + r)2 (1 + r)3  2          debt r yt yt+1 yt+2 yt+3 V ar(c ) =   V ar  + V ar  + V ar  + V ar  + ... t 1 + r (1 + r)0 (1 + r)1 (1 + r)2 (1 + r)3  2 20  21  22 debt r 1 1 1 V ar(c ) =      V ar(yt) +    V ar(yt) +    V ar(yt) + ... t 1 + r 1 + r 1 + r 1 + r

since V ar(yt) = V ar(yt+1) = V ar(yt+2) = ... = V ar(yt+j)  2  20  21  22  debt r 1 1 1 V ar(c ) =   V ar(yt)   +    +    + ... t 1 + r 1 + r 1 + r 1 + r  2  j ∞ 2 debt r X 1 V ar(ct ) =   V ar(yt)   1 + r j=0 1 + r  2 debt r 1 V ar(ct ) =   V ar(yt) 1 1 + r 1 − (1+r)2  2 debt r 1 V ar(c ) =   V ar(yt) t 1 + r (1+r)2−1 (1+r)2 2   2 debt r (1 + r) V ar(c ) =   V ar(yt) t 1 + r (1 + r)2 − 1 2 2 debt r (1 + r) V ar(c ) = V ar(yt) t (1 + r)2 (1 + r)2 − 1 2 debt r V ar(c ) = V ar(yt) t (1 + r)2 − 1 2 debt r V ar(c ) = V ar(yt) t 1 + 2r + r2 − 1 2 debt r V ar(c ) = V ar(yt) t 2r + r2 2 debt r V ar(c ) = V ar(yt) t r(2 + r) debt r V ar(c ) = V ar(yt) t (2 + r) (1.13) (11) and (12) give the ﬁrst and second moments for the consumption allocation under the debt economy.

Chapter 1 53 Essays on Macroeconomics and Development

First notice that the variance of the debt economy’s consumption allocation is lower than r the variance of the oslc consumption allocation. It just suﬃces to show that 2+r < 1 Pj (1 − 1+r κ=1 Πκ). Note that the lower bound of the rhs of the inequality is when j = S, Pj hence the κ=1 Πκ) = 1 takes its maximum value, as a result the lower bound of the 1 r r r rhs is (1 − 1+r ) = 1+r which is larger than 2+r . As result, it holds that 2+r < (1 − 1 Pj r 1 Pj 2 1+r κ=1 Πκ) ∀j and therefore 2+r < (1 − 1+r κ=1 Πκ)

∗ r Pj Note that for d < 1+r κ=1 Πκ[E(yj)−yk] and combining (9) and (10) I obtain condition 1 for second order stochastic dominance.

Since conditions 1 and 2 with these assumptions hold, then I infer that the consumption allocation under the debt economy second order stochastically dominates the consumption allocation from the oslc economy. Therefore it will provide the household with higher expected utility as shown below: Deﬁnition: Distribution D second order stochastically dominates distribution O if for every x ∈ C: Z x Z x D(c)dc ≤ O(c)dc ymin ymin What I want to show is that when the cumulative distribution D second order stochastically dominates distribution then the risk averse agent prefers D since it delivers higher expected utility ie

Z ymax Z ymax Z ymax Z ymax u(c)d(c)dc ≥ u(c)o(c)dc → u(c)d(c)dc − u(c)o(c)dc ≥ 0 ymin ymin ymin ymin I take the RHS of the above expression:

Z ymax Z ymax u(c)d(c)dc − u(c)o(c)dc [integrating by parts] ymin ymin ymax Z ymax ymax Z ymax u(c)D(c) − u0(c)D(c)dc − u(c)O(c) + u0(c)O(c)dc ymin ymin ymin ymin Z ymax Z ymax 0 0 u(ymax) − 0 − u (c)D(c)dc − u(ymax) + 0 + u (c)O(c)dc ymin ymin Z ymax u0(c) O(c) − D(c) [integrating by parts] ymin Z c Z c also deﬁne So(c) = O(c)dc and Sd(c) = D(c)dc ymin ymin ymax Z ymax 0 00 u (c) So(c) − Sd(c) − u (c) So(c) − Sd(c) dc ymin ymin Z ymax 0 00 u (ymax) So(ymax) − Sd(ymax) − u (c) So(c) − Sd(c) dc ymin

note So(ymax) − Sd(ymax) = 0 Z ymax 00 − u (c) So(c) − Sd(c) dc ymin

I know that D second order stochastically dominates O, so that So(c) > Sd(c) ∀c ∈ C

54 Chapter 1 Essays on Macroeconomics and Development and u00(c) < 0 so the above last expression is positive indicating that the risk averse agent derives higher expected utility under the debt economy regime compared to the oslc.

Note that above I have made the assumption that the cumulative distribution of the PR regime second order stochastically dominates the cumulative distribution under autarky and as a direct result, the expected utilities are related as shown.

B Figures

Chapter 1 55 Essays on Macroeconomics and Development Nations Source: West Africa: Land Use and Land Cover Dynamics & United Figure B.2: Absence of Ownership Documents in African Countries ( pri 2019 )) Figure B.1:

56 Chapter 1 Essays on Macroeconomics and Development Phase I Phase I Phase I Phase II Phase II (2009-2012) (2009-2012) (2009-2012) (2012-2014) (2012-2014) (national level) (17 communes) (17 communes) (additional 30 communes) (additional 30 communes) 1. Support government’s efforts to develop and implement improved rural land legislation and torevise develop, and implement other legalframeworks. and procedural 2. Significant public outreach programpeople to about inform the new legislationsbenefits and its expected 1. Improve institutional capacity toland deliver services in rural areas. 2. Funding of series ofdecentralization land services. registration, mapping and 1. Site specific land rights2. formalization Provision sub-activities of APFR certificates 3. Preparation of land titlesareas. and leases in selected project Rural Land Governance Project Burkina Faso Rural Land Governance Project Impact Evaluation (IMPAQ, 2015) Table 1.5: and and Site-Specific Land Tenure Interventions Communication Capacity Building Institutional Development Legal and Procedural Change Activities Title Description Phases Activity No 1 Activity No 2 Activity No 3

Chapter 1 57 Essays on Macroeconomics and Development Note: Tenure insecurity: : % of people who believe it is somewhat or Figure B.4: very likely that they couldtheir lose will their in the right next to 5 use years. their property or part of it against MCC - Baseline Survey - Conﬂict Resolution Figure B.3:

58 Chapter 1 Essays on Macroeconomics and Development

Figure B.5: Communes that participated to Phase I and Phase II of the RLG

Chapter 1 59 Essays on Macroeconomics and Development

Figure B.6: Traditional leaders’ inﬂuence in local community

Figure B.7: Who is primarily responsible for allocating land

60 Chapter 1 Essays on Macroeconomics and Development

Figure B.8: Role of leaders in allocating land by administrative province

Figure B.9: Distribution of plot size

C Tables

Production value based productivity measure

Chapter 1 61 Essays on Macroeconomics and Development

Figure B.10: Distribution of household land holdings

Figure B.11: Diﬀerence in plot size by region

62 Chapter 1 Essays on Macroeconomics and Development

Figure B.12: Diﬀerence in household land holdings per region

Figure B.13: Diﬀerence in plot size at the individual level

Chapter 1 63 Essays on Macroeconomics and Development

Figure B.14: Diﬀerence in plot size at the household level

Table 1.6: On the ﬁrst column is the number of parcels for which production data are available, in the second is the number of parcels for which prices per local unit of measure are imputed. The low percentages are due to the diﬀerent local units of measurement used.

Min. Q1 Median Mean Q3 Max Obs Baseline 0.01 0.41 1 1.59 2.01 12 5,680 Interim 0.01 0.37 0.91 1.51 2 14 9,108

Table 1.7: Size of parcels (in ha)

64 Chapter 1 Essays on Macroeconomics and Development

Min Q1 Median Mean Q3 Max Baseline 1.00 1.00 2.00 2.48 3.00 18.00 Interim 1.00 1.00 2.00 2.73 3.00 17.00

Table 1.8: Number of parcels per household

Min Q1 Median Mean Q3 Max Baseline 0.01 1.58 3.00 3.94 5.02 34.00 Interim 0.01 1.75 3.24 4.11 5.25 45.50

Table 1.9: Size of HH’s land holdings (in ha)

Min Q1 Median Mean Q3 Max plot level -99.5 -17.3 0.00 -3.36 9.44 100 individual level -99.5 -18.20 0.00 -4.10 8.89 100 hh level -99.5 -18.86 0.00 -5.15 7.95 100

Table 1.10: Distribution of land size diﬀerence

Chapter 1 65 Essays on Macroeconomics and Development Phase I Phase I Phase I Phase II Phase II (2009-2012) (2009-2012) (2009-2012) (2012-2014) (2012-2014) (national level) (17 communes) (17 communes) (additional 30 communes) (additional 30 communes) 1. Support government’s efforts to develop and implement improved rural land legislation and torevise develop, and implement other legalframeworks. and procedural 2. Significant public outreach programpeople to about inform the new legislationsbenefits and its expected 1. Improve institutional capacity toland deliver services in rural areas. 2. Funding of series ofdecentralization land services. registration, mapping and 1. Site specific land rights2. formalization Provision sub-activities of APFR certificates 3. Preparation of land titlesareas. and leases in selected project Rural Land Governance Project Burkina Faso Rural Land Governance Project Impact Evaluation (IMPAQ, 2015) and and Table 1.11: Site-Specific Land Tenure Interventions Communication Capacity Building Institutional Development Legal and Procedural Change Activities Title Description Phases Activity No 1 Activity No 2 Activity No 3

66 Chapter 1 Essays on Macroeconomics and Development

D Naive productivity measure

Even though the production value based productivity measure is highly accurate and allows for cross-crop comparison, it restricts the sample signiﬁcantly, due to low market participation from the side of the individual farmers and sizeable heterogeneity across local measurement units. To address this issue I proceed in performing the same analysis by employing a naive productivity measure, agnostic to the type of crops.

For plot p ∈ P where P is the set of all plots in the survey, I calculate the following

qp agricultural productivityp = κp where qp is the quantity produced and κp is the size of the plot p.

Table 1.12 shows the estimation results from regression (1.5) with the naive productivity measure used to build individual productivity. The results are robust to this speciﬁcation as well.

Table 1.12: Results from (4)

Dependent variable:

∆κi,2012/09 treated EAs control EAs

(1) (2)

below-median×ind.productivity 0.197∗∗ 0.066 (0.079) (0.170)

Individual controls Yes Yes Household control Yes Yes Observations 1,620 1,343 R2 0.220 0.185 Adjusted R2 0.213 0.176

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

Chapter 1 67 Essays on Macroeconomics and Development

As a robustness check, the focus is swifted on the direction of the diﬀerence in land size so as to account only individuals that experienced an increase in land after the rerorm. Therefore, the analysis proceeds in building a dummy variable Increase which takes the value 1 if the change in plot size between the two waves is positive and 0 if not.  1 ∆κ2012/09 > 0 Increasei =  0 ∆κ2012/09 ≤ 0

Since the dependent variable is a dummy, I need to use a non-linear specification to compute the effect of the individual productivity for an individual ranking below the village median productivity to the difference in land size between the two waves. The specification to be estimated is the following probit model:

0 P r[increase = 1|X] = Φ(β1Below-village-median1,2009∗Individual productivityi,2009+X β2) (1.14) where Φ() is the cdf of the standard normal distribution and X is a matrix of control variables addressing individual and household characteristics.

Estimating the probit model, specified in equation (1.14) raises the following results (table 1.13). As it can be inferred, the sample has been divided into treated and control areas, where treatment is defined as the implementation of the first stages of the land reform. The result is in line with the theoretical proposition 3, since for those individuals that rank below the village median and hence can be classified as low productivity, an one unit raise in their productivity would affect positively the probability of the individual obtaining more land in 2012 compared to those individuals that rank above the village median. The interesting part is that this effect is not observed in the control areas. For the control areas not only the effect of the interaction term is insignificant but its sign points toward an efficient allocation of land, in which low productivity individuals lose land.

68 Chapter 1 Essays on Macroeconomics and Development

Table 1.13: Results from (5)

Dependent variable:

Pr(increase = 1| X) treated control

(1) (2)

below-median×ind.productivity 0.123∗ -0.058 (0.075) (0.098)

Individual controls Yes Yes Household control Yes Yes Observations 1,324 1,230

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

Chapter 1 69 Chapter 2

Mutual insurance and land security in rural Ghana

joint with Karol Mazur (University of Oxford)

70 Essays on Macroeconomics and Development

“ Every individual is born into a certain status relative to all other members of the community. It is not a personal status, as it might be in an individualistic society, and does not imply rank, but reciprocal obligations and beneﬁts which he incurrs as a member of the community”

— J. H. Driberg, (1934), The African conception of law

2.1 Introduction

The agricultural sector in Ghana constitutes one of the major employers at the country level. Almost 45% of the Ghanaian labour force is engaged in agricultural activities translating into more than 7 million individuals operating a farm (Oxford Business Re- port Ghana, 2020). Despite the sector’s economic importance, agriculture in Ghana is predominantly small-scale, managed under traditional rules and norms, and characterized by negligible mechanisation and irrigation investments (SRID, 2001).

As in the broader African context, extended families and traditional local authorities in Ghana have extensive power over organization of public life and provision of local common and public goods. Ghanaian land management and ownership system is a particularly striking example of economic institutions based on overlapping customary and statutory rules (Michalopoulos and Papaioannou, 2020). Even as of today, around 80% of Ghanaian land is held under customary law tenure (Fenrich et al., 2011). Although details of this customary law governing land conveyances vary between diﬀerent parts of the country, its one invariably common feature is the predominant and ample role of chiefs and family heads in organizing public life of rural communities, and especially allocating land and settling tenure issues.1

Furthermore, because of its characteristics, a lot of agricultural production in Ghana is aimed at satisfying rural population’s subsistence requirements. Given severity of consumption ﬂuctuations in such an environment coupled with the absence of government- provided social insurance, village communities have devised safety nets based on cash and in-kind transfers, gifts or loans. While not regulated by any statutory law, these insurance schemes are largely based on reciprocal interactions taking place in close-knit societies. As such, they prove to be robust to default and consistently provide resources at a ﬂexible basis to the ones in need.

In this study, we explore how land security aﬀects mutual insurance and land re-allocations in rural Ghana. We do so by tracing the economic interaction between (i) the take-up of

1See the evidence from Afrobarometer 2008-2009 survey compiled in ?.

Chapter 2 71 Essays on Macroeconomics and Development land titles providing a formal proof of land ownership; and (ii) local institutions governing land re-allocations and mutual insurance nets between neigbours and relatives. We hy- pothesize that understanding this interaction is particularly important due to the fact that land rights may have significant impact on informal institutions in rural Ghana through affecting incentives for this kind of voluntary co-operation. In particular, stronger land rights may allow for more secure sharing of land with potentially more productive members of the society, leading to higher overall agricultural output. On the other hand, by structurally altering the nature of outside options for all community members, land titles may affect incentives for co-operation over other margins, such as the mutual insurance.

We address the question in hand both empirically and theoretically. On the empirical front, we use the Ghana Socioeconomic Panel Survey (Yale EGC-ISSER) on agricultural and economic activities of Ghana’s rural population. This rich dataset allows us to test empirically the link between land rights and land disputes, land fluidity, agricultural productivity, consumption and, ultimately, informal risk sharing arrangements. We first show that the number of land disputes due to multiple claims declines with increases in the degree of land rights (measured as the share of village’s land with titles). Second, we document that land re-allocations are more intense in rural areas experiencing increases in land formality. Third, we show that increases in land security provided by titles also strongly correlate with increases in overall village productivity, signifying that informal land markets become not only more fluid, but also more efficient. Fourth, our measure of land formality translates further into higher aggregate consumption at the village level. Finally, we uncover improvements in risk sharing in communities with stronger rights over land. Taken together, this empirical evidence suggests that, although not necessarily utilized in equilibrium, land titles may well complement traditional village institutions, such as mutual insurance and land re-allocations analyzed here.

In the second part of our analysis, we build a dynamic model rationalizing our empirical findings. Our village economy features co-operation over insurance transfers and land re-allocations in the presence of voluntary participation (or limited commitment) constraints. We assume that the (endogenously chosen) level of land rights alters the amount of land securely owned by the household, whether cultivated or rented out. Consequently, this assumption alters the households’ outside option attainable had they chosen to break out of the co-operation with the rest of the community. This component is crucial to understand the nature of informal co-operation patterns under scrutiny as it effectively endogenizes the functioning of informal institutions. For sufficiently low values of land right, the dynamics generated by the model are entirely in line with the empirical evidence presented. As land security increases, land allocations become more efficient, aver-

72 Chapter 2 Essays on Macroeconomics and Development age income and consumption increase, and consumption smoothing against idiosyncratic productivity shocks is improved. However, there exists a level of land rights above which the co-operation over risk and land sharing unravels, pushing the whole society into a bad equilibrium where they can still trade land, but have to rely on self-insurance only. Thus, our model uncovers potentially non-linear eﬀects of land formalization eﬀorts and as such provides a candidate explanation for strong persistence of informal institutions in rural areas of developing countries.

The work unfolds as follows. In section 2.2, we present the related literature. In section 2.3 we describe the customary norms surrounding land conveyances in rural Ghana. In section 2.4 we exhibit how we constructed the main variables of interest from the dataset in hand, we present certain suggestive evidence of the main mechanisms under examination and we proceed in showing the formal empirical analysis. In section 2.5 we present the quantitative model and associated results. We conclude in section 2.6.

2.2 Literature Review

The observed large agricultural productivity gaps in developing countries (Gollin et al., 2014) have incited various explanations. One of the most recent and predominant ones is land misallocation due to missing formal land markets. For instance, exploiting rich household-level data from Malawi, Restuccia and Santaeulalia-Llopis(2017) find substantial misallocation of both land and productive capital. The authors attribute the observed misallocation to the absence of formal land markets, as land in Malawi is allocated mostly by village chiefs. Chen et al.(2017) associate the level of land rental markets to the degree of misallocation in Ethiopia, by exploiting a land reform granting farmers the right to rent out land. They find that the existence of rental markets correlates with lower degree of misallocation while it boosts productive technological investment and agricultural productivity. Adamopoulos and Restuccia(2019) study a land reform in Philippines which extensively redistributed land above a certain threshold and subsequently banned market transactions of the redistributed land. They estimate quantitatively the counterfactual of complete land markets and they infer a significant decrease in the negative impact the land reform exerted on both farm-size and agricultural productivity. Chari et al.(2017) examine the effects of a land reform in China, that allowed farmers to lease their land. They find that the reform led to an increased participation in formal land markets, im- proving land allocation and raising aggregate land productivity. In a similar vein, Chen (2017) attributes international agricultural productivity differences to the existence of untitled land in the developing world. In his work, land misallocation is the result not only of missing land markets, but also distortions to individual occupational choices created

Chapter 2 73 Essays on Macroeconomics and Development due to the existence of untitled land.

Related to the above is the literature studying broader role of land rights on agricultural investment and productivity. Due to the unclear land tenure regime observed in the developing world, land rights might be endogenous to a variety of factors and especially to investment, thus resulting to conflicting results regarding their role. Besley(1995) studies theoretically links between land rights and investment incentives and finds some empirical support for the positive feedback of land rights for investment in the data from two regions in Ghana. Exploiting a land tenure reform regulating sharecroppers’ rights over land in West Bengal, Banerjee et al.(2002) find that increased land security (even though not full) enhance agricultural productivity. Goldstein and Udry(2008) assert that higher land security stemming from higher local political positions, result in higher investment and output. On the other hand, Carter et al.(1994) argue that increased land rights might not deliver the desired productivity gains, due to lack of access to land markets for a large part of the rural population. Employing data from Burkina Faso, Brasselle et al.(2002) evince that there is no clear evidence of a positive effect on agricultural investment from increased formal land rights. The latter work as well as Townsend(1994), Udry(1994) and Gollin and Udry(2019) claim that across the developing world informal institutions can substitute for market frictions within weak institutional environments leading to some sort of constrained efficient allocations.

We contribute to this literature in two ways. First, we do not restrict our analysis only to the degree of land markets development but we directly assess the impact of land formalization efforts on the intensity of overall (formal or not) land re-allocations. Second, we investigate the impact of these efforts on another, so far overlooked in the literature, key margin of mutual insurance. Third, we identify land rights as a determinant of functioning of rural informal institutions that may at first sight seem unrelated to land markets (such as mutual insurance networks). Fourth, our quantitative model shows that effects of increasing land rights may be highly non-linear and can lead to a break up in community co-operation.

In order to develop the informal mutual insurance framework we build on the risk-sharing with limited commitment literature. Ligon et al.(2002) were among the ﬁrst to explain the observed imperfect risk-sharing in village economies with limited commitment constraints. Attanasio and Rıos-Rull(2000) in their work they ﬁnd parameter spaces where public insurance can both improve or deteriorate risk sharing. Mazur(2020) ana- lyzes interaction between informal insurance against idiosyncratic shocks and village- or government-managed irrigation reducing aggregate shocks, and shows that the two may complement or substitute each other depending on the irrigation’s management structure.

74 Chapter 2 Essays on Macroeconomics and Development

Manalis(2019) assesses the effect of a land reform granting individual land rights on rural communities’ mutual insurance and land re-allocation arrangement and documents the arising trade-off between output efficiency and extent of risk-sharing. Finally, Morten (2019) and Meghir et al.(2019) employ related models of migration and risk sharing with limited commitment in order to evaluate interaction between these two margins.

Our contribution to this strand of literature is that we account for land not only as the major (and often sole) production factor, but also as an important common good inﬂuencing the degree of risk-sharing within local communities. Land is the common feature underlying mutual and self-insurance thus linking the informal contract to the outside alternative. In this way, we are able to capture more accurately the interaction between informal institutions and formal land markets in rural Ghana as well as evaluate its implications on agricultural productivity.

2.3 Statutory and customary land institutions in Ghana

Land management in Ghana is predominantly a combination of statutory and customary practices. Those two pillars interact and overlap with each other, thus creating a dynamic and ﬂuid institutional framework governing land issues. Whilst Ghana in its post-colonial years went through much turmoil regarding its formal legal system, customary norms and especially family land law has remained vastly accepted by the population, regardless of national politics (Woodman, 2003).

During the recent legal history of Ghana, notable advancements in the legal framework regarding land were the National Land Policy in 1999 and the subsequent first and second Land Administration Projects in 2003 and 2010 respectively (Kline et al., 2019). Those acts constitute a first attempt from the Ghanaian government to establish an efficient land administration system and led to the current legislation as delineated by the Land Bill of 2016. The latter act advances Ghanaian land management by ensuring equitable access to land for women, registering and protecting customary land and facilitating land conveyances among other provisions. An indication of the focus on customary norms by statutory law is that the Constitution of 1992 formally recognizes customary land ownership as long as the state law is not violated (Kline et al., 2019). Moreover, the law even though does no longer recognize legal courts centered around Ghanaian Chiefs, customary groups are allowed to manage the so called Alternative Dispute Resolution (ADR) programs that mediate disputes arising among members of the community (Kline et al., 2019).

Under customary norms, land can either belong to “Stool lands” or to “Skin lands”. Those

Chapter 2 75 Essays on Macroeconomics and Development two types diﬀer in the rules conditioning inheritance of land. The former are inherited to heirs of patrilineal descent, while the latter of matrilineal. “Stool land” is usually found in the southern part and refers to the wooden carved stool. “Skin land” is found in the northern part of the country and refers to animal’s skin. In both cases, these are invariably symbols of the Chieftanship, signifying the importance of customary rules for land management in rural Ghana.

Driberg(1934) emphasizes the diﬀerence between the European and African concepts of law by highlighting their disparate foundations. European law is founded on the individualistic assumption, while the African law is based upon collectivist organization. This exact collectivist view characterises land management under customary law in Ghana. Land might either belong to the community, to the village or the family but never to the individual (Daniels, 1996). At the same time, all Ghanaian land is owned by someone (Kline et al., 2019).2 As a result, under customary norms the head of the extended family or a traditional leader are entitled to extended power over land among other issues.

Our work builds on the premise that customary land management exhibits a certain degree of eﬃciency. Land is allocated among members of the community as to maximize total agricultural output of the village and can be re-allocated if this is in the best collective interest. Moreover, communities manage to insure their members against adverse shocks by a system of reciprocal transfers among them. Turpin(1963) in a review of the Ollennu et al.(1962) book on customary land law in Ghana, writes:

In illustration of the eﬃciency of this law in ensuring the fullest exploitation of land may be mentioned the reversion to the donor or vendor stool (community) of land not reduced to occupation by subjects of the acquiring stool; the unlimited right of a subject to exploit unappropriated land; the necessity for continuous rights of members of the general community (stool or family) to the natural produce of stool or family land; and the right of a pledgee of land to exploit it for his own beneﬁt.

This extract succinctly summarizes a customary system which prescribes land re-allocation among members of the community according to their needs and in order to fully utilize land. Additionally, every member of the extended family is entitled to the produced agricultural output, if in need. This collectivist organizational mode is embedded in our approach to the research question.

2“There is a Ghanaian saying to the eﬀect that there is no land without an owner” (Daniels, 1996)

76 Chapter 2 Essays on Macroeconomics and Development

2.4 Empirical analysis

A Data

For the empirical component of the work, we use the Yale EGC-ISSER Ghana Socioeco- nomic Panel Survey, providing regionally representative data for all 10 regions of Ghana. In total, 5010 households from 334 enumeration areas (EAs) were interviewed in two waves (2009 and 2014).3 We drop data from urban areas, leaving us with a sample of 4512 households living in 198 villages.4 The dataset provides comprehensive information at the plot, individual and household level regarding consumption, income, ﬁnancial and farm assets, labour, land, agricultural and business activities and many more.

Income and consumption

We construct our measures of income and consumption using the detailed information on several sources and forms of both recorded in the dataset. First we combine information on the main and secondary occupation income of the sample. This is reported for the past 7 days from the interview date. We recover the annual value of the main and secondary occupation income by multiplying their income in the past 7 days by the number of weeks they have been employed during the past year.

The main source of income within our rural sample comes from crops sales recorded during the major and minor seasons, which span the whole calendar year. We construct value of crop sales by exploiting information on sold quantities and amount of money received in exchange. On the other hand, land rentals are surprisingly low both in the extensive and intensive margin.

In order to capture broader sources of household income, we also use the data on non- farm enterprise income. This refers to income stemming from operating a small business outside the agricultural sector. The survey asks for detailed report of enterprise income and costs on a monthly basis for the past year. As a result, combining this information with the percentage of enterprise income that belongs to the household, we obtain a complete picture of non-farm enterprise income per annum.

The ﬁnal source of potential income is from household assets. In our survey, household assets are categorized into farm assets and ﬁnancial assets. We include in our income

3The third wave of the dataset is expected to become available sometime in 2021. 4In principle, the panel contains data from 200 villages over two waves, but we lose 59 village- wave observations (because of these villages having no household providing answers about land titles). Similarly, the full ample has 4723 household-wave observation, but we lose 211 records due to missing data about non-ﬁnancial income.

Chapter 2 77 Essays on Macroeconomics and Development measure the former category, which includes animals, farming tools and durable farming inputs. However, we exclude from it ﬁnancial assets (savings, borrowing, lending, remittances, in-transfers and out-transfers received or sent), since they do not stem from productive activities and so are likely to act as means of insurance.

As far as consumption is concerned, the Ghana Socioeconomic Panel Survey reports consumption of food (purchased, home produced or received as gift) and expenditures on clothing. In particular, the food list consists of 85 items and provides information on the way those were obtained. The consumption section also includes expenditures on durable goods with two additional categories: “other items” (including expenditures on ceremonies, vehicles and household repairs); and “fuels” (including electricity, gasoline and crops byproducts). Those two expenditure categories are reported for the past 12 months.

Our consumption and income variables are sums of all the above mentioned items. In order to obtain an annual measure for the food and clothing expenditures which are reported for the last month, we simply multiply by 12.5 Finally, we express all income and consumption variables in per capita terms.6

Land rights measure

A primary pillar of our work is how easily farmers in rural Ghana can access formal land markets. The Ghana Socioeconomic Panel Survey provides the plot-level information whether the current user carries the right to sell the parcel, or not.

As most of our analysis below pertains to co-operation at the community level, we aggregate selling rights information at the village level using a weighted (by the relative size of the plots) share of plots that can be sold in the sample: X κp,t sell-rightsv,t = sell-rightsp,t P (2.1) p∈v p∈v κp,t where p is plot, v is the village (EA), sell-rightsp,t is a dummy variable indicating whether

5We acknowledge the seasonality that this measure suﬀers from. In particular, there is large depen- dence on whether the interview took place before or after the harvest season. In order to deseasonalize this measure we are planning to employ the method of De Magalhães and Santaeulàlia-Llopis(2018) who regress period consumption on seasonal dummies. However, our dataset only contains two periods, and more than two calendar years are needed. As a result, we are planning to perform this upon publication of the third wave of the survey. 6We compute per capita quantities by dividing household-level quantities by a composite age-adjusted measure of household members, derived as in Townsend(1994) by adding the following numbers: for adult males, 1.0; for adult females, 0.9; for males aged 13-18, 0.94; for females aged 13-18, 0.83; for children aged 7-12, 0.67; for children aged 4-6, 0.52; for toddlers 1-3, 0.32; and for infants 0.05.

78 Chapter 2 Essays on Macroeconomics and Development

the user of the plot has selling rights over it, κp,t is the size of the plot in hectares.

Land ﬂuidity measure

In order to formalize the degree of land variation we build a measure of size variation at the village level by exploiting the panel structure of our dataset. Our land fluidity index (LF Iv) is defined as a normalized (by the total size of land at the EA level) sum of squared differences in households’ landholdings over time.

2 P h∈v landh,t+1 − landh,t LF Iv = (2.2) landv,t/t+1 where landh,t is the sum of size of all plots cultivated by members belonging to the same household in waves 1 (t) and wave 2 (t + 1), and landv,t/t+1 is the mean of plots’ size for all plots within the village (EA) in both waves.

Productivity measure

In order to construct our measure of agricultural productivity, we process data on crop sales prices, produced quantities and land size to construct a comparable measure of productivity based on the market value of agricultural production. Productivity of plot p of size κp in which individual i grows crops c in wave t is calculated as the market value of produced crops per unit of land cultivated, i.e. using the following formula:

pc,t · qc,p,t plot − prodp,t = ∀p, t (2.3) κp,t where pc,t is the crop market price in wave t and qc,p,t is the produced quantity of crops c.

Two characteristics of Ghanaian agricultural markets impede a direct derivation of productivity. First, very few farmers do actively participate in the crop market. Harvested crop is mostly directed to self-consumption, or is gifted to other households if not lost due to pests or weather conditions. As a result, large part of the sample reports high produced quantity but low or none sold quantity. This fact hinders the derivation of market prices. To overcome this diﬃculty we proceed as follows. If the individual i has sold part of the harvest in the market, then pc ≡ pi,c, the price in which i sold. If individual i has not participated to the market, we ﬁrst generalize at the village level by setting the crops’ price pc ≡ pc,v, which is the average price of all other farmers in the same village that sold the same crops.7 If no other farmers from the same EA (village) have participated to the

P p−i,c 7 −i∈v More precisely, pc,v is deﬁned as pc,v = , where −i indicates all other households living n−i in village v other than household i.

Chapter 2 79 Essays on Macroeconomics and Development market selling the same crops we generalize at the district, regional and country level, by setting pc equal to pc,d, pc,r and pc,g respectively. In this way we manage to derive prices for as many individuals as possible.

The second characteristic is the existence of multiple local units of measurement. Because of this, the sample records transactions for a given crop denominated in multiple units of measurement. In the described process, we need to account for diﬀerent units of measurement. Thus, we create a three level id, EA-crop-unit, which allows us to match the same crops expressed in the same unit among diﬀerent individuals residing in the same village.8

Lastly, throughout the regression analysis, we aim to explore the link between land rights and productivity at the macro-village level. To this end, we construct the EA productivity metric using the following formula: X pc,t ∗ qc,p,t κp.t village-productivityv,t = P ∀t, v (2.4) p∈v κp,t p∈v κp,t where each plot’s productivity belonging to a certain EA has been weighted by the relative size of the plot to the EA’s total land size.

B Suggestive empirical observations

Before delving into formal econometric evidence, we motivate our study by providing descriptive statistics on the main elements of our work. The left bar plots in Figure C.1 document that land acquisition in rural Ghana is mostly informal and conﬁned within the limits of the extended family. Only 15% (20%) of the respondents have obtained their cultivated plots from a person to whom they are not related in wave 1 (wave 2). At the same time, more than 65% of the sample in wave 1 have obtained the plot either from the family head (∼ 30%) or a relative (∼ 35%). In wave 2, where only the option “relative” is provided, more than 70% of the plots were given to current users by a relative.

Furthermore, the right bar plots in Figure C.1 evince a complete absence of formal land sales. Purchased plots amount to less than 5% of the sample, while there is a strong indication of plots being transferred through lineage, as inheritance constitutes the main acquisition way with more than 40% (50%) in wave 1 (wave 2).

Despite the absence of formal land markets, households’ land holdings are experiencing considerable variation in size, signifying the mechanism of informal land re-allocation. In Figure C.2, we present changes in aggregate land holdings at the household level across

8We proceed accordingly at the region, district and country level with the id accounting for the granularity level

80 Chapter 2 Essays on Macroeconomics and Development the two waves. As most of the observations lie oﬀ the 45 degree line, household land holdings go through signiﬁcant changes over time, an observation that is consistent across all ten Ghanaian regions.

Furthermore, Figure C.3 shows that while the share of villages’ land with selling rights on it is small on average, it is varying signiﬁcantly both over time and regions. This conﬁrms existence of pluralistic land tenure regime documented in the literature.9

Finally, because the central focus of this paper is on the link between land formality and risk-sharing, we document how the communities in our sample benefit from and provide insurance. Firstly, we show in Figure C.4 that vast majority of households in our sample has financial savings lower than 20% of annual consumption and that this does not change much over time. Secondly, Figure C.5 documents that more than 60% of the loans given are coming from relatives, neighbours and friends. This signifies tight links among members of the same extended family and community, as well as a lack of well-developed formal credit markets. Complementary to the latter evidence, Figure C.6 shows that households in our sample mostly used the received loans either for consumption or to purchase agricultural inputs. Lastly, we observe in Figure C.7 that more than 80% of the loans did not require any form of collateral. This finding evinces that loans are given on the basis of trust and potentially constitute an informal insurance mechanism developed within the community.

C Regression analysis

In this part, we analyse econometrically the channel through which land rights aﬀect co- operation over land use and mutual insurance at the village level. This investigation will largely rely on rich empirical variations documented above.

Intuitively, increases in land formality may give rise to two opposite forces. On one hand, stronger selling rights might undermine informal land and risk sharing co-operation within communities by establishing formal land markets that make the title holders better-off on their own, and so effectively reduce incentives of these individuals to co-operate with community. On the other hand, a farmer who feels more secure about her or his rights over land, might be more willing to share it (via renting or sharecropping) with other households, if efficient (or desirable). We will show that the latter force dominates in our

9A potential concern with this data is that part of our measured changes in land formality is due to attrition of plots between waves of our data. In order to address this issue, Figure C.10 presents equivalent of Figure C.3, that has been constructed using a subsample of plots that households reported to have been consistently cultivating between the ﬁrst and second wave. As can be seen, the general patterns of changes in land formality survive.

Chapter 2 81 Essays on Macroeconomics and Development sample (but we will also show that the former exists in our quantitative model).

In order to build our argument we proceed as follows. First, we show that in villages with a higher share of selling rights, land changes hands more often. This suggests that land holders feel more secure about their assets. Second, we show that in those villages, the average village productivity is also higher, suggesting that more intense land re-allocations are more eﬃcient. Third, we conﬁrm that increases in selling rights are positively related to increases in average village consumption. Finally, because of the increased agricultural productivity and implied higher consumption, we document that increases in land formality improve informal insurance against idiosyncratic shocks (or risk sharing).

Land rights and land security

First, we argue that higher take up of land titles translates into higher land security in communities. To this end, we regress the village-level share of land with disputes due to multiple claims on our measure of selling rights (2.1):

multiple − land − claimsv,t = α + β1 · sell-rightsv,t + β2 · Xv,t + v,t (2.5)

where Xv,t is a vector of time-varying village and land controls containing size of village population, average years of use of land, share of land that can be fallowed, share of land with chemical use and share of land with irrigation.

Results in Table 2.1 show that the increases in the share of land in a village with formal titles is signiﬁcantly negatively correlated with the number of land disputes due to multiple claims over land. As intuitive as it is, this ﬁnding suggests that land titles provide higher security for land owners, which will have important implications for the ensuing analysis below.

Land rights and land ﬂuidity

In order to examine the role of selling rights in village-level land re-allocation, we regress our measure of land ﬂuidity (2.2) on our measure of selling rights (2.1) as follows:

LF Iv = α + β1 · sell-rightsv,t/t+1 + β2 · Xv,t/t+1 + v (2.6)

Notice that regression (2.6) is based on the cross-sectional data. This is due to the fact that the land ﬂuidity index is employing the diﬀerence in land holdings’ size over time. We adjust the right hand side of equation (2.6) by taking the average selling rights at the village level over the two available waves (sell-rightsv,t/t+1). Similarly, vector Xv,t/t+1

82 Chapter 2 Essays on Macroeconomics and Development includes the average values over the two waves for multiple land characteristics we control for.

Table 2.2 presents results of the estimation. We observe that higher values of selling rights at the EA level are correlated with higher values of the LF I index. This evinces that land markets in villages with more selling rights tend to be more ﬂuid. Importantly, Table 2.8 shows that increases in land rights do not trigger increases in formal land sales. As such, our ﬁndings suggest that households are more inclined to share land informally, when they hold formal rights.

Land rights and productivity

Next, we investigate how improvements in land security aﬀect agricultural productivity at the aggregate level. We investigate this using the following regression:

log(village-productivityv,t) = α + β1 · sell-rightsv,t + β2 · Xv,t + v,t (2.7) where the productivity measure is as described in (2.4).

Table 2.3 shows the results of estimation. We observe that villages with higher level of selling rights tend to be more productive. These gains realize because land is reallocated more intensively, and - as it turns out - more eﬃciently.

Land rights and average consumption

Furthermore, we investigate whether increases in aggregate productivity translate into increases in aggregate consumption using the following regression:

cv,t = α + β1 · sell-rightsv,t + β2 · Xvt + γd,t + v,t (2.8)

where cv,t is the mean per capita consumption at the village level and γd,t are district-wave ﬁxed eﬀects controlling for district-level aggregate shocks.

Table 2.4 shows that the degree of land rights at the village level is indeed positively associated with the mean village consumption. Because this result informs us only about the aggregate eﬃciency impact of increases in land security, we turn to investigating the consumption risk sharing channels in what follows next.

Cross-sectional evidence: risk-sharing ratio

We provide evidence on informal risk-sharing in two ways. First, we construct a risk- sharing ratio accounting for the cross-sectional variation in consumption and income at

Chapter 2 83 Essays on Macroeconomics and Development the village level. We deﬁne it as follows:

Varh(ch,v,t) RSv,t = (2.9) Varh(yh,v,t) where ch,v,t and yh,v,t are the per capita consumption and income of household h in village v at time t.

Notice that the lower is the risk-sharing ratio, the better is the insurance against income shocks in the cross-section at the village level. For instance, a value of RSv,t equal to 1 signiﬁes that consumption and income exhibit the same variation. Furthermore, a negative relationship between RSv,t and sell-rightsv,t implies potential improvements in risk-sharing as land markets become more formal.

We test whether increases in land formality are associated with improvements in the risk-sharing ratio using the following speciﬁcation:

log(RSv,t) = α + β1 · sell-rightsv,t + β2 · Xv,t + (2.10)

where Xv,t is a vector of our village and land controls extended by mean village consumption allowing us to control for aggregate shocks aﬀecting all villagers in a given period, and the main independent variable is the degree of selling rights at the EA level as deﬁned in expression (2.1).

Table 2.5 presents the results from estimating equation (2.10). Under all specifications, which differ at the calculation of the risk-sharing ratio components of consumption and income (in levels, and per capita) the degree of formal land rights is negatively correlated with the risk-sharing ratio, suggesting that increases in selling rights may not only bring efficiency gains, but also contribute to reductions in consumption inequality.

Panel evidence: consumption smoothing test

In our second measure of risk sharing, we exploit the panel dimension of the data allowing us to control for household fixed effects and run an extended consumption smoothing test in spirit of Townsend(1994). In particular, we estimate the effect of sell-rightsv,t on the elasticity of consumption with respect to idiosyncratic income shocks as follows:

ihs(ch,v,t) = α+β1·ihs(yh,v,t)+β2·ihs(yh,v,t)·sell-rightsv,t+β3·sell-rightsv,t+β4·Xh,v,t+h,v,t (2.11) where ch,v,t and yh,v,t are per capita consumption and income of household h, in village v at time t. We have used the inverse hyperbolic sine transformation for the dependent variable in (2.11) to numerous observations with negative income, while preserving properties

84 Chapter 2 Essays on Macroeconomics and Development

10 of log-transformation. Xh,v,t is a vector of land controls and mean consumption of every other household in the village accounting for aggregate shocks (ihs(cv,t)). Notice that controlling for the average village consumption (ihs(cv,t)), allows us to interpret the coeﬃcient β1 as the elasticity of consumption with respect to idiosyncratic income shocks.

Results in Table 2.6 show that households in our sample are able to smooth their consumption very well. First, we find that the elasticity of consumption is low, but statistically significant. Second, we see that selling rights are positively correlated with the level of household consumption (confirming our findings above). Third, and in line with evidence from the risk sharing ratio, we see that the income and selling rights interaction term is statistically significant and negative, implying that increases in selling rights improve consumption smoothing.

Importantly, notice that consumption smoothing can be achieved either through self- insurance or mutual-insurance. To show that our results on risk sharing do not reﬂect self-insurance, we test whether households’ stock of savings responds to changes in land formality:

savingsh,v,t = α + β1 · sell-rightsh,v,t + β2 · Xh,v,t + ζv,t + h,v,t (2.12) where savingsh,v,t is the household’s h stock of savings in period t either kept at home, in a bank or in any other saving society or group (e.g. Susu). Importantly, now sell-rightsh,v,t indicate the selling rights at the household level. Moreover, due to cattle being the predominant form of assets in rural Africa, we estimate the same specification with the value of cattle owned by the household as the dependent variable. Under both specifications and as shown in Table 2.7, selling rights exhibit no significant relationship with household savings. These findings allow us to conclude that the documented above improvements in consumption smoothing upon increasing land formality are not due to better self- insurance, but rather reflect informal risk sharing channels in rural communities.

The empirical analysis so far aims to establish an empirical relationship between land rights and risk-sharing. Those components interact through an informal mechanism of joint consumption and land sharing. In particular, a higher degree of rights over land, as expressed by the ability of the user to sell the parcel, ensure farmers of the safety of their land. They are less afraid of losing land thus they are more willing to participate to land re-allocation and share their land more eﬃciently bringing productivity gains and higher agricultural output that can consequently be redistributed for consumption of the ones hit by bad shocks.

10In particular the inverse hyperbolic sine transformation is deﬁned as ihs(x) = ln(x + (x2 + 1)0.5), which accepts zero values of x.

Chapter 2 85 Essays on Macroeconomics and Development

2.5 Quantitative model

In this section, we propose a dynamic model featuring co-operation over insurance and land in presence of voluntary participation (or limited commitment) constraints. We characterize the outside option (of self-insurance and land trade), ﬁrst best and limited commitment benchmarks, and we argue that the framework proposed can rationalize all of the empirical ﬁndings documented above.

A Environment

The village economy is inhabited by two ex-ante homogeneous households deriving utility from consumption according to function u(c), satisfying u0(c), −u00(c) > 0. Time t is discrete and agents are inﬁnitely lived.

In every period t, household i uses land zi,t (chosen in period t−1) to produce crop output according to the following production function:

α yi,t = φtθi,tzi,t (2.13)

where φ ∈ Φ = (φ1, ..., φNφ ) and θi ∈ Θ = (θ1, ..., θNθ ) denote aggregate and idiosyncratic productivity shocks following Markov chains with transition matrices πφ and πθ.

More importantly, the supply of land is ﬁxed in every period, i.e. z1,t + z2,t = 1 ∀t. Finally, because we assume ex-ante homogeneity, households start their lives with the inherent allocation of land z1,0 = z2,0 = 0.5.

The state of the economy is given by vector xt = (z1,t, z2,t, φt, θ1,t, θ2,t) and the decisions of the households are at = (c1,t, c2,t, z1,t+1, z2,t+1).

B Outside Option

We begin by characterizing the outside option of households. First, we assume that households have access to formal land titles, and so they acquire and sell land through participation in the formal land market.

Second and related, we assume self-insurance: agents can consume and store (in form of land) as much as they want, without sharing their goods with other households.

Third, we assume that in the very period of household’s deviation tdev from the contract on land and risk sharing (LRS, described below), the deviator own the claim to and produces zout ψ · z − ψ · zLRS ψ ∈ , using land equal to i,tdev = i,0 + (1 ) i,tdev (with [0 1] standing for the degree zLRS of land rights and i,tdev being exogenously given here). In other words, the higher is the degree of land rights ψ, the higher is the weight on the household’s inherent land in it’s

86 Chapter 2 Essays on Macroeconomics and Development portfolio in the period of deviation. From the second period of deviation onward (i.e. for t > tdev), the deviating household keeps the title to economic beneﬁts of the entirety of land they possess.

Formally, given the degree of land rights ψ, the period’s tdev value of outside option for i x zout , θ the household with current state i,tdev = ( i,tdev i,t), that decides to deviate from the contract, reads:

out out out i,tdev Vi,t (xi,tdev ; ψ) = max u(ci,t ) + βEθVi,t +1(xi,tdev+1|x ) dev {cout ,zout } dev dev (2.14) i,tdev i,tdev+1 subject to:

α out out out out t = tdev : ci,t + qz,tzi,t+1 ≤ φtθi,t zi,tdev + qz,tzi,tdev (2.15)

out out out α out t > tdev : ci,t + qz,tzi,t+1 ≤ φtθi,t(zi,t ) + qz,tzi,t (2.16)

out LRS LRS where zi,tdev = ψzi,0 + (1 − ψ)zi,t , xi,tdev = (ψzi,0 + (1 − ψ)zi,t , θi,t) and xt+1 = out (zi,t+1, θi,t+1) and qz,t+1 is a land price such that zi,t ∈ [0, 1] ∀i and the market clearing condition z1,t + z2,t = 1 ∀t holds in equilibrium.

The associated land-FOC reads:

ui,ct+1 out α−1 qz,t = βEθ (αθi,t+1(zi,t+1) + qz,t+1) (2.17) ui,ct

Land holdings are chosen in (2.17) such that the marginal cost of an additional unit of land (that costs qz,t) is equalized with its expected marginal gain, given by the marginal increase in crop output and the capital value of land (all in terms of the marginal utility of consumption).

Notice that the degree of land rights’ chosen has important implications for the value function of deviators:

Proposition 1. V out x ψ ψ zLRS < z i,tdev ( i,tdev ; ) is increasing in for the household with i,tdev i,0.

Intuitively, higher land rights increase land security of agents: whenever they deviate from the co-operation with other villagers, they maintain rights to their inherent amount of land. Consequently, this implies that the value of deviation for the individuals with a relatively high (low) land allocation within the LRS contract would decrease (increase) with improvements in land rights. This dynamic will be very important for characterization of the LRS contract below.

Chapter 2 87 Essays on Macroeconomics and Development

C First best

In the ﬁrst best allocation with Pareto weights λ1, λ2, the social planner’s (village chief or family head) problem reads:

2 FB X FB FB t V = max λiu(c ) + βEθV (xt+1|x ) t FB FB i,t t+1 {ci,t ,zi,t+1} i=1 2 2 X FB X FB α (2.18) subject to : ci,t ≤ φtθi,t(zi,t ) i=1 i=1

z1,t + z2,t = 1, zi,t ∈ [0, 1] where si,t+1 ≥ 0, xt+1 = (z1,t+1, z2,t+1, θ1,t+1, θ2,t+1, φt+1). Notice that prices qz,t are not present in this benchmark as they are not relevant to the planner choosing land allocations directly (subject to the land market clearing condition).

Then, the associated FOCs read as follows:

u1,c λ2 t = (2.19) u2,ct λ1

α−1 α−1 Eθ1,t+1 [u1,ct+1 θ1,t+1]λ1z1,t+1 = Eθ2,t+1 [u2,ct+1 θ2,t+1]λ2(1 − z1,t+1) (2.20)

This benchmark diﬀers in three important ways. First, the planner directly redistributes consumption between agents according to (2.19) with exogenously given Pareto weights.

Second, land is allocated according to (2.20), such that the marginal output gain of an additional unit of land given to household 1 is equalized with the one of household 2 (all in marginal utility terms). In other words, the planner ﬁrst ensures that the land is allocated in the most productive way maximizing the aggregate crop output, and then redistributes the realized output according to her preferences.

Third, the land rights are irrelevant in this benchmark. This is a direct consequence of the fact that the planner is assumed to be benevolent and to be able to enforce all allocations she desires. We relax the latter assumption in what follows.

D Land and risk sharing with limited commitment

This benchmark differs from the first best in that it subjects the planner’s decisions to the limited commitment constraints. Satisfying those requires the planner to make such consumption and land transfer decisions that make both households participating in the contract in every period and state realization at least as well off as in the outside option.

88 Chapter 2 Essays on Macroeconomics and Development

Formally, the planner solves the following problem:

2 LRS X LRS LRS 1 V (ψ) = max λ u(ci,1) + βEθV (x2|x ) (2.21) 1 LRS LRS i,0 i,2 {ci,t ,zi,t+1} i=1 subject to: LRS X LRS X LRSα ζt ci,t ≤ φtθi,t zi,t (2.22) i i

∞ LRS X t0−t LRS out LRS t µi,t Et β u(ci,t0 ) ≥ Vi,t ψ · zi,0 + (1 − ψ) · zi,t , θi,t ∀i, x (2.23) t0=t

z1,t + z2,t = 1, zi,t ∈ [0, 1]

LRS where µi,t denotes the Lagrange multiplier on the limited commitment constraint for household i in period t.

In this case, the FOCs read:

LRS LRS LRS LRS LRS u1,ct λ2,t−1 + µ2,t ct : ζt = ui,ct · λi,t−1 + µi,t ∀i ⇒ = LRS LRS ∀i =6 j (2.24) u2,ct λ1,t−1 + µ1,t

" out # LRS LRS LRS α−1 LRS ∂V1,t+1 z1,t+1 :Eθ u1,ct+1 λ1,t + µ1,t+1 θ1,t+1α z1,t+1 − µ1,t+1 LRS = ∂z1,t+1 " out LRS # LRS LRS LRS α−1 LRS ∂V2,t+1 ∂z2,t+1 Eθ u2,ct+1 λ2,t + µ2,t+1 θ2,t+1α 1 − z1,t+1 − µ2,t+1 LRS LRS ∂z2,t+1 ∂z1,t+1 (2.25)

out ∂V1,t+1(˜xt+1) out LRS α−1 where LRS = u1,ct+1 · (1 − ψ)(qz,t+1 + αθ1,t(ψ · z1,0 + (1 − ψ) · z1,t+1) )) ∂z1,t+1 out LRS ∂V2,t+1(˜xt+1) ∂z2,t+1 out LRS α−1 and LRS · LRS = −u2,ct+1 · (1 − ψ)(qz,t+1 + αθ2,t(ψ · z2,0 + (1 − ψ) · (1 − z1,t+1) )). ∂z2,t+1 ∂z1,t+1 First of all, notice the adjusted consumption sharing rule. In any period t, after the idiosyncratic shocks are realized, consumption is determined by Pareto weights inherited from period t − 1 that are additionally updated to take care of potentially binding limited commitment constraints. This is done so that in equilibrium no agent defaults on the contract and so that the long-run co-operation is achievable. Following Marcet and Marimon LRS LRS LRS (2019), the Pareto weights are updated according to λi,t (xt) = λi,t−1(xt−1) + µi,t .

The planner’s decision (2.25) about land allocations is a version of condition (2.20) in the ﬁrst best benchmark, adjusted for expected next period’s binding limited commitment out ∂Vi,t+1(˜xt+1) constraints whenever land rights are incomplete, i.e. ψ < 1. Because LRS > 0 ∀i, ∂zi,t+1

Chapter 2 89 Essays on Macroeconomics and Development the planner needs to shade the land allocated to agent i in period t, if she expects this household to have binding participation constraints in the next period.

Overall, the extent of eﬃcient consumption risk sharing is pinned down in equilibrium by “how slack” are the limited commitment constraints of the agents. In particular, the ability of the planner to provide consumption insurance to both households is an increasing function of the distance between the (inside) value of co-operation and the (outside) value of deviating in the enforcement constraints (2.23) of both households.

While a similar statement holds true about the eﬃciency of land allocations, the degree of land rights chosen ex-ante has important implications for the co-operation between households in equilibrium. To make this clear, we make the following three observations. First, we know from the literature11 that the limited commitment constraints in a given period are usually binding for individuals with currently higher productivity. Second, achieving economic eﬃciency calls for allocating relatively more land to the more productive household (i.e. the one with higher θ). Third, we showed in Proposition 1 that the value of deviation is increasing in ψ for households with current allocation of land LRS satisfying zi,t < zi,0.

Taking these observations together implies that higher degree of land rights ψ will drive down the outside option (and so relax the limited commitment constraints) of the more productive households. Because it is precisely these agents who determine the degree of land and risk sharing co-operation in equilibrium, increases in ψ will improve the eﬃciency of both margins.12 As a direct consequence, the aggregate crop output will also increase, and so will the average consumption.

E Preliminary quantitative results

In what follows, we present preliminary results from the quantitative model. Table E.1 presents the parameter values assumed. Importantly, we chose relatively "standard" values of parameters in order to demonstrate functioning of our model. A more serious calibration is the next priority on our to-do list.

In Figure E.2, we show how key statistics of village productivity, risk sharing and land reallocation change as we increase the degree of land rights ψ in the benchmark allocation with limited commitment, ﬁrst best and outside option. First of all, we observe that for low enough values of ψ the land misallocation is reduced. This can be seen through the

11See e.g. Alvarez and Jermann(2000) 12Notice that even if land rights are complete, i.e. ψ = 1, the land allocation under limited commitment will be distorted (relative to the ﬁrst best benchmark) due to potential adjustments in Pareto weights.

90 Chapter 2 Essays on Macroeconomics and Development

Time preference β 0.85 Utility fn u(c) log(c)

Idiosyncratic prod. shocks (θL, θH = 1 − θL) (0.1, 0.9)

Persistence of idiosyn. prod. shocks ρθ 0.65

Aggregate prod. shocks (φL, φH ) (0.9, 1.1)

Persistence of aggregate prod. shocks ρφ 0.5 Production f-n parameter α 0.7

Figure E.1: Parametrization of quantitative model convergence of the correlation between current idiosyncratic household productivity and the amount of land cultivated towards the ﬁrst best level. As a consequence, the village level agricultural level productivity increases, as reﬂected by increases in the average village consumption.

Importantly, these changes bring about further improvements in risk sharing, as measured by reduction in elasticity of consumption w.r.t. idiosyncratic productivity shocks. These dynamics arise because increases in land formality (for low enough values of ψ) relax limited commitment constraints of the more productive household, which has relatively higher incentives to deviate and so pins down the equilibrium patterns of co-operation.

The above can be seen in E.3 depicting changes in the current value of deviating for low- and high-productivity θ households. In line with out discussion in the theoretical outline of the model, we see that the value of outside option of the high productivity declines and the one of the low productivity agent increases, as we increase the degree of land rights ψ. Crucially, however, notice that the pace at which the latter increases is signiﬁcantly higher than that one of the high productivity households. Taken together, this implies that it is possible that at some point the values of deviation are aﬀected so much that there is no scope for the planner to elicit equilibrium joint co-operation over land and risk sharing. This is precisely what we found in Figure E.2 for values of land rights above approx. 35%.

F Outlook on quantitative analysis

In next steps, we plan to calibrate the quantitative version of the model using the sim- ulated method of moments. Our broad goal is to use this experimental laboratory to study the impact of changes in formality of land markets on properties of (endogenous) informal institutions in rural Ghana. We plan to use the quantitative model to investigate e.g. what is the size of equilibrium feedback forces between land re-allocations and

Chapter 2 91 Essays on Macroeconomics and Development

Figure E.2: Numerical example results

Figure E.3: Changes in value of outside option

92 Chapter 2 Essays on Macroeconomics and Development consumption insurance, how large is the land misallocation due to either low take up of land titles or enforcement frictions.

In a more distant future, we also plan to extend the model in order to conduct a further policy experiments quantifying the eﬀects of land collaterisation (which is essentially non-existent in our data sample) on economic welfare. This margin may be particularly important to analyze as on the one hand it may improve investment, but on the other it may also expand the farmers’ set of self-insurance possibilities, harming the informal insurance networks in our communities of interest.

2.6 Conclusion

In this paper, we address one of the major frictions aﬀecting rural areas of Ghana: pluralistic land tenure and weak enforcement of land rights. Complementing existing literature, we not only study how missing formal land rights form a barrier for eﬃcient allocation of land and agricultural productivity, but also how this friction interacts with the informal mutual insurance networks that are of critical importance in largely subsistence-focused Ghanaian agriculture.

Interestingly and contrary to basic intuition, we find that attempts of governments at increasing formality of land institutions relevant to the living of rural populations need not crowd-out the other informal margins of co-operation, but can even complement them. In particular, we first find that rural communities that hold more of formal titles allowing for land sales face less disputes over land. Second, we find that land changes hands more frequently in these communities with higher land security provided by formal land rights. Third, we find that these reallocations are not only more intense, but also more efficient, as reflected by increases in agricultural productivity and average consumption. Finally, we also find that societies with higher degree of land rights enjoy better consumption smoothing against idiosyncratic income shocks, suggesting improvements in functioning of the mutual insurance networks.

We construct a dynamic model of village co-operation over consumption transfers and land re-allocations that rationalizes our empirical ﬁndings. We endogenize the functioning of these two informal institutions by subjecting them to limited commitment constraints, stipulating that every household has to be made always better-oﬀ within the co-operation that what he could get outside of it by relying on self-insurance and trading land in formal markets. In this environment, adopting higher land rights allows for better anchoring of individuals’ outside value of deviation by specifying ex-ante land ownership. Because the threat of losing one’s land is reduced as the degree of land rights increases, land

Chapter 2 93 Essays on Macroeconomics and Development reallocations are more eﬃcient leading to gains in aggregate output. Because this implies increases in the (inside) value of co-operation, this higher output can be shared more eﬃciently as insurance against idiosyncratic shocks.

Interestingly, our model also uncovers an important non-linearity showing that there may exist an upper bound on the endogenously chosen degree of land formality, crossing which may trigger a complete unraveling of local co-operation. This may happen because informal rural institutions usually span multiple margins of co-operation and, as such, changes in one of those may have ﬁrst order equilibrium eﬀects on all the other ones. Because maintaining a well-functioning nexus of these institutions may be critical in developing parts of the world, traditional societies may rationally choose to remain informal at large, in spite of formal markets being present. As such, we provide a candidate explanation for often observed low take up of formal land titles in developing countries.

94 Chapter 2 Essays on Macroeconomics and Development

References

Adamopoulos, T. and D. Restuccia (2019). Land reform and productivity: A quantitative analysis with micro data. Technical report, National Bureau of Economic Research.

Alvarez, F. and U. J. Jermann (2000). Eﬃciency, equilibrium, and asset pricing with risk of default. Econometrica 68 (4), 775–797.

Attanasio, O. and J.-V. Rıos-Rull (2000). Consumption smoothing in island economies: Can public insurance reduce welfare? European economic review 44 (7), 1225–1258.

Banerjee, A. V., P. J. Gertler, and M. Ghatak (2002). Empowerment and eﬃciency: Tenancy reform in west bengal. Journal of political economy 110 (2), 239–280.

Berry, L. B. (1995). Ghana: A country study, Volume 550. US Government Printing Oﬃce.

Besley, T. (1995). Property rights and investment incentives: Theory and evidence from ghana. Journal of political Economy 103 (5), 903–937.

Brasselle, A.-S., F. Gaspart, and J.-P. Platteau (2002). Land tenure security and investment incentives: puzzling evidence from burkina faso. Journal of Development Economics 67 (2), 373–418.

Carter, M. R., K. D. Wiebe, and B. Blarel (1994). Tenure security for whom? diﬀerential eﬀects of land policy in kenya. Searching for land tenure security in Africa, 141–168.

Chari, A., E. M. Liu, S.-Y. Wang, and Y. Wang (2017). Property rights, land misallocation and agricultural eﬃciency in china. Technical report, National Bureau of Economic Research.

Chen, C. (2017). Untitled land, occupational choice, and agricultural productivity. Amer- ican Economic Journal: Macroeconomics 9 (4), 91–121.

Chen, C., D. Restuccia, and R. Santaeulàlia-Llopis (2017). The eﬀects of land markets on resource allocation and agricultural productivity. Technical report, National Bureau of Economic Research.

Daniels, W. E. (1996). The impact of the 1992 constitution on family rights in ghana. Journal of African Law, 183–193.

De Magalhães, L. and R. Santaeulàlia-Llopis (2018). The consumption, income, and wealth of the poorest: An empirical analysis of economic inequality in rural and urban sub-saharan africa for macroeconomists. Journal of Development Economics 134, 350– 371.

Chapter 2 95 Essays on Macroeconomics and Development

Driberg, J. H. (1934). The african conception of law. Journal of Comparative Legislation and International Law, 230–245.

Fenrich, J., P. Galizzi, and T. E. Higgins (2011). The future of African customary law. Cambridge University Press.

Goldstein, M. and C. Udry (2008). The proﬁts of power: Land rights and agricultural investment in ghana. Journal of political Economy 116 (6), 981–1022.

Gollin, D., D. Lagakos, and M. E. Waugh (2014). The agricultural productivity gap. The Quarterly Journal of Economics 129 (2), 939–993.

Gollin, D. and C. Udry (2019). Heterogeneity, measurement error and misallocation: Evidence from african agriculture. NBER Working Paper (w25440).

Kline, A., É. Moore, E. Ramey, K. Hernandez, L. Ehrhardt, M. Reed, M. Parker, S. Hen- son, T. Winn, and T. Wood (2019). Whose land is it anyway? navigating ghana’s complex land system. Texas A&M Law Review 6 (1), 1–22.

Ligon, E., J. P. Thomas, and T. Worrall (2002). Informal insurance arrangements with limited commitment: Theory and evidence from village economies. The Review of Economic Studies 69 (1), 209–244.

Manalis, G. (2019). Land rights and risk sharing in rural west africa.

Marcet, A. and R. Marimon (2019). Recursive contracts. Econometrica 87 (5), 1589–1631.

Mazur, K. (2020). Sharing risk to avoid tragedy: Informal insurance and irrigation in village economies. Available at SSRN 3326582 .

Meghir, C., A. M. Mobarak, C. D. Mommaerts, and M. Morten (2019). Migration and informal insurance. Technical report, National Bureau of Economic Research.

Michalopoulos, S. and E. Papaioannou (2020). Historical legacies and african development. Journal of Economic Literature 58 (1), 53–128.

Morten, M. (2019). Temporary migration and endogenous risk sharing in village india. Journal of Political Economy 127 (1), 1–46.

Ollennu, N. A. et al. (1962). Principles of customary land law in ghana. Principles of customary land law in Ghana..

Restuccia, D. and R. Santaeulalia-Llopis (2017). Land misallocation and productivity. Technical report, National Bureau of Economic Research.

96 Chapter 2 Essays on Macroeconomics and Development

Townsend, R. M. (1994). Risk and insurance in village india. Econometrica: Journal of the Econometric Society, 539–591.

Turpin, C. C. (1963). Principles of customary land law in ghana. by the hon. mr. justice n. a. ollennu, judge of the high court of ghana, honorary professor of law in the university of ghana. number 2 in the law in africa series. [london: Sweet amp; maxwell, ltd.1962. xxvi, 164 and (appendix and index) 108 pp. £ 10s. net.]. The Cambridge Law Journal 21 (2), 321–323.

Udry, C. (1994). Risk and insurance in a rural credit market: An empirical investigation in northern nigeria. The Review of Economic Studies 61 (3), 495–526.

Ward, W. E. F. (1967). A history of Ghana. Allen & Unwin.

Woodman, G. R. (2003). Family law in ghana under the constitution 1992. Int’l Surv. Fam. L., 195.

Chapter 2 97 Essays on Macroeconomics and Development

2.7 Appendix

A Regression analysis

Table 2.1: Land Disputes (multiple claims)

Dependent variable:

share of village land w/ multiple claimsv,t

(1) (2)

∗∗ ∗ sell-rightsv,t -0.074 -0.074 (0.034) (0.043)

Land Controls Yes Yes Village Fixed Eﬀects Yes Yes District Clustered se No Yes Observations 341 341

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Total sample of rural villages is 396. 55 rural villages exhibit NA values in our main explanatory variable, so the sample is 341 observations.

Table 2.2: Land ﬂuidity index

Dependent variable:

Land Fluidity Index

∗ sell-rightsv,t/t+1 70.808 (40.951)

Constant -8.350 (20.385)

Village Land controls Yes District Fixed Eﬀects Yes Observations 197 R2 0.771 Residual Std. Error 31.353 (df = 89) F Statistic 2.804∗∗∗ (df = 107; 89)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Total sample of rural villages is 198 per wave. One rural village exhibits NA values in our main explanatory variable for both waves, so the sample is 197 observations.

98 Chapter 2 Essays on Macroeconomics and Development

Table 2.3: Village productivity

Dependent variable:

log(village-productivityv,t)

∗∗ sell-rightsv,t 2.345 (0.922)

Village Land controls Yes District Clustered se Yes Observations 340 R2 0.325 F Statistic 8.175∗∗∗ (df = 8; 136)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Total sample of rural villages for both waves that have non-NA values in selling rights is 341. Here we lose one village due to our dependent variable

B Endogeneity of selling rights

Importantly, notice that our proxy for selling rights can be endogenous to many different factors. First of all, it may respond to informal risk-sharing arrangements at the community level, generating a reverse causality problem. To address this very issue, we build the quantitative model of risk sharing and land re-allocations that explicitly contains this channel and so allows us for isolating these effects (see more below). However, the degree of land formality may be also differentially affected by the relative strength of formal and informal institutions present in each community, or the geographical distance of these communities to institutions allowing for formalization of land rights. To address this issue, we devise an IV strategy building on the fact that regions of Ghana have been colonized by the Great Britain for centuries, up to the country’s independence in 1957. The British rule was mostly concentrated in coastal regions due to their main interest in trade. Therefore, we plan to instrument land rights with proximity of the villages in the sample to major cities in the coastal south, proxying distance to modern British institutions. In this way, we obtain exogenous variation in our explanatory variable.

IV strategy

Our IV strategy relies on features of the British rule of the Gold Coast during the 19th century. As colonizers of Ghana were interested in trading opportunities that the country’s mineral resources oﬀered, they were focused on the coastal area were most of the commercial activity was taking place. One additional reason why also the British activity was concentrated in the Ghanaian South was the conﬂictual relations with the Ashanti,

Chapter 2 99 Essays on Macroeconomics and Development

Table 2.4: Village mean consumption per capita

Dependent variable:

village mean consumption per capita

(1) (2)

∗ ∗ sell-rightsv,t 0.281 0.281 (0.150) (0.170)

Constant 7.085∗∗∗ 7.085∗∗∗ (0.321) (0.319)

Village + Land controls Yes Yes District x Wave FE Yes Yes District Clustered se No Yes Observations 341 341 R2 0.904 0.904 Adjusted R2 0.795 0.795 Residual Std. Error 0.265 (df = 160) 0.265 (df = 160) F Statistic 8.347∗∗∗ (df = 180; 160) 8.347∗∗∗ (df = 180; 160)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

100 Chapter 2 Essays on Macroeconomics and Development

Table 2.5: Risk-sharing ratio

Dependent variable:

pc pc2 log(RSv,t) log(RSv,t )

(1) (2)

∗∗ ∗∗ sell-rightsv,t -1.720 -1.611 (0.855) (0.807)

log(mean village consumption pc) 0.370 (0.296)

log(mean village consumption pc2) 0.528∗ (0.298)

Village+Land Controls Yes Yes District Clustered se Yes Yes Village FE Yes Yes Observations 341 341 R2 0.125 0.136 Adjusted R2 -1.155 -1.129 F Statistic (df = 7; 138) 2.821∗∗∗ 3.103∗∗∗

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Columns (1), (2) are in levels, columns (3), (4) are per capita with equal weights on household members, columns (5), (6) are per capita with different weights on household members according to gender and age. Income is gross income. Selling land rights are calculated as the weighted mean of land at the village level that can be sold - weights are based on plot size. The sample is restricted to rural areas. .Consumption and Income has been winsorized at 2.5% level one of the largest ethnic groups of inland Ghana. At the same time, coastal states (Fante, Ewe, Ga) were also in conflict with the Ashanti over control of their lands. Because of this, leaders of these states also resorted to the British protection against their inland enemies.13 Several treaties were signed that sealed the cooperation between the coastal South and the British occupants. The most notable of all was the Bond of 1844, a special treaty with local chiefs of Fante and others from the coastal areas which laid the legal foundations of the British rule in Ghana.14 Even though in principle, the treaty granted limited judicial powers to the British, underlying efforts to extend their judicial authority were so successful that in 1850 they considered establishing European courts in the place of African ones (Berry, 1995). However, the influence of the British judicial power was confined to the coast (at least up until the end of the nineteenth century, as the Ashanti and Northern regions were annexed only in 1896 and 1902, respectively - see Figure C.9a). As a result, the penetration of the British rule across Ghana exhibits substantial heterogeneity. It was predominant in the coastal area, limited in the Ashanti region (central Ghana) and negligible in the Northern part (Michalopoulos and Papaioannou, 2020). This heterogeneity is also mirrored in the timing of annexation for Ghana’s different regions

13..under Maclean’s administration, several coastal tribes had submitted voluntarily to British protection Berry(1995) 14This document obliged local leaders to submit serious crimes such as murder and robbery to British jurisdiction Berry(1995)

Chapter 2 101 Essays on Macroeconomics and Development

Table 2.6: Townsend Panel Regression Results

Dependent variable:

ihs(consumption per capita) ihs(consumption per capita) per capita (1) per capita (2)

(1) (2) (3) (4)

∗∗∗ ihs(yh,v,t) (gross pc income) 0.035 (0.007)

∗∗∗ ihs(yh,v,t) (gross pc income exc. ﬁn. wealth) 0.026 (0.006)

∗∗∗ ihs(yh,v,t) (gross pc2 income) 0.028 (0.007)

∗∗∗ ihs(yh,v,t) (gross pc2 income exc. ﬁn. wealth) 0.021 (0.005)

∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ sell-rightsv,t 0.377 0.328 0.372 0.322 (0.143) (0.118) (0.138) (0.114)

∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ihs(cv,t) - village consumption 0.937 0.969 0.925 0.953 (0.031) (0.034) (0.028) (0.030)

∗∗ ihs(yh,v,t) (gross pc income)·sell-rightsv,t -0.063 (0.029)

∗∗ ihs(yh,v,t) (gross pc income exc. ﬁn. wealth)·sell-rightsv,t -0.055 (0.026)

∗∗ ihs(yh,v,t) (gross pc2 income)·sell-rightsv,t -0.061 (0.027)

∗∗ ihs(yh,v,t) (gross pc2 income exc. ﬁn. wealth)·sell-rightsv,t -0.052 (0.025)

Village + Land controls Yes Yes Yes Yes HH Fixed eﬀects Yes Yes Yes Yes Village Clustered se Yes Yes Yes Yes Observations 4,723 4,512 4,723 4,512 R2 0.588 0.589 0.597 0.598 Adjusted R2 -0.044 -0.074 -0.022 -0.051 F Statistic 295.531∗∗∗ (df = 9; 1863) 275.028∗∗∗ (df = 9; 1724) 306.420∗∗∗ (df = 9; 1863) 285.394∗∗∗ (df = 9; 1724)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Column (1) and (2) correspond to equal weights on household members. Col- umn (3) and (4) correspond to different weights on household members according to gender and age. Gross income includes in-transfers and credit. Gross income excluding financial wealth ignores financial assets. Consumption and income have been winsorized at the 2.5% level and have been transformed using the inverse hyperbolic sine transformation. Selling land rights are calculated as the weighted mean of land at the village level that can be sold - weights are based on plot size. The sample is restricted to rural areas. 4,723 households report non negative values of gross income and 4,512 non-negative values of gross income excluding financial wealth. The difference between the two captures households reporting only financial wealth.

Table 2.7: Savings

Dependent variable:

savingsh,v,t lsdv panel

(1) (2)

sell-rightsh,v,t -77.585 -37.508 (152.442) (472.114)

Constant -193.106 (179.784)

Land Controls Yes Yes HH Fixed Eﬀects No Yes Village×Time Fixed eﬀects Yes No Village Clustered se Yes Yes Observations 2,782 2,782 R2 0.170 0.025 Adjusted R2 -0.027 -3.090 Residual Std. Error 2,190.035 (df = 2246) F Statistic 0.862 (df = 535; 2246) 1.060 (df = 16; 663)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

102 Chapter 2 Essays on Macroeconomics and Development

Table 2.8: Land Purchases

Dependent variable:

share of purchased landv,t (1) (2)

sell-rightsv,t 0.031 0.031 (0.032) (0.029)

Land Controls Yes Yes Village Fixed Eﬀects Yes Yes District Clustered se No Yes Observations 341 341

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

(see Figure C.9a).

Our IV approach exploits this form of heterogeneity to instrument for land rights according to the following assumption. Areas directly under the British control would be more likely to have adopted a land tenure regime closer to the western individualistic norm rather the customary collective one. This approach requires us to quantify the degree of influence of the British rule in different regions of Ghana and in particular in the villages appearing in our sample. We plan to do so by measuring the effective distance from the coastal centre of the British rule (taking into account rail lines or roads) and the distance from other commercial centres of the country, such as Kumasi (which did not belong to the south but constituted a strong commercial pole of the era). The minimum distance out of those two measures could accurately capture the proximity of the enumeration areas in the sample to the British rule.

Chapter 2 103 Essays on Macroeconomics and Development

C Suggestive empirical observations

Figure C.1: From whom and how did you obtain the plot?

Figure C.2: Diﬀerence in hh land size

104 Chapter 2 Essays on Macroeconomics and Development

Figure C.3: Diﬀerence in villages’ selling rights

Figure C.4: Savings as % of consumption

Chapter 2 105 Essays on Macroeconomics and Development

Figure C.5: Source of the loan

Figure C.6: Purpose of the loan

106 Chapter 2 Essays on Macroeconomics and Development

Figure C.7: Collateral request

Figure C.8: British annexation of coastal, central, northern and east regions

(a) Source: Berry(1995) with information from Ward(1967)

Chapter 2 107 Essays on Macroeconomics and Development

Figure C.10: Diﬀerence in villages’ selling rights

108 Chapter 2 Chapter 3

Contagion as a dealmaker? The eﬀect of ﬁnancial spillovers on regional lending programs

joint with Alica Ida Bonk (Swiss National Bank) and Alexandra Fotiou (International Monetary Fund) 1

1The views expressed here are those of our own and do not represent those of the International Monetary Fund and the Swiss National Bank

109 Essays on Macroeconomics and Development

3.1 Introduction

The cooperation between the IMF and the European lending mechanisms during the European debt crisis was generally deemed successful. However, "different views over program strategy emerged because of differential emphasis on potential spillovers within the euro area" (IMF, 2017). The IMF’s main priority is to ensure "the stability of the international monetary and financial system", while the European Commission (EC) and the European Central Bank (ECB) - representing all members of the euro area - put heightened emphasis on reducing the risk of contagion within the currency union (Darvas, 2017) (table 3.1). At the peak of the sovereign debt crisis, starting in 2011, the ESM

IMF EC-ECB Help Latvia in a way Latvia Restore stability & that sets good precedent (2008) promote growth for others & helps the stability of neighbours Ensure stability in the Greece Restore stability & Euro-Area (fear of contagion), (2010/11) promote growth address the Greek debt later Ensure stability in the Ireland Restore stability & Euro-Area (2010/11) promote growth (fear of contagion)

Table 3.1: Priorities of the IMF and the European institutions involved in lending programs (source: Darvas(2017)) provided loans with a longer maturity and at a lower interest rate than the IMF (Corsetti et al., 2017) (table 3.2).23 A potential explanation is that the threat of contagion perceived by the average euro area member, which is internalized by the ESM, is larger than the one of the average IMF member country due to closer financial and trade linkages within the former group.4 In addition, the IMF has the capacity and ability to pool 2The European Financial Stability Facility (EFSF) was created as a temporary crisis resolution mechanism by the euro area Member States in June 2010. The European Stability Mechanism (ESM) was set up in 2012 as a successor of the EFSF of permanent nature. 3The evolution of the different loan conditions provided from the European mechanisms in comparison to the IMF is insightful and reflects the development in the design of the European loan-contracts. Initially, the EFSF shadowed the expertise and contract-design of the IMF. However, after 2011 we observe a change in the conditions offered by the European mechanisms. 4For a graphical illustration of financial linkages in the euro area see Appendix A. Germany held Spanish banking debt worth 49% of German GDP and France held Greek banking debt worth 28% of

110 Chapter 3 Essays on Macroeconomics and Development

Oﬃcial Loan Terms EFSF/ESM IMF Country Interest Interest Date Maturity Maturity (in programme) Rate Rate May 2010 5 yrs 4.041 5 yrs 3.23 June 2011 10 yrs 3.78 5 yrs 3.53 Greece March 2012 20 yrs 2.07 8 yrs 3.13 December 2012 30 yrs 0.93 8 yrs 3.07 December 2010 7.5 yrs 5.25 7 yrs 3.37 Ireland July 2011 15 yrs 2.74 7 yrs 3.53 June 2013 22 yrs 2.32 7 yrs 3.07 May 2011 7.5 yrs 5.47 7 yrs 3.37 Portugal July 2011 15 yrs 3.15 7 yrs 3.53 June 2013 22 yrs 2.19 7 yrs 3.07 Spain November 2012 12.5 yrs 0.78 -- Cyprus May 2013 15 yrs 1.03 4 yrs 3.07

Table 3.2: Official Loan Terms. source: Corsetti et al. (2017) based on IMF, EC, EFSF, ESM and Bloomberg risks together among different countries globally. This reinforces the bargaining power of troubled countries within the euro area with respect to the regional mechanisms. As a result, ESM lending conditions might be more benign compared to those offered by the IMF and the degree of leniency granted by the ESM is increasing in the strength of spillover risk.

A motivating example is the case of Ireland and Cyprus. Both countries faced a banking crisis in 2008 and 2012/13, respectively. Consequently, they entered into a lending agreement with the IMF and the European lending mechanisms. Despite the fact that both countries were aﬀected by the same type of crisis, the remedies applied diﬀered. In both cases, the IMF proposed the bail-in of the banking sector. However, European mechanisms strongly opposed a bail-in of Irish banks’ creditors, while in the case of Cyprus, they agreed.

In his review of the Irish crisis, Eichengreen mentions a potential reasoning behind the European mechanisms’ diﬀering approaches:

"[...] some commentators [...] suggest that Trichet insisted (telephonically) that the Irish authorities not bail in bank bondholders on the grounds that

French GDP, when the respective ﬁrst ﬁnancial assistance program was agreed upon by the European lending mechanisms.

Chapter 3 111 Essays on Macroeconomics and Development

doing so would damage the big French and German banks holding Irish bank paper" (Eichengreen, 2015)

In contrast, shareholders of the Cyprus’ banking sector were primarily located in Russia. Hence, spillover eﬀects onto other euro area countries resulting from a Cypriot bail-in were considered low compared to the Irish case. Indeed, ﬁgures A.1 and A.2 in the Appendix illustrate that Euro-core countries such as Germany, Belgium, the Netherlands and France held substantially larger debt-amounts of Irish banks than of Cypriot banks when the respective ESM packages were approved.

We construct a new dataset that records all the dates that official meetings (e.g., Eu- rogroup meetings) between the different institutions and lending mechanisms (i.e., EC, ESM, IMF etc.) took place or relevant statements were made by officials for a crisis-hit country (Greece, Ireland, Portugal, Spain and Cyprus).5 We distinguish between announcements made by the IMF and the European mechanisms and assess their impact on a measure of interconnectedness.6 Lending related announcements – irrespective of being related to future loans or current period disbursements – by the European mechanisms reduced the spillover effect that Ireland, Greece and Portugal had on the rest of the union, but had the opposite effect for Cyprus. In contrast, lending related announcements by the IMF did not significantly affect spillovers of debtor countries. We embrace the view that the European lending mechanisms successfully accomplished their goal of reducing the risk of contagion within the euro-area, which we use as motivating evidence to turn our focus to the design of lending programs by regional mechanisms.

The aim of this paper is to assess how the presence of contagion within a region or currency union, in our context, may impact the design of lending programs oﬀered by regional mechanisms.7 In other words, we explore whether (and when) the risk of spillovers gives

5In the newly built dataset we also observe all the loan-disbursements with the corresponding maturity and interest rate. Therefore, we have full information on the conditionality of the loan, which is later used to validate the theoretical predictions of our model. Moreover, we are able to track the timing of the announcements regarding a loan disbursement and distinguish between the anticipated date of a disbursement, the realized date, and in some cases the revised date. The timing can help to study promises of loan amounts, interest rates and maturity dates and whether these promises were materialized or revised/ updated. 6Our measure of interconnectedness is based on an estimated time-varying correlation, derived from a bivariate-GARCH model. 7Our narrative examples and empirical evidence draw from the experience of the European Union. However, our underlying hypothesis and theoretical model has a broader application to all Regional Fi- nancing Arrangements (RFA)(RFA members are the European Stability Mechanism, the Arab Monetary Fund, the BRICS Contingent Reserve Arrangement, the Chiang-Mai Initiative Multilateralisation, the Eurasian Fund for Stabilization and Development, the EU Balance of Payments Facility, the European

112 Chapter 3 Essays on Macroeconomics and Development countries in trouble bargaining power that results in more favorable borrowing conditions. To answer this question, within a simple recursive contract framework, we present a dynamic lending contract between a lending mechanism and a borrowing country, in which we introduce spillover costs. We structure the lending mechanism and the borrowing country in the spirit of a principal-agent model. The model frames the financial assistance contract as a bargaining game, where the principal (lending mechanism) offers a contract to the agent (borrowing country), which consists of current and future transfers. We adopt the view that financial assistance programmes worked as a carrot-and-stick.

The contract requires voluntary participation from the side of the borrower, who always has the option to leave and resort to autarky. Therefore, for the contract to be self- enforcing, the loan terms have to be designed in a way that ensure voluntary participation of the borrower. We interpret a binding participation constraint as a credible threat of default, since if the borrower resorts to autarky, he reneges its obligations to the principal. The key feature of the theoretical part comes from the fact that in all states in which the borrower credibly threatens to default, the principal incurs a cost. This cost arises from the interconnectedness between the lending and borrowing economy. It captures spillover costs which emerge when tightly linked economies participating in a lending agreement, face the contingency of one-sided default. The principal, when designing the contract, internalizes the spillover costs weighted by the corresponding probability of a binding participation constraint. This allows to account for the fact that the European mechanisms took contagion into serious consideration in the design of the lending programs.

To better understand the theoretical predictions of our model, we compare it to a benchmark in which spillover costs are absent. The latter case corresponds to a simple one- sided limited commitment dynamic contract. We find that a lending program featuring spillovers between the contracting parties will put more weight on future payments compared to the benchmark model. The reason is that the principal promises higher future transfers to lower spillover costs. Therefore, the principal finds it more profitable to equate the binding participation constraint and thus ensure voluntary participation of the borrower through substituting current consumption transfers with future ones. As a result, in the presence of spillover costs, the contract back-loads consumption at a higher degree than the benchmark case.

We test and validate the main prediction of our theoretical model empirically. More precisely, we show empirical evidence that higher interconnectedness between the country

Financial Stabilisation Mechanism, the Latin American Reserve Fund (FLAR), and the North American Financial Agreement).

Chapter 3 113 Essays on Macroeconomics and Development in trouble and the rest of the euro-area partners has a negative effect on the country’s current consumption. In particular, the borrowing country appears to be worse off in the short term, but better off in the long term under a lending contract which involves high spillovers between the two parties. We proxy current consumption by a measure of current cumulative loan disbursements.

Recent relevant work in the literature (Gourinchas et al., 2018) discusses the size of cross- country transfers in the euro-area during the sovereign debt crisis, showing that transfers were large and heterogeneous among the different countries. In their two-period model, lenders transfer resources to debtors to avoid default. In our paper, we focus on contagion as a channel for the willingness of the creditors to provide financial assistance. Our theoretical part represents financial assistance programmes between a lending mechanism and a sovereign borrower as a constrained efficient allocation. An extensive strand of literature adopts the same representation, advocating its advantage to explicitly model the underlying frictions. In particular, Dovis(2019) juxtaposes the aforementioned approach to the quantitative incomplete markets approach on sovereign debt, as initiated by the seminal paper by Eaton and Gersovitz(1981), in order to evince its suitability for policy analysis. Atkeson(1991) is one of the first works representing sovereign borrowing as an constrained optimal contract, followed by Thomas and Worrall(1994), Alvarez and Jermann(2000) and Kehoe and Perri(2002) who either study limited commitment or moral hazard as main frictions. This work also draws on more recent literature such as Abrahám et al.(2018) and Muller et al.(2019). In particular, we borrow form the latter work the interpretation of the arising limited commitment friction in a lending programme as a credible threat of default by the borrower.

We contribute to this strand of the literature by explicitly modelling spillover costs. These costs are directly linked to the underlying limited enforcement friction, characterising the contract. In our model, the lender incurs certain costs, whenever the borrower credibly threatens to leave the contract. This additional element signiﬁcantly changes the structure of the loan oﬀered by the lender and the weight she puts on the available means to sustain the contract.

In addition, part of the contribution of this paper relies on the newly built dataset. This is the first dataset to record all the dates that announcements and statements of EU and IMF officials regarding programme-countries were made. Moreover, we extend and enrich the dataset of Corsetti et al.(2017) on loan conditions (i.e., disbursements, maturities, interest rates) beyond 2011 and in higher frequency. We distinguish between the timing of announcements on the different conditionality measures and record information on announcements related to a payment of a future disburesement, the realization of a

114 Chapter 3 Essays on Macroeconomics and Development disbursement, or its revision.

There is enough empirical evidence in the literature of existence of contagion, with a broad range of definitions for contagion (see e.g. Forbes(2012)). Beirne and Fratzscher (2013) study contagion in the euro area focusing on sovereign credit risk, and show that between 2000 and 2011 sovereign spreads and Credit Default Swaps were on average 31bp and 44bp respectively higher in the euro area due to cross-country spillovers. Constancio et al.(2012) also shows evidence of contagion during the euro-area sovereign-debt crisis. He focuses on comovements in sovereign bond yields filtering out long-run trend and attributing the remaining correlation to cross-country contagion. His findings indicate that in 2011, 38 percent of the variance of Italian and Spanish government yields was explained by contagion from Greece, Ireland, and Portugal. We add to this existing evidence and follow a widely-used approach in finance (Diebold and Yılmaz, 2009), which is based in VAR variance decompositions, to compute an aggregate spillover index in the subsamples before and after the peak of the crisis. However, in order to explore better the time-dimension and the frequency of our data, we estimate dynamic conditional correlations (Engle, 2002), which we us a measure to proxy contagion between countries.

The remaining of the paper is organized as follows. In Section 2, we describe the newly built dataset. In Section 3, we provide empirical motivation regarding our theoretical modelling assumptions. Section 4 presents the model and its predictions, which we validate empirically in Section 5. Section 6 concludes.

3.2 A new dataset on loan announcements and disbursements during the Euro crisis

In order to determine the eﬀect of announcements and disbursements by the European lending mechanisms and the IMF on spillovers between Euro area countries we create a novel database. We use press releases, memos and other oﬃcial statements of the European Commission (EC), the ESM and the IMF. In ?? we present some examples of these statements. For each day, we record the following information:8

1. whether a statement relating to a lending program has been issued by the EC/ESM

8EC statements contain information on joint (planned) actions by the Troika (EC, ECB and IMF) or explicitly mention ﬁnancial assistance given by the EFSF, the EFSM or the ESM. While the content of statements issued by the EC tends to overlap with those from the ESM, the latter provides additional information in some cases. Press releases and oﬃcial statements issued by the aforementioned institutions can be found here: https://www.esm.europa.eu/newsroom/press-releases, https://ec.europa.eu/ commission/presscorner/home/en, https://www.imf.org/en/News/

Chapter 3 115 Essays on Macroeconomics and Development

or the IMF (includes mentions of current or future loan disbursements as well as more general announcements related to e.g. the initialization of a lending program, completed review missions or maturity extensions),

2. whether an EC/ESM statement mentions a concrete future or current-period loan disbursement,

3. whether a realized EC/ESM disbursement was anticipated,

4. at which time horizon a future EC/ESM disbursement is anticipated to take place (within the current month or the next 1, 2, 3, 6, 12, 24 or 36 months), and

5. the announced or realized loan size, interest rate and maturity of the loan.

We record data for all euro area crisis countries (Ireland, Greece, Spain, Portugal and Cyprus) from April 29, 2009 until September 20, 2018.9 Information mentioned in points 1.-3. above is recorded in form of dummy variables which equal one if such an information was provided on a given day and zero otherwise.

To the best of our knowledge we are the first to collect such information. Our approach differs from Corsetti et al.(2017) who collect information on changes in maturities and interest rates offered to debtor countries which occurred in 2011 and 2013. In contrast, our database contains higher frequency (daily) data on lending related announcements. Furthermore, while Corsetti et al.(2017) are interested in the effect of loan conditions on sovereigns’ bond market access, our aim is to identify the causal effect of financial assistance on spillovers across euro area countries. As will be highlighted below, any disbursement or change in loan conditions through the Troika, had been previously signalled which eliminates the surprise effect of policy actions. To remedy this issue, we focus on announcements by the EC and the IMF rather than policy changes.

From the descriptive statistics in 3.4, three facts are striking. First, Greece and Ireland were the most common subject of statements and press releases by the EC and the ESM followed by Spain, Portugal and finally Ireland. Second, the difference between "all" announcements made by the European creditors and the number of concrete references to future or current disbursements is large especially for Spain, reflecting the Spanish government’s hesitation to ask for financial support. Third, for Greece, future disbursements were announced three times more often than actual disbursements whereas the ratio was below 1 for Ireland, potentially reflecting the lengthy bargaining process with Greece over austerity measures.

9When a statement or press release is issued during the weekend, we record it on the following Monday (when markets open).

116 Chapter 3 Essays on Macroeconomics and Development

To visualize the dynamics of events, C.1 plots the annualized count measure of announcements for each country. The height of the bars reflects the timing of the crises and subsequent lending programs.10 For example, for Greece announcements by the IMF and the EC jumped up in 2010 and stayed elevated through 2015. Indeed, in April 2010, the Greek authorities requested an initial loan of €45 billion and in May 2010 they announced a number of austerity measures to secure a three-year €110 billion loan. A second bailout loan of €130 billion was approved in February 2012. In 2013, the Greek parliament approved new austerity measures paving the way for a new tranche worth almost 7 billion euros. Finally, in 2015, the Troika approved the third bailout package worth 86 billion euro. As reflected in C.1, the Irish debt crisis started simultaneously but ended sooner than in Greece (announcements by the IMF and the EC after 2014 are mostly concerned with post-program surveillance missions to Ireland). Portugal received a €78 billion bailout package but already regained complete access to financial markets in September 2013. Spain became a major concern for the Euro-zone in June 2012 as indicated by the spike in EC announcements.11 The large number of EC statements on Spain in 2012 potentially reflects the size of the Spanish economy and hence the threat it posed to the rest of the euro area. On 9 June 2012, the Eurogroup decided to grant Spain a financial support package with a maximum value of 100 billion. Although Cyprus already requested financial assistance in 2012, the package worth 10 billion was only approved in March/April 2013 which is mirrored in the chart.

Overall, C.1 shows that IMF and EC statements tended to move in parallel but initial IMF statements occurred earlier. Furthermore, actual disbursement had always been pre- announced (not visible in our annual chart) resulting in a likely anticipation eﬀect that our empirical strategy in sectionC will take into consideration.

3.3 Empirical analysis

In this section we assess the risk of contagion empirically. The sovereign debt crisis highlighted the relevance of transmission of shocks and contagion in the euro area due to strong financial and trade interlinkages. At the same time, several Member States had to take up financial support from the IMF and the Regional financial arrangements (RFA) (e.g. Ireland, Cyprus, Greece). For this reason, the EA appears to be the ideal candidate to illustrate our analysis.

As a ﬁrst step in our empirical analysis, we want to check whether there is evidence

10For a detailed description of events during the European debt crisis see Corsetti et al.(2017). 11The interest rate on Spain’s 10-year bonds reached 7% and the country had diﬃculty accessing bond markets

Chapter 3 117 Essays on Macroeconomics and Development of contagion across the financial markets of the different EU countries. To understand how interconnected the different markets are, we first construct an overall measure of contagion – a Spillover Index – based on a Vector Autoregressive (VAR) model (Diebold and Yılmaz, 2009). Then, in order to understand how contagion may have changed over time, we examine a time-varying measure of correlations that we estimate using a Bivariate-GARCH Dynamic Conditional Correlation Model (DCC) (Engle, 2002).

To study the different spillover channels we use different financial and macroeconomic data. More precisely, for the sovereign channel, we obtain daily and monthly data from Datastream on sovereign CDS of 5-year seniority and 10-year government bond total returns denominated in euro. For the banking channel, we make use of the 5-year senior bank CDS daily data of the biggest banks of each country from Datastream denominated in euro, equity total return indices from MSCI Bloomberg, and consolidated banking data from the BIS. Regarding the bank CDS, we create a weighted aggregate bank CDS index for each country, where the weights depend on the size of assets holdings of each bank.12 Moreover, we control for common currency financial factors, such as the Eurostoxx 50 index, the 3-month Euribor and EONIA rate. Our final sample covers the period between 2007 to 2018.

A The Spillover Index - Vector Autoregressive Model

Following Diebold and Yılmaz(2009), within a VAR we examine variance decompositions, which allow us to aggregate spillover eﬀects across countries and derive a single measure of contagion. The idea is that for each ﬁnancial asset i (i.e. the government bond of country i), we focus on the shares of its forecast error variance coming from shocks on asset j (i.e. the government bond of country j). Then we sum up each county’s shares which gives us a measure of interconnectedness.

More precisely, if we take a simple covariance stationary VAR

yt = Ayt−1 + ut, where ut ∼ wn(0, Σu), we can derive the moving average (MA) representation, since our model is covariance stationary. The MA representation reads as

−1 yt = (I − AL) ut = B(L)ut, where B(L) = (I − AL)−1. For the derivation of the variance decompositions, it is easier to write the MA representation in an equivalent form

yt = Γ(L)εt, 12Data on the size of asset holdings is obtained from the S&P Global Market Intelligence platform

118 Chapter 3 Essays on Macroeconomics and Development

−1 0 −1 where Γ(L) = B(L)Qt and εt = Qtut, E(εtεt) = I and Qt is the unique lower-triangular

Cholesky factor of the covariance matrix of ut. The one-step-ahead forecasting at time t is yt+1,t = Ayt with the corresponding one-step-ahead error vector being

et+1,t = yt+1 − yt+1,t = Γ0εt+1

0 0 with covariance matrix E(et+1,tet+1,t) = Γ0Γ0. Using the H-step-ahead forecasts, for a pth order N-variable VAR, the Spillover index is

2 ΣΣγh,ij S = 0 Σtrace(ΓhΓh)

In our analysis, we estimate a model of two lags for the endogenous variables (y). The vector of endogenous variables consists of ﬁrst log-diﬀerences of daily CDS for Bel- gium, Cyprus, France, Germany, Greece, Ireland, Italy, Portugal, and Spain, i.e. yt =

(yBEL,t, yCY P R,t, yF RA,t, yDEU,t, yGRE,t, yIRL,t, yIT A,t, yP RT,t, yESP,t). We analyze daily data on sovereign CDS (2008-2018) (bank CDS etc.)

Our results show evidence of contagion between EA countries, but since contagion is not one-to-one (i.e. the spillover measures are not 100 percent), this is an indication that there is room for risk-sharing. We ﬁnd evidence of heterogeneous interlinkages among the Member States. Peripheral countries, such as Italy, Spain, Portugal and Ireland are a great source of spillovers among themselves and appear to aﬀect also a number of core countries. Instead, Cyprus does not contribute to the variability of the sovereign and banking markets of the rest of the countries. This result is in line with our motivating narrative evidence.

B Bivariate-GARCH Dynamic Conditional Correlation Model

Exploiting the high frequency of the data, we study time-varying correlations and the extent to which official announcements for programme-countries affect the decisions of other countries’ investors. For this purpose we use the Bivariate-GARCH Dynamic Conditional Correlation (DCC) model of Engle (2002). Correlations are a critical input for financial management and DCC is a flexible, parametric model widely used for forecasting, and estimation of time-varying correlations. A Fisher transformation is applied in the resulting correlation to ensure that the correlation are normalized and always lie between -1 and 1, which is the measure that we use for the analysis that follows.

In Figure ... we plot the estimated DCC together with dates of the oﬃcial statements and announcements for the programme-countries (Greece, Ireland, Portugal, Cyprus). In addition, we include the realized disbursements of the loans that these countries received.

Chapter 3 119 Essays on Macroeconomics and Development

An interesting observation coming from the different subplots and the DCC is the existence of two different "contagion-regimes". In the beginning of the sample, in 2009, Greece, Ireland and Portugal were interconnected with the different euro-area markets at a correlation level of 0.6-0.7. Instead during 2012, we observe a regime-switch, investors decide to move away from these markets and diversify their portfolios, moving into a "low-contagion" regime.

Since we are interested in the initial period of the euro-area sovereign debt crisis, we analyze the sub-sample of the "high-contagion" regime. We re-run the VAR model and assess the variance decompositions to derive the spillover index for the period of 2008- 2011. As expected, the euro-area markets were highly interconnected in the ﬁrst part of the eurozone crisis, at a level of 76.63 percent. The spillover for the subsample is higher compared to 54.09 percent of the entire sample, but there is still evidence that there is room for risk-sharing. Looking also at the contributions of each country, there is spillover-heterogeneity across the diﬀerent countries with Spain contribution the most to the variation of most of the countries, in particular to Italy, Portugal, and Ireland The contribution of Ireland and Greece to the neighbor countries of the Union, is also pretty high. Instead, Cyprus appears to have the smallest contribution level in the sovereign CDS of the other countries, with it being very low.

C Spillovers and lending during the Euro crisis - A linear regression analysis

The eﬀect of lending programs on spillovers

To verify whether providing financial support to crisis-hit countries mitigated their spillover effects on other members of the Eurozone we rely on our newly constructed database on loan announcement and disbursements. For each country we regress the DCC on a dummy that equals one on days of announcements and zero otherwise. As reported in 3.11, announcements by the European lending mechanisms are negatively related to spillovers for all countries except Cyprus. Hence, the realized effect on spillovers is in line with the European commissions’ objective. In contrast lending related announcements by the IMF have no significant effect on spillovers for any of the debtor countries (see 3.12).

Disaggregating European mechanisms’ announcements by type and controlling for Eu- rostoxx50 returns and the Euribor, we find that both announcements of future loan disbursements and of current-period disbursement significantly reduce spillovers for Ireland (3.13) and Portugal (3.14) providing evidence against anticipatory effects. Neither of the coefficients is significant for Greece (3.15) and Cyprus (3.16).

120 Chapter 3 Essays on Macroeconomics and Development

3.4 A recursive contract model with spillover costs

In the present section, we perform an alteration to an otherwise standard recursive contract model with limited commitment to understand the implications of spillover eﬀects for contractual terms. Building on Muller et al.(2019), we represent the lending agreement between a lender and a sovereign country (borrower) as a recursive dynamic contract inﬂicted by one sided limited commitment.

In our dynamic principal-agent framework, the principal is the lending mechanism and the agent is the distressed sovereign country requesting ﬁnancial assistance. Throughout the analysis, the lender is assumed to be fully committed, while the borrower can leave the contract at any time and resort to autarky. As a result, the lender has to design the contractual terms in such a way that the borrower always honors them at every state.

The limited commitment friction is represented through the addition of a participation constraint which has to be satisﬁed, in order for the contract to be self-enforcing. The critical idea is the interpretation of a binding participation constraint as a credible threat of default. Under a binding participation constraint the borrower is indiﬀerent between between honoring the contract the contract and resorting to autarky. Hence, defaulting on lending terms becomes a credible contingency. The Principal’s optimal response to a binding participation constraint, within a textbook model of one-sided limited commitment, is an increase in consumption and promised utility. This can be interpreted as a renegotiation of the loan terms. As Muller et al.(2019) put it, the lender "sweetens the deal" when the sovereign credibly threatens to default.

Extending the idea of a binding participation constraint as a credible threat of default, we focus on its subsequent effects on lender’s profits due to the existence of spillover costs. In our theoretical model, the lender incurs a certain cost whenever the participation constraint binds. This cost is the result of lender’s exposure to the borrower and aims to capture the losses incurred by the lender when the contingency of borrower’s default arises. In addition, the theoretical model allows to examine the response of the lender when this spillover cost exists and when it is absent. In this way, we are able to study in what way loan terms differ when the lender is vulnerable to spillover costs due to high exposure to the borrower compared to the case of a well protected lender without spillover effects from a borrower’s default.

A Model

∞ Every period the agent receives a stochastic endowment stream {yt}t=0, where for each t ≥ 0 yt is iid according to the discrete probability distribution P rob(yt = ys) = Πs, where

Chapter 3 121 Essays on Macroeconomics and Development

s ∈ {1, 2, ..., S}, with ys < ys+1, ∀s. In each period, the borrower gives its endowment to the lender, receiving back consumption (ct) and a stream of future utilities namely promised utility (wt).

In each period, the borrower can either stay within the contract and receive current consumption (ct) and promised utility (wt) or renege the contractual agreement and live in autarky - consuming its endowment forever after. Therefore the participation constraint (PC) takes the following form:

u(ct) + βwt ≥ u(yt) + βvaut ∀t [PC]

P∞ t PS Where vaut = t=0 β s=1 Πsu(yt), is the discounted value of future utility derived from consumption of own endowment. Essentially, the participation constraint ensures that at each period and state, the contract option is at least equal to the outside option, namely autarky.

We assume that the sole friction of the model is limited commitment from the side of the borrower, while the lender is fully committed to the contract. An additional constraint of the model in recursive form is that the lender delivers the past promised utility, namely the promise keeping constraint (PKC).

S X Πs u(ct) + βwt ≥ v [PKC] s=1

Where v is the promised utility with which the contract enters current period and it acts as a state variable. The promise keeping constraint makes sure that what the contract allocates to the borrower is at least equal to what has been promised to the borrower (v).

Moreover, we add another state variable sˆ = s : ys = max{y0, y1, ..., yt} which records the maximum state realised up until the current period. This implies that at any state s, if next period’s state s0 is higher than sˆ, then the participation constraint will be binding, while if s0 is lower than sˆ, then the participation constraint will be slack. Due to the iid nature of the endowment’s stochastic process the participation constraint at the future state s0 will be binding or slack according to the following probability distribution.

 Psˆ−1 slack, with prob = s=1 Πj PC at s0 =  PS binding, with prob = s=ˆs Πj

This is critical for the functioning of the model, since the future contingency of a binding participation constraint gives rise to a credible threat of default from the side of the borrower and the lender is taking this into account when designing the loan terms.

122 Chapter 3 Essays on Macroeconomics and Development

The model departs from the textbook one-sided limited commitment structure due to the existence of spillover cost. As already mentioned, the spillover cost arises in the case of a binding participation constraint and mirror the losses incurred by a lender who is exposed to the borrower. Therefore the proﬁt of the lender is represented by the following value function.

S sˆ−1 S X X 0 X 0 0 P (v, sˆ) = max Πj{yj − cj} +β ΠjP (wj, sˆ ) + Πj[P (wj, sˆ ) −ζ(wj, sˆ )] {cj ,wj } j=1 j=1 j=ˆs | {z } | {z } | {z } spillovers when PC binds expected current transfer exp.prof. (PC slack) | {z } exp.prof. (PC binds)

At each period, the lender is maximizing its profit function over current consumption and promised utility allocated to the borrower. The profit function consists of the current transfer denoted as the difference between the currently realised endowment and the corresponding current consumption allocated to the borrower and the discounted expected Psˆ−1 profit which is divided in two parts. First, with probability s=1 Πj, the borrower is realised with an endowment lower to the maximum endowment that has been realised up to date and hence the participation constraint is slack. Under this case, no credible threat of default arises. With certainty, the borrower finds it optimal to stay within the PS contract. Second, with probability s=ˆs Πj the borrower next period is realised with an endowment that is higher than what has been realised so far and hence the participation constraint is binding. Under this case, the outside option of autarky now appears to be more lucrative and the threat of default becomes credible.

Due to the exposure of the lender to the borrower, spillover cost arises. This is denoted by the ζ(wj, sˆ) function with the following properties, strictly decreasing and convex in future promised utility (ζw < 0, ζww > 0). The fact that spillover cost is decreasing in future promised utility is rationalised through the fact that the more favourable future terms the lender oﬀers to the borrower, the less lucrative the outside option appears to the borrower and thus the contingency of default appears to be more distant.

The problem takes the following form, in which the lender maximizes its objective function with respect to current consumption and future promised utility under the promise keeping

Chapter 3 123 Essays on Macroeconomics and Development and participation constraint. S sˆ−1 S X X 0 X 0 0 P (v, sˆ) = max Πj{yj − cj} + β ΠjP (wj, sˆ ) + Πj[P (wj, sˆ ) − ζ(wj, sˆ )] {c ,w } j j j=1 j=1 j=ˆs S X s.t. Πj{u(cj) + βwj} ≥ v [PKC] j=1

u(cj) + βwj ≥ u(yj) + βvaut , ∀j [PC]

wj ∈ [vaut, v¯]

sˆ(t) = {j : yj = max{y0, y1, ..., yt}}

Proposition 1: For a given promised utility v, when the participation constraint is non- binding, the constrained eﬃcient allocation prescribes constant consumption and promised utility and equal to cs = g1(v) and ws = v. When the participation constraint binds then consumption, promised utility satisfy equations1 and2 respectively .

0 Πj u (cj) = − (3.1) θΠj + λj

0 0 λj P (wj) − ζ (wj) = −(θ + ) (3.2) Πj where λj is the state contingent lagrange multiplier assigned to the participation constraint and θ is the lagrange multiplier assigned to the promise keeping constraint.

Proof: See Appendix

Comparative statics with and without spillover eﬀects

We are primarily interested in juxtaposing a lending agreement between a lending mechanism and a sovereign country as the borrower under the following two cases. The ﬁrst refers to the model described in the previous section, in which the lender incurs spillover costs in the contingency of a default from the side of the sovereign. The second refers to the absence of spillover cost and represents cases in which the lender is not exposed to the borrower’s economy.

The former case can be thought as lending agreements between members of the same union, in which contracting economies are highly interlinked. The latter case eﬀectively represents agreements between a sovereign and an external organisation whose exposure level to the borrower’s economic condition is low. Under this case, spillover costs are unlikely to arise and most importantly to aﬀect the lending terms and conditions.

In theoretical terms, the latter case of absent spillover cost is represented by a standard 0 one-sided limited commitment model without the convex cost function (ζ(wj, sˆ )) in the

124 Chapter 3 Essays on Macroeconomics and Development lender’s objective function (Ljungqvist and Sargent, 2018). In the present section, we attempt to trace the diﬀerences in the choice variables of the two models and infer in what way the existence of spillover cost aﬀects the conditionality of the resulting contract.

We focus on the case of a binding participation constraint and the corresponding first order conditions, because this is the case in which the default threat becomes credible and the two models differ in the subsequent effect on the lender’s objective function.

No Spillover Spillover 0 no spil 1 0 spil 1 foc wrt cj u (cj ) = − 0 no spil u (cj ) = − 0 spil 0 spil P (wj ) P (wj )−ζ (wj ) no spil spil 0 no spil 0 λj 0 spil 0 spil 0 λj foc wrt wj P (w ) = P (v) − P (w ) − ζ (w ) = P (v) − j Πj j j Πj

Table 3.3: First order conditions under a binding PC in the two relevant cases

The ﬁrst order condition with respect to current consumption shows that the lender equates its marginal rate of transformation between current consumption and future 1 0 promised utlity (− 0 ) to the marginal rate of substitution of the borrower (u (cj)) P (wj ) in both cases. Additionally the form of the ﬁrst order condition with respect to future promised utility has taken this form after using the envelope condition (see appendix).

We start our analysis from the future promised utility under the two speciﬁcations with the following proposition.

Proposition 2: For the same past promised utility v, the contract with present spillover cost allocates higher promised utility and lower current consumption compared to the contract with absent spillover cost. spil no spil wj > wj spil no spil cj < cj Proof: See Appendix

3.5 Discussion of model predictions

From the preceding analysis, we infer that the presence of spillover costs aﬀects the loan terms diﬀerently when we distinguish current consumption from future promised utility. On the one hand, current consumption in the spillover case is lower than what would be prescribed in the absence of inter-linkages between the lender and the borrower. On the other hand, future promised utility is higher in the presence of spillovers.

As a result, the borrower’s bargaining power from the spillover cost incurred by the lender manifests its beneﬁts only in the future. Essentially, the presence of spillovers

Chapter 3 125 Essays on Macroeconomics and Development enhances the main characteristic of the one-sided limited commitment model of "back- loading consumption" in the future. Theoretically, the reason why this result occurs, is due to the anticipation of a credible threat of borrower’s default from the lender. The presence of spillover cost in the contingency of a binding participation constraint induces the lender to increase future promised utility in order to render the contract more attractive option compared to autarky.

The reason why the lender does so through promised utility - i.e. future consumption and not current consumption is due to lower costs. The lower costs emerge from the 0 spillover cost function (ζ(wj, sˆ )) being decreasing in wj. As a result, when relaxing the participation constraint through current consumption, leads to a decrease in current proﬁts for the lender, when doing so by increasing promised utility the lender can also enjoy the beneﬁts from lower spillover costs.

This is an interesting outcome of the model which suggests different loan terms in lending agreements within a union with strong links between the borrower and the lender, compared to lending agreements with an external institution not exposed to the borrower. In the former case, the model predicts benefits for the borrower in the long term, while in the latter, the borrower is better off in the short term.

A Empirical Validation of Model Predictions

The European debt crisis constitutes a suitable environment to test the predictions of the model. During the European debt crisis, the IMF and the European mechanisms collaborated closely to address the financing needs of the countries in trouble. Even though the two entities agreed upon the structural reforms to be followed by the borrowing countries, they provided loans of different amounts and different terms. We attribute this difference to the different economic interconnectedness between each borrowing country and the lending mechanism. On the one hand, the IMF does not incur any spillover cost, since it constitutes an external institution funded by a sizeable number of countries whose interests are affected by the European crisis at a minimum level. On the other hand, European lending mechanisms are comprised of European economies whose interests are very much aligned and directly linked to each other.

Spillovers and the disbursement structure of European lending programs

Our aim is to document how the risk of spillovers across euro area countries aﬀects the speed with which loan payments are disbursed. To do so, we construct a daily measure of current consumption for which we calculate the ratio of cumulative disbursements and the total loan amount promised by the European lending mechanisms. This provides us

126 Chapter 3 Essays on Macroeconomics and Development with an indicator of how much is paid out to a certain country in the current period relative to how much a country expects to still receive in the future. The data for both, the numerator and the denominator, is obtained from reading press releases of the ESM and the Euopean Commission and are part of the dataset described in 3.2. Both take into account disbursements and promised loans through the EFSF, the EFSM, the ESM and the Greek Loan facility (GLF). The denominator remains largely unchanged for Cyprus, Ireland and Portugal, i.e. the amounts promised at the start of the lending programs did not change. However, for Greece the denominator changes with each of the three lending programs (GLS, EFSM and ESM). E.1 plots our measures of spillovers and current consumption for Greece, Portugal, Ireland and Cyprus over time, revealing that except for Ireland the cumulative amounts disbursed fell short of what was initially promised.13

To empirically validate our theoretical prediction that the presence of spillovers costs results to lower current consumption, we run the following panel regression:

MA Current Consumtpionit = c + βDCCit + uit (3.3)

We regress Current Consumption, which is the measure of disbursements that we discuss above, on past moving averages of the DCC over 30 and 90 days before period t. We use this measure to mitigate potential problems with reverse causality and frequently fluctuating DCCs measures. However, we also include the results of the standard measure of DCC. 3.10 displays the results from the panel regression which shows that higher spillovers are negatively related to current consumption. Hence, posing a threat to other countries in the euro area seemed to have influenced the timing of disbursements which were back-loaded, i.e. relatively higher consumption was promised in the future. The regression coefficients remain negative and significant at the 5% and 10% level respectively for all of our DCC measures. Furthermore, our results are robust to linear de-trending.

Overall, we observe that the effect of spillovers on "back-loading" consumption is present under the European financial assistance programme. Table 3.10 presents the results from regressing the level of spillovers and a measure of current consumption disbursements. Higher spillover costs imply lower current disbursement, a result that is robust under various specifications using the moving average of spillover costs the past 30 and 90 days respectively. By construction this result implies that higher spillover costs imply higher consumption allocated to future transfers.

13Note that disbursements started on a diﬀerent day for each crisis countries.

Chapter 3 127 Essays on Macroeconomics and Development

3.6 Conclusion

In this paper we assess how the risk of contagion aﬀected the design of lending contracts to crisis-hit countries within the euro-area. We introduce a simple recursive contract model with spillover costs. We ﬁnd that the presence of contagion reduces current consumption for the country in trouble, and back-loads future consumption.

Moreover, we present a newly built dataset that records all the dates that official meetings of the IMF and the Europeans took place, and official announcements were made. This record includes rich information on the loan conditions that were offered in the different programme-countries and the timing regarding the realization of a loan disbursement.

Overall, our results indicate that accounting for spillover costs in the design of lending contracts is relevant. We support our argument by showing evidence that the European Mechanisms’ decision to provide ﬁnancial support successfully mitigated spillovers.

128 Chapter 3 Essays on Macroeconomics and Development

References

Abrahám, A., E. Cárceles Poveda, Y. Liu, and R. Marimon (2018). On the optimal design of a ﬁnancial stability fund.

Alvarez, F. and U. J. Jermann (2000). Eﬃciency, equilibrium, and asset pricing with risk of default. Econometrica 68 (4), 775–797.

Atkeson, A. (1991). International lending with moral hazard and risk of repudiation. Econometrica: Journal of the Econometric Society, 1069–1089.

Beirne, J. and M. Fratzscher (2013). The pricing of sovereign risk and contagion during the european sovereign debt crisis. Journal of International Money and Finance 34, 60–82.

Constancio, V. et al. (2012). Contagion and the european debt crisis. Financial Stability Review 16, 109–121.

Corsetti, G., A. Erce, and T. Uy (2017). Oﬃcial Sector Lending Strategies During the Euro Area Crisis. ADEMU Working paper series No. 070.

Darvas, Z. (2017). Regional and global ﬁnancial safety nets : the recent European experience and its implications for regional cooperation in Asia. Bruegel working paper No. 06.

Diebold, F. X. and K. Yılmaz (2009). Measuring Financial Asset Return and Volatility Spillovers, with Application to Global Equity Markets. The Economic Journal 119 (1), 158–171.

Dovis, A. (2019). Eﬃcient sovereign default. The Review of Economic Studies 86 (1), 282–312.

Eaton, J. and M. Gersovitz (1981). Debt with potential repudiation: Theoretical and empirical analysis. The Review of Economic Studies 48 (2), 289–309.

Eichengreen, B. (2015). The Irish Crisis and the EU from a Distance. IMF seminar presentation January 20, 1–14.

Engle, R. (2002). Dynamic conditional correlation: A simple class of multivariate generalized autoregressive conditional heteroskedasticity models. Journal of Business & Economic Statistics 20 (3), 339–350.

Forbes, K. J. (2012). The “big c”: identifying and mitigating contagion. In Proceedings- Economic Policy Symposium-Jackson Hole, pp. 23–87. Federal Reserve Bank of Kansas City.

Chapter 3 129 Essays on Macroeconomics and Development

Gourinchas, P. O., P. Martin, and T. Messer (2018). The economics of sovereign debt, bailouts and the eurozone crisis. In West Coast Workshop International Finance & Open Economy Macroeconomics. Working Paper, University of California, Berkeley, Berkeley, CA.

IMF (2017). Collaboration between Regional Financiang arrangements and the IMF. IMF Background paper July 2015.

Kehoe, P. J. and F. Perri (2002). International business cycles with endogenous incomplete markets. Econometrica 70 (3), 907–928.

Ljungqvist, L. and T. J. Sargent (2018). Recursive macroeconomic theory. MIT press.

Muller, A., K. Storesletten, and F. Zilibotti (2019). Sovereign debt and structural reforms. American Economic Review 109 (12), 4220–59.

Thomas, J. and T. Worrall (1994). Foreign direct investment and the risk of expropriation. The Review of Economic Studies 61 (1), 81–108.

130 Chapter 3 Essays on Macroeconomics and Development

3.7 Appendix

A Financial linkages across the euro area

The following figures A.1- A.5 illustrate the size of outstanding bank claims for each of the five crisis in the year in which their respective first financial assistance program was agreed upon by the European lending mechanisms. The data is taken from the BIS Consolidated banking statistics and we express the level of indebtedness of banks in Cyprus, Ireland, Spain, Portugal and Greece as shares of creditors’ GDP. In each figure the debtor country is coloured in green and debtor countries take different shades of red, where darker shades reflect higher claims outstanding. No data was available for countries coloured in grey.

What becomes apparent from these graphs is that the main creditors were Germany, France, the Netherlands and Belgium. Overall, the most striking bilateral debt relation- ships are: Germany that held Spanish banking debt worth 49% of German GDP and France that held Greek banking debt worth 28% of French GDP.

Furthermore, there seem to be some "neighborhood" eﬀects in the sense that Portugal was especially exposed to Spanish banking debt (10.4% of Portuguese GDP) and vice versa (Spain held Portuguese banking debt worth 5.6% of Spanish GDP). Similarly, Greece held Cypriot debt worth 4.8% of Greek GDP. Furthermore, the banking sectors with the largest debt abroad (wrt. other Euro area creditors in our sample) were Ireland (124% of its GDP) and Cyprus (99% of its GDP). However, measured as a share of creditor countries’ GDP, Cypriot banks posed the lowest threat to Euro area members compared to the other four crisis countries which is partly driven by the small size of the Cypriot economy.

Chapter 3 131 Essays on Macroeconomics and Development

Figure A.1: Outstanding claims on Irish banks as a share of creditor country’s GDP in 2010 (Data source: BIS Consolidated banking statistics)

132 Chapter 3 Essays on Macroeconomics and Development

Figure A.2: Outstanding claims on Cypriot banks as a share of creditor country’s GDP in 2011 (Data source: BIS Consolidated banking statistics)

Chapter 3 133 Essays on Macroeconomics and Development

Figure A.3: Outstanding claims on Greek banks as a share of creditor country’s GDP in 2011 (Data source: BIS Consolidated banking statistics)

134 Chapter 3 Essays on Macroeconomics and Development

Figure A.4: Outstanding claims on Spanish banks as a share of creditor country’s GDP in 2011 (Data source: BIS Consolidated banking statistics)

Chapter 3 135 Essays on Macroeconomics and Development

Figure A.5: Outstanding claims on Portuguese banks as a share of creditor country’s GDP in 2011 (Data source: BIS Consolidated banking statistics)

136 Chapter 3 Essays on Macroeconomics and Development

Figure A.6: Share of general government gross debt held by foreign investors (Data source: IMF Sovereign Investor Base Dataset for Advanced Economies)

Chapter 3 137 Essays on Macroeconomics and Development

B Theoretical appendix

Assigning non-state contingent lagrange multiplier θ to the PKC, and a state contingent

λj to PC, I obtain the Lagrangian:

S S X X L = Πj {yj − cj} + βP (wj) − β Πjζ(wj)+ j=1 j=ˆs S X +θ Πj[u(cj) + βwj] − v j=1 S X + λj u(cj) + βwj − u(yj) − βvaut j=1

Note I have to break down the PC term in the Lagrangian from j = 1 → sˆ− 1 and sˆ → S

S S X X L = Πj {yj − cj} + βP (wj) − β Πjζ(wj)+ j=1 j=ˆs S X +θ Πj[u(cj) + βwj] − v j=1 sˆ−1 X + λj u(cj) + βwj − u(yj) − βvaut j=1 | {z } >0, PC slack, λj =0 S X + λj u(cj) + βwj − u(yj) − βvaut j=ˆs | {z } =0, PC binds, λj >0

t Note however, that I know that if in the next period j ∈ [1, max {s}0 −1], the PC is slack, t and λj = 0, while if j ∈ [max {s}0,S], the PC is binding, and λj > 0, so the Lagrangian can be written as:

S S X X L = Πj {yj − cj} + βP (wj) − β Πjζ(wj)+ j=1 t j=max{s}0 S X +θ Πj[u(cj) + βwj] − v j=1 S X + λj u(cj) + βwj − u(yj) − βvaut j=max {s}t | {z } 0 =0, PC binds, λj >0

I distinguish two cases with respect to whether the PC binds next period, and I take focs wrt to the relevant variables:

t Case 1: Non-binding PC , λj = 0 and j ∈ [1, max {s}0 − 1]

∂L 0 0 1 = 0 → Πj(−1) + θΠju (cj) = 0 → u (cj) = − ∂cj θ

138 Chapter 3 Essays on Macroeconomics and Development

∂L 0 0 = 0 → ΠjβP (wj) + θΠjβ = 0 → P (wj) = −θ ∂wj

t Case 2: Binding PC , λj > 0 and j ∈ [max {s}0,S]

∂L 0 0 0 0 Πj = 0 → Πj(−1) + θΠju (cj) + λsu (cj) = 0 → u (cj)[θΠj + λj] = −Πj → u (cj) = − ∂cj θΠj + λj

∂L 0 0 0 0 λj = 0 → ΠjβP (wj) − βΠjζ (wj) + θΠjβ + βλj = 0 → P (wj) − ζ (wj) + (θ + ) = 0 → ∂wj Πj

0 0 λj P (wj) − ζ (wj) = −(θ + ) Πj

Envelope Condition P 0(v) = −θ

Case 1: Since the PC does not bind, the promised utility that the principal sets remains constant and so does consumption. As a result, in the slack PC case, the contract attains its ﬁrst best allocation

00 0 0 P ()<0 P (wj) = −θ = P (v) −−−−→ wj = v

0 1 0 1 1 0−1 1 u (cj) = − → u (cj) = 0 = 0 → cj = u 0 = constant θ P (wj) P (v) P (v)

Case 2: Since the PC binds, the allocation receives a wedge represented by the lagrange multiplier on the PC. On top of this, when the PC binds, the threat of a default becomes credible, and hence spillover costs inﬂict the allocation.

0 Πj u (cj) = − θΠj + λj and 0 0 λj θΠj + λj P (wj) − ζ (wj) = − θ + = − Πj Πj Combining the two I obtain:

−1 0 0 0 u (cj) = P (wj) − ζ (wj)

Note that combining the foc wrt to promised utility with the envelope condition I obtain the following:

0 0 0 λj Env:P (v)=−θ) 0 0 0 λj P (wj) − ζ (wj) = − θ + −−−−−−−−→ P (wj) − ζ (wj) = P (v) − Πj Πj

Chapter 3 139 Essays on Macroeconomics and Development

spil no spil14 Proposition 2: Given that λj < λj the comparison focuses on the two equations:

spil 0 spil 0 spil 0 λj P (wj ) − ζ (wj ) = P (v) − [Spillovers case] Πs and no spil 0 no spil 0 λj P (wj ) = P (v) − [No Spillovers case] Πs Note that P 0(v) is constant and equal to −θ in both cases (from the envelope condition).

Subtracting the two expression we obtain the following:

spil no spil 0 spil 0 spil 0 no spil 0 λj 0 λj P (wj ) − ζ (wj ) − P (wj ) = P (v) − − P (v) + → Πs Πs no spil spil λspil<λno spil 0 spil 0 spil 0 no spil λj − λj j j P (wj ) − ζ (wj ) − P (wj ) = −−−−−−−→ Πs 0 spil 0 spil 0 no spil P (wj ) − ζ (wj ) −P (wj ) > 0 → | {z } <0, ∀wj 0 spil 0 spil 0 no spil P (wj ) − ζ (wj ) > P (wj ) | {z } <0, ∀wj

0 spil 0 spil We know that P (wj ) − ζ (wj ) < 0, so the last expression becomes:

0 no spil 0 spil 0 spil P (wj ) < P (wj ) − ζ (wj ) < 0 → 0 spil 0 spil 0 no spil ζ (wj ) < P (wj ) − P (wj ) < 0

Focusing on the second inequality:

0 spil 0 no spil P (wj ) − P (wj ) < 0 → 00 0 spil 0 no spil P ()<0 P (wj ) < P (wj ) −−−−→ spil no spil wj > wj

Hence, the presence of spillover cost, is leading to an increased promised utility. Following

14The Lagrange multiplier expresses the marginal effect to the objective value function by relaxing the constraint by one unit. Suppose that you relax the constraint by one unit increase in wj then in the 0 case of no spillovers, the objective function of the principal will change by P (wj), it will decrease by 0 that amount. Instead, in the case of present spillovers the objective function will also decrease by P (wj) 0 but there is also going to be an opposite effect amounting for −ζ (wj), partially offsetting the decrease in profits. Hence in total, relaxing the PC by one unit, causes a larger change in the case of no spillovers, compared to the case of spillovers.

140 Chapter 3 Essays on Macroeconomics and Development from above, 0 no spil spil 0 spil P (wj ) < P (wj − ζ (wj )) → 1 1 > → 0 no spil spil 0 spil P (wj ) P (wj − ζ (wj )) 1 1 − < − → 0 no spil spil 0 spil P (wj ) P (wj − ζ (wj )) 00 0 no spil 0 spil u ()<0 u (cj ) < u (cj ) −−−−→ no spil spil cj > cj

C Descriptive statistics from database on loan announcements and disbursements

Example statements

03/05/2010: Euro area and IMF agreement on ﬁnancial support programme for Greece

"To support the Greek government’s efforts to get its economy back on track, euro area Member States on 2 May pledged a three-year programme total of €80 billion in bilateral loans. Under the conditions set out in the Eurogroup statement of 11 April, up to 30 billion out of this programme will be made available for 2010. Its first disbursement will be made by 19 May. In addition, IMF reached an agreement with the Greek authorities to support this program with a stand-by arrangement of about 30 billion, bringing the joint commitment to a total financing of 110 billion." (available here: https: // ec. europa. eu/ economy_ finance/ articles/ eu_ economic_ situation/ 2010-05-03-statement-commissioner-rehn-imf-on-greece_ en. htm )

14/09/2011: Commission proposes better ﬁnancial terms for EU loans to Ire- land and Portugal

"Two proposals were adopted by the European Commission today, suggesting reduced interest rate margins and extended maturities for loans granted by the European Union (EU) to Ireland and Portugal. The loans are provided by the EU under the European Financial Stabilisation Mechanism (EFSM) as part of ﬁnancial assistance packages to the two countries. (...) similar conditions are expected to be adopted for the lending that the European Financial Stability Facility (EFSF) is providing to Ireland and Portugal. (...) Both countries should pay lending rates equal to the funding costs of the EFSM, i.e. reducing the current margins of 292.5 bps for Ireland and of 215 bps for Portugal to zero. The reduction in margin will apply to all instalments, i.e. both to future and to already disbursed tranches. Furthermore, the maturity of individual future tranches

Chapter 3 141 Essays on Macroeconomics and Development to these countries will be extended from the current maximum of 15 years to up to 30 years. As a result the average maturity of the loans to these countries from EFSM would go up from the current 7.5 years to up to 12.5 years. In addition to the substantial cash savings for Ireland and Portugal, the new financial terms will bring benefits such as enhanced sustainability and improved liquidity outlooks. Moreover, indirect confidence effects through the enhanced credibility of programme implementation should result in improved borrowing conditions for the sovereign as well as the private sector. (...)" (available here: https: // ec. europa. eu/ commission/ presscorner/ detail/ en/ MEMO_ 11_ 602)

03/06/2013: Statement by the EC and the ECB following the conclusion of the third review of the ﬁnancial assistance programme for Spain

"A delegation from the European Commission, in liaison with the European Central Bank, the European Stability Mechanism and the European Banking Authority, carried out the third review of the financial sector assistance programme for Spain from 21 May to 31 May 2013. The International Monetary Fund also participated in the review, fulfilling its role as an independent monitor. On the basis of the review, it can be concluded that the programme remains on track. Spanish financial markets have further stabilised since the last review, with sovereign and corporate bond yields dropping amidst lower volatility. In parallel, the liquidity situation of the Spanish banking sector has further improved. This allowed Spanish banks to further regain access to funding markets and to reduce reliance on central bank financing. (...) Progress has also continued with respect to horizontal financial-sector conditionality. Thereby, compliance with the requirements in the Memo- randum of Understanding is nearly complete and achievements toward strengthening the governance, regulatory and supervisory framework of the Spanish banking sector have been made. (...) The next review is foreseen to take place in September 2013." (available here: https: // ec. europa. eu/ commission/ presscorner/ detail/ en/ MEMO_ 13_ 489)

Descriptive statistics

142 Chapter 3 Essays on Macroeconomics and Development

EC announcem. IMF announcem. All Future loan Disbursem. All CYP 28 13 10 41 ESP 49 4 2 n.A. GRE 60 15 5 57 IRL 57 18 21 51 PRT 46 12 10 55 1 Notes: The ﬁrst four columns contain a count of the number of days on which a statement was issues by the European lending mechanisms (via the EC and/or ESM website) and the IMF between April 29, 2009 and September 20, 2018. For the former creditor, statements are further dis-aggregated into a general count of "all" announcements, statements announcing future loan payments as well as those referring to disbursements made in the current period. Note that announcements of future and current loan disbursements may occur on the same day.

Table 3.4: Descriptive statistics of loan announcements and disbursements

Dynamics of announcements

Chapter 3 143 Essays on Macroeconomics and Development

Figure C.1: Frequency of announcements by country

1 Notes: "EC announcements" and "IMF

announcements" stand for statements issued by the EC (which may be on behalf of the ESM, EFSF or EFSM) and the IMF, respectively. For the former creditor group, statements are further dis-aggregated into statements announcing future loan payments as well as those referring to disbursements made in the current period.

144 Chapter 3 Essays on Macroeconomics and Development 17.75 49.87 32.16 104.45 116.77 147.03 202.20 168.43 171.63 54.09% From Others Contribution Spillover Index 2.60 8.00 5.40 ESP 20.15 20.79 32.73 58.64 40.10 94.01 191.07 2.97 6.99 5.82 16.92 18.49 31.37 46.23 98.24 42.78 PRT 168.89 ) 3.42 9.27 5.45 ITA 24.23 25.14 34.36 98.04 45.86 54.29 202.02 Sovereign CDS ∆( 2.28 7.50 4.50 IRL 16.58 17.59 95.05 34.03 32.15 28.88 143.52 Generalized 2.72 0.42 4.27 2.67 4.72 5.80 5.46 4.52 98.72 GRE 29.87 Table 3.5: 4.85 0.65 2.94 6.95 9.08 6.72 6.90 10.50 94.60 DEU 48.59 2.90 9.47 4.90 18.02 91.42 17.54 23.47 16.96 17.66 FRA 110.93 0.98 1.22 0.41 0.14 1.73 1.92 1.01 0.91 8.31 96.13 CYP 2.50 5.56 3.02 BEL 91.40 18.77 17.62 23.73 17.49 18.38 107.07 ) IRL ITA ESP BEL PRT FRA CYP GRE DEU To Others Generalized Sovereign CDS Contribution ∆( D Empirical analysis Evidence of Contagion - VAR Analysis and Spillover Index

Chapter 3 145 Essays on Macroeconomics and Development 91.47 39.75 397.79 352.24 389.15 340.25 157.73 247.04 366.48 265.07 76.17% From Others Contribution Spillover Index 5.68 ESP 38.83 13.72 30.67 23.17 31.74 60.74 26.57 81.91 32.70 263.83 9.69 2.74 61.10 78.29 77.31 15.37 23.32 81.23 25.91 NLD 79.51 373.25 8.22 ITA 33.51 15.62 27.48 15.82 21.53 86.45 20.51 61.52 27.88 232.09 6.63 IRL 18.32 10.70 14.92 11.00 82.83 21.56 12.03 33.10 17.34 145.60 1.41 4.14 2.30 2.47 4.99 7.77 2.52 5.79 74.45 GRE 4.00 35.39 6.71 3.04 58.64 74.68 81.70 13.60 18.73 77.03 23.19 DEU 74.68 345.31 10 yr Government Bonds Total Returns 3.06 64.39 10.75 83.94 72.34 16.81 30.00 75.76 28.96 FRA Table 3.6: 79.82 381.10 2.56 FIN 16.30 87.16 16.68 12.95 15.54 18.31 15.21 17.87 15.59 131.02 9.94 3.74 BEL 84.24 64.67 58.05 19.36 36.38 60.32 38.25 66.18 351.98 AUT 84.23 64.83 10.19 72.13 72.13 4.09 18.79 30.22 76.52 30.48 386.62 IRL ITA FIN ESP BEL FRA NLD GRE AUT DEU To Others Total Returns Contribution 10 yr Gov Bonds

146 Chapter 3 Essays on Macroeconomics and Development 332.61 354.68 214.34 474.19 402.39 220.97 246.11 390.45 433.99 377.11 79.38% From Others Contribution Spillover Index ESP 39.56 35.54 21.26 59.04 48.47 34.88 22.44 60.85 47.88 90.97 369.92 42.18 62.03 30.79 72.26 60.51 24.05 37.67 51.65 87.46 48.85 NLD 430.01 ITA 40.28 33.97 28.06 64.48 50.52 30.52 22.39 88.10 51.22 61.95 383.39 IRL 30.97 39.71 15.62 33.14 29.94 12.23 90.49 22.93 37.86 21.85 244.24 30.06 18.19 10.39 29.10 23.52 89.96 10.94 29.59 23.69 34.13 GRE 209.61 Equity Total Returns 39.65 46.45 35.57 74.49 82.98 25.10 32.62 54.53 64.15 52.40 DEU 424.95 Table 3.7: 45.71 51.61 42.67 89.33 70.48 30.31 34.24 65.06 72.21 60.44 FRA 472.72 FIN 16.54 20.32 90.71 48.88 41.08 11.33 19.28 32.08 36.33 24.55 250.39 BEL 47.66 93.12 16.68 49.74 42.21 18.13 36.94 34.11 59.38 35.07 339.92 92.15 46.88 13.29 43.07 35.67 34.43 29.62 39.62 41.26 37.87 AUT 321.71 IRL ITA FIN ESP BEL FRA NLD GRE AUT DEU Equity To Others Total Returns Contribution

Chapter 3 147 Essays on Macroeconomics and Development 17.21 74.13 43.28 220.73 221.41 181.53 141.66 208.11 58.77% From Others Contribution Spillover Index 2.92 7.73 52.95 56.50 16.67 40.20 29.90 96.43 NLD ) 206.87 7.47 7.78 0.82 3.45 6.83 7.24 7.27 IRL Bank CDS 98.46 40.85 ∆( 1.97 8.95 7.46 27.93 29.08 26.17 95.71 28.11 PRT 129.69 Generalized 2.43 7.87 ITA 48.39 46.93 12.57 97.66 30.21 41.77 190.18 Table 3.8: 2.69 3.76 BEL 16.48 15.07 94.89 12.56 10.32 16.35 77.24 2.87 1.80 2.68 1.86 2.07 0.88 2.14 98.38 GRE 14.31 2.74 7.79 64.64 98.57 14.07 46.85 31.29 57.96 FRA 225.34 3.64 7.79 97.32 64.24 15.73 47.06 30.63 54.49 DEU 223.58 ) CDS IRL ITA BEL PRT FRA NLD GRE DEU Bank To Others Generalized ∆( Contribution

148 Chapter 3 Essays on Macroeconomics and Development 46 304 272 235 242 275 327 325 335 73.63% From Others Contribution ) Spillover Index 357 5.94 97.4 ESP 53.86 42.71 36.31 39.89 48.18 65.29 64.89 Sovereign CDS ∆( 5.4 336 53.4 45.28 41.38 34.35 45.47 47.87 95.53 63.31 PRT 7.4 332 ITA 53.58 38.66 34.54 38.11 44.34 93.90 52.73 62.82 271 37.2 3.18 45.3 IRL 30.25 26.97 37.54 94.49 42.67 47.69 249 4.19 31.54 29.17 22.31 94.94 39.95 38.33 44.62 38.68 GRE Subsample (2008-2011) Generalized 221 4.87 35.27 41.20 89.44 20.17 23.98 33.69 30.02 32.15 DEU Table 3.9: 264 6.76 30.6 43.04 91.78 41.58 27.13 37.30 39.13 38.31 FRA 22 3.25 1.91 2.42 1.98 4.28 1.38 2.29 4.018 92.07 CYP 308 8.13 BEL 95.60 45.02 37.03 31.30 38.44 51.71 44.39 51.84 ) IRL ITA ESP BEL PRT FRA CYP GRE DEU To Others Generalized Sovereign CDS Contribution ∆(

Chapter 3 149 Essays on Macroeconomics and Development

E Spillovers and loan conditions

Table 3.10: Results from regressing a measure of current consumption on spillovers (DCC).

(1) (2) (3) Cum.loans/Overall Cum.loans/Overall Cum.loans/Overall DCC -2.049∗∗ (-7.61)

MA DCC[-30] -2.136∗∗ (-6.36)

∗ MA DCC[-90] -2.242 Notes: (-5.06)

Constant 1.227∗∗∗ 1.249∗∗∗ 1.278∗∗ (19.89) (16.16) (12.41) Observations 7716 7716 7716 t statistics in parentheses ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

"DCC" is the current dynamic conditional correlation (DCC) of sovereign CDS spreads whereas "MA DCC[-30]" and "MA DCC[-90]" denote the moving average over the last 30 days and 90 days, respectively. Countries included in the panel: Greece, Ireland, Portugal, Cyprus. We include ﬁxed eﬀects and cluster standard errors at the country level.

150 Chapter 3 Essays on Macroeconomics and Development

Figure E.1: Spillover measures and current consumption by country

1 Notes: The current consumption ratio is calculated as the ratio of cumulative loan amounts disbursed

over the total amounts promised by the European lending mechanisms.

Chapter 3 151 Essays on Macroeconomics and Development

F The eﬀect of announcements on spillovers - Summary tables

Table 3.11: Results from regressing the DCC on EFSM/EFSF/ESM (EC) announcements

(1) (2) (3) (4) IRL_DCC GRE_DCC PRT_DCC CYP_DCC EC announcements -0.0338∗∗∗ -0.0381∗∗ -0.0430∗∗∗ 0.00697 (-3.49) (-2.99) (-3.48) (1.40)

Constant 0.321∗∗∗ 0.109∗∗∗ 0.308∗∗∗ 0.0922∗∗∗ (193.28) (48.19) (166.38) (151.52) Observations 2452 2452 2452 2452 t statistics in parentheses ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001 Controls: Euribor, Eurostoxx50 return

Table 3.12: Results from regressing the DCC on IMF announcements

(1) (2) (3) (4) IRL_DCC GRE_DCC PRT_DCC CYP_DCC IMF announcements -0.0151 -0.0000469 -0.0185 -0.00451 (-1.47) (-0.00) (-1.70) (-1.09)

Constant 0.321∗∗∗ 0.108∗∗∗ 0.308∗∗∗ 0.0923∗∗∗ (192.52) (47.85) (165.04) (151.64) Observations 2452 2452 2452 2452 t statistics in parentheses ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001 Controls: Euribor, Eurostoxx50 return

152 Chapter 3 Essays on Macroeconomics and Development

G The eﬀect of announcements on spillovers - Country details

Table 3.13: Results from regressing the DCC for Ireland on EFSM/EFSF/ESM and IMF lending announcements referring to Ireland

(1) (2) (3) (4) IRL_DCC IRL_DCC IRL_DCC IRL_DCC All EFSM/EFSF/ESM announcem. -0.0338∗∗∗ (-3.49)

EFSM/EFSF/ESM loan announcem. -0.0392∗ (-2.28)

EFSM/EFSF/ESM loan disbursem. -0.0441∗∗ (-2.78)

IMF announcements -0.0151 (-1.47)

Eurostoxx50 return 0.00142 0.00145 0.00171 0.00142 (1.25) (1.27) (1.50) (1.25)

Euribor 0.131∗∗∗ 0.131∗∗∗ 0.131∗∗∗ 0.130∗∗∗ (51.67) (51.47) (51.58) (51.50)

Constant 0.321∗∗∗ 0.321∗∗∗ 0.321∗∗∗ 0.321∗∗∗ (193.28) (193.39) (193.47) (192.42) Observations 2452 2452 2452 2452 t statistics in parentheses ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

Chapter 3 153 Essays on Macroeconomics and Development

Table 3.14: Portugal

(1) (2) (3) (4) PRT_DCC PRT_DCC PRT_DCC PRT_DCC All EFSM/EFSF/ESM announcem. -0.0430∗∗∗ (-3.48)

EFSM/EFSF/ESM loan announcem. -0.0503∗ (-2.17)

EFSM/EFSF/ESM loan disbursem. -0.0622∗ (-2.56)

IMF announcements -0.0185 (-1.70)

Eurostoxx50 return 0.00137 0.00133 0.00114 0.00131 (1.08) (1.05) (0.90) (1.03)

Euribor 0.135∗∗∗ 0.135∗∗∗ 0.135∗∗∗ 0.135∗∗∗ (48.11) (47.96) (47.99) (47.90)

Constant 0.308∗∗∗ 0.307∗∗∗ 0.307∗∗∗ 0.308∗∗∗ (166.38) (166.38) (166.49) (165.04) Observations 2452 2452 2452 2451 t statistics in parentheses ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

154 Chapter 3 Essays on Macroeconomics and Development

Table 3.15: Greece

(1) (2) (3) (4) GRE_DCC GRE_DCC GRE_DCC GRE_DCC All EFSM/EFSF/ESM announcem. -0.0381∗∗ (-2.99)

EFSM/EFSF/ESM loan announcem. -0.00708 (-0.28)

EFSM/EFSF/ESM loan disbursem. 0.0377 (0.86)

IMF announcements -0.0000469 (-0.00)

Eurostoxx50 return 0.000473 0.000494 0.000495 0.000499 (0.31) (0.32) (0.32) (0.32)

Euribor 0.218∗∗∗ 0.217∗∗∗ 0.217∗∗∗ 0.217∗∗∗ (63.73) (63.48) (63.48) (63.35)

Constant 0.109∗∗∗ 0.108∗∗∗ 0.108∗∗∗ 0.108∗∗∗ (48.19) (48.05) (48.06) (47.85) Observations 2452 2452 2452 2452 t statistics in parentheses ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

Chapter 3 155 Essays on Macroeconomics and Development

Table 3.16: Cyprus

(1) (2) (3) (4) CYP_DCC CYP_DCC CYP_DCC CYP_DCC All EFSM/EFSF/ESM announcem. 0.00697 (1.40)

EFSM/EFSF/ESM loan announcem. -0.00641 (-0.88)

EFSM/EFSF/ESM loan disbursem. -0.00159 (-0.19)

IMF announcements -0.00451 (-1.09)

Eurostoxx50 return 0.0000396 0.0000338 0.0000307 0.0000385 (0.10) (0.08) (0.07) (0.09)

Euribor 0.00345∗∗∗ 0.00338∗∗∗ 0.00339∗∗∗ 0.00338∗∗∗ (3.74) (3.68) (3.69) (3.68)

Constant 0.0922∗∗∗ 0.0923∗∗∗ 0.0923∗∗∗ 0.0923∗∗∗ (151.52) (152.28) (152.28) (151.64) Observations 2452 2452 2452 2452 t statistics in parentheses ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

156 Chapter 3 Essays on Macroeconomics and Development

Notes on the Portuguese lending programme

Portugal received ﬁnancial assistance during 2011-2014. The programme was supported by loans from the European Union amounting to 52 bn and a 26 bn frin the Extended Fund Facility with the IMF.

European mechanisms’ support was equally divided to the EFSM and the EFSF each responsible for disbursing 26 bn . On table 12 all realised disbursements are presented. It should be noted that EFSM did not disburse the total amount as this was scheduled in the beginning of the programme but 1.7 bn less.

The outline of the programme was published in June 2011 under the title "The Eco- nomic Adjustment for Portugal" almost a month after Portuguese government’s request for ﬁnancial assistance.

Monitoring the implementation of the programme, was realised through 11 reviews of the progress. In each of those reviews - a statement was released which outlined the scheduled disbursements and their speciﬁc timeline.

Below we record what was outlined in each of those reviews regading future payments and compare them to the realised disbursements from table 12. No particular reference to which institution EFSF or EFSM was responsible for disbursing the tranches, however this can be inferred by taking into account the total amount to be disbursed and the timing of the sceduled payments.

1. 19 May 2011: EU and EFSF funding plans to provide ﬁnancial assistance for Portugal and Ireland

"Various borrowing operations by EFSM and EFSF will take place between 23 May and 15 July to cover ﬁrst disbursements to Portugal and Ireland for a total of 15.3 billions".

• In the aforementioned period 6 disbursements took place [(1)-(6)], Four of which from the EFSM and the remaining two from the EFSF. (1)-(4) amount for 6.5 bn (EFSM) and (5)-(6) from EFSF amounting for 5.9 bn.

2. 12 August 2011: Statement by the EC, ECB, and IMF on the First Review Mission to Portugal

"Approval of the conclusion of this review will allow the disbursement of €11.5 billion (€7.6 billion by the EU, and €3.9 billion by the IMF). This disbursement can take place in September".

• In the aforementioned period the only disbursements were done by the EFSF

Chapter 3 157 Essays on Macroeconomics and Development

and correspond exactly to the amount of money announced (7.6 bn) – these are the disbursements (7)-(9). The EFSF did not do any payments for this period.

3. 16 November 2011: Statement by the EC, ECB, and IMF on the Second Review Mission to Portugal

"Approval of the conclusion of this review will allow the disbursement of €8 billion (€5.3 billion by the EU, and €2.7 billion by the IMF). These disbursements could take place in December and January.".

• In the aforementioned period, EFSM made one payment (12) while EFSF made three payments (10)-(11) and (13). All those together amount for 5.2 billions (0.1 billion less than announced).

4. 28 February 2012: Statement by the EC, ECB, and IMF on the Third Review Mission to Portugal

"Approval of the conclusion of this review will allow the disbursement of €14.9 billion (€9.7 billion by the EU, and €5.2 billion by the IMF). These disbursements could take place in April subject to the approval of the IMF Executive Board and ECOFIN and EUROGROUP".

• In the aforementioned period, EFSM made 2 disbursements (14)-(15) amounting for 4.5 billions, and EFSF made another two (16)-(17) amounting for 5.2 billions. In total 9.7 billions as announced.

5. 04 June 2012: Statement by the EC, ECB, and IMF on the Fourth Review Mission to Portugal

"Approval of the conclusion of this review will allow the disbursement of €4.1 billion (€2.7 billion by the EU, and €1.4 billion by the IMF). These disbursements could take place in July".

• In the aforementioned period, only the EFSF made two disbursements, amounting for 2.6 billions (0.1 bn less than announced). (18)-(19) disbursements.

6. 11 September 2012: Statement by the EC, ECB, and IMF on the Fifth Review Mission to Portugal

"Approval of the conclusion of this review will allow the disbursement of €4.3 billion (€2.8 billion by the EU, and €1.5 billion by the IMF). These disbursements could take place in October subject to the approval of the IMF Executive Board and ECOFIN and EUROGROUP.".

158 Chapter 3 Essays on Macroeconomics and Development

• In the aforementioned period, two disbursements were made. One from the EFSM (20) of 2 billions, and one from EFSF (21) of 0.6 billions amounting in total of the announced amount.

7. 19 November 2012: Statement by the EC, ECB, and IMF on the Sixth Review Mission to Portugal

"Approval of the conclusion of this review will allow the disbursement of €2.5 billion (€1.6 billion by the EU, and €0.9 billion by the IMF). These disbursements could take place in January 2013".

• In the aforementioned period, January 2013 – no disbursement was made by either of EFSM, EFSF. The only disbursement close to this period was 0.6 bn by EFSF in February 2013 (22) in the table. (This is 0.8 bn less than what was promised).

8. 15 March 2013: Statement by the EC, ECB and IMF on the Seventh Review Mission to Portugal

"The conclusion of this review could take place in May, subject to the approval of ECOFIN and EUROGROUP and of the IMF Executive Board, and will allow the disbursement of €2.0 billion (€1.3 billion by the EU, and about €0.7 billion by the IMF)".

• In the aforementioned period only EFSF made two disbursements (23) and (24) amounting to 2.1 bn euros (Note here that this amount is more than the announced. HOWEVER, from previous reviews we have 0.1 bn not disbursed from what was promised in the 2nd Review, 0.1 bn not disbursed from the 4th review and 0.8 bn not disbursed from the 5th review. Adding those disbursements to the 1.3 promised bn sums up to the 2.1 bn disbursed in the aforementioned period).

9. 3 October 2013: Statement by the European Commission, ECB and IMF on the eighth and ninth review mission to Portugal

"The conclusion of the 8th and 9th reviews could take place in Novem- ber, subject to the approval of ECOFIN and Eurogroup and of the IMF Executive Board. This would allow for the disbursement of €5.6 billion (€3.7 billion by the EU, and about €1.9 billion by the IMF) following the approval of the current reviews."

• In the aforementioned period the only payment was done by EFSF and amounts to 3.7 bn (25)

Chapter 3 159 Essays on Macroeconomics and Development

10. 16 December 2013 Statement by the EC, ECB and IMF on the Tenth Review mission to Portugal

"Approval of the conclusions of this review will allow the disbursement of EUR 2.7 billion (EUR 1.8 billion by the EU and EUR 0.9 billion by the IMF), bringing the total amount disbursed to Portugal to EUR 74 billion representing roughly 94 percent of total available ﬁnancial assistance".

• Only EFSM made one disbursement (26) of exactly 1.8 bn as announced.

11. 28 February 2014 Statement by the EC, ECB and IMF on the Eleventh Review Mission to Portugal

"The conclusion of the 11th review could take place in April 2014, subject to the approval of ECOFIN and Eurogroup and of the IMF Executive Board. This would allow for the disbursement of €2.5 billion (€1.6 billion by the EU, and about €0.9 billion by the IMF) following the approval of the current review."

• In the aforementioned period, EFSF made a disbursement (27) amounting to 1.2 bn euros (0.4 bn less than promised).

12. 2 May 2014: Statement by the European Commission, ECB, and IMF on the Twelfth Review Mission to Portugal

"The conclusion of the 12th review could take place in June, subject to the approval of the ECOFIN Council and of the IMF Executive Board. This would allow for the disbursement of €2.6 billion (€1.7 billion by the EU, and about €0.9 billion by the IMF) following the approval of the current review."

• Note that at this point EFSM and EFSF owes to Portugal 0.4 bn euros from the 11th review plus another 1.7 bn euros promised by the 12th review. However, the last payment disbursed takes place on 12 November 2014 by the EFSM and amounts to 0.4 bn euros (probably what was owed by the 11th review. The 1.7 bn euros from the 12th review – were never disbursed)

• This can be conﬁrmed by the fact that in total the EFSM had agreed to disburse in total 26 bn euros throughout the programme, however, it only did disburse 24.3 bn (less than agreed by the missing 1.7bn disbursement)

160 Chapter 3 Essays on Macroeconomics and Development

Loans to EFSM EFSF Portugal Disbursement Disbursement Total Date Cumulative Cumulative (bn euros) (bn euros) Cumulative (01) 31/05/2011 1.75 1.75 0.00 1.75 (02) 01/06/2011 1.50 3.25 0.00 3.25 (03) 01/06/2011 2.25 5.50 0.00 5.50 (04) 01/06/2011 1.00 6.50 0.00 6.50 (05) 22/06/2011 6.50 3.70 3.70 10.20 (06) 29/06/2011 6.50 2.20 5.90 12.40 (07) 21/09/2011 5.00 11.50 5.90 17.40 (08) 29/09/2011 2.00 13.50 5.90 19.40 (09) 06/10/2011 0.60 14.10 5.90 20.00 (10) 20/12/2011 14.10 1.00 6.90 21.00 (11) 12/01/2012 14.10 1.70 8.60 22.70 (12) 16/01/2012 1.50 15.60 8.60 24.20 (13) 19/01/2012 15.60 1.00 9.60 25.20 (14) 24/04/2012 1.80 17.40 9.60 27.00 (15) 04/05/2012 2.70 20.10 9.60 29.70 (16) 30/05/2012 20.10 3.50 13.10 33.20 (17) 30/05/2012 20.10 1.70 14.80 34.90 (18) 17/07/2012 20.10 1.50 16.30 36.40 (19) 17/07/2012 20.10 1.10 17.40 37.5 (20) 30/10/2012 2.00 22.10 17.40 39.5 (21) 03/12/2012 22.10 0.80 18.20 40.30 (22) 07/02/2013 22.10 0.80 19.00 41.10 (23) 27/06/2013 22.10 1.05 20.05 42.15 (24) 27/06/2013 22.10 1.05 21.10 43.20 (25) 22/11/2013 22.10 3.70 24.80 46.90 (26) 25/03/2014 1.80 23.90 24.80 48.70 (27) 28/04/2014 23.90 1.20 26.00 49.90 (28) 12/11/2014 0.40 24.30 26.00 50.30

Table 3.17: Loans disbursed to Portugal under the lending programme, by EFSM and EFSF

Chapter 3 161 Essays on Macroeconomics and Development

Notes on the Greek lending programme

Greece was under a ﬁnancial assistance programme for almost a decade. Financial assistance provided by European mechanisms to Greece can be divided into three distinctive programmes plus an enhanced surveillance framework, which in total disbursed a history high of 283.7 bn . In particular, the ﬁrst programme, namely the Greek Loan Facility (GLF) began in May 2010 and ended in June 2013. It disbursed 80 billions 15 in total and consisted of billateral loans pooled by the European Commission. This programme was also part of a joint package, assisted by the IMF which disbursed in total 30 billion through a Stand By Arrangement (SBA).

The second programme in contrast to the ﬁrst one, provided ﬁnancial assistance to Greece through the EFSF, which was fully operational since August 2010. The second programme started in March 2012 and ended in June 2015. In total it disbursed 141.8 billion with loan repayments scheduled from 2023 to 2070 and a weighted average maturity of 42.45 years. It committed all unreleased payments remaining from the GLF (11.8 bn ) plus 130 billion .

The third programme, namely the ESM stability support programme lasted for three years, from August 2015 to August 2018. It disbursed a total amount of 61.9 bn with loan maturities ranging from 2034 to 2060 (32.5 years average loan maturity). This last programme was completely under ESM’s responsibility.

1st Programme for Greece - Greek Loan Facility

Loans to Greece GLF 1st Programme Disbursement Date Cumulative (bn euros) (1) 18 May 2010 14.50 14.50 (2) 13 September 2010 6.50 21.00 (3) 19 January 2011 6.50 27.50 (4) 16 March 2011 10.90 38.40 (5) 15 July 2011 8.70 47.10 (6) 14 December 2011 5.80 52.90

Table 3.18: Loans under the 1st programme (Greek Loan Facility - GLF)

15The amount was subsequently reduced by 2.7 billion since Slovakia refused to participate and Portugal and Ireland were expempted after entering a lending programme themselves.

162 Chapter 3 Essays on Macroeconomics and Development

Loans to Greece EFSF 2nd Programme Disbursement Date Sub-Tranches Cumulative (bn euros) 12 March/10April/25April 2012 29.7 29.7 12 March/10April/25April 2012 4.9 34.6 19 March 2012 5.9 40.5 (1) March/June 2012 10 April 2012 3.3 43.8 (74 bn) 19 April 2012 25.0 68.8 10 May 2012 4.2 73 28 June 2012 1.0 74 17/19 December 2013 34.3 108.3 (2) December 2012/May 2013 28/31 January 2013 9.2 117.5 (49.1 bn) 28 February 2013 2.8 120.3 3 May 2013 2.8 123.1 (3) May/June 2013 17 May 2013 4.2 127.3 (7.5 bn) 25 June 2013 3.3 130.6 (4) July/December 2013 31 July 2013 2.5 133.1 (3 bn) 18 December 2013 0.5 133.6 28 April 2014 6.3 139.9 (5) April/August 2014 9 July 2014 1.0 140.9 (8.3 bn) 14 August 2014 1.0 141.9

Table 3.19: Loans received by Greece under the 2nd programme (EFSF)

Chapter 3 163 Essays on Macroeconomics and Development

Loans to Greece ESM 3rd Programme Date Sub-tranches Disbursement Cumulative 20 August 2015 13.00 13.00 24 November 2015 2.00 15.00 (1) August/December 2015 01 December 2015 2.70 17.70 (18.7 bn) 08 December 2015 2.70 20.40 23 December 2015 1.00 21.40 (2) June/October 2016 21 June 2016 7.50 28.90 (10.3 bn) 26 October 2016 2.80 31.70 (3) July/October 2017 10 July 2017 7.70 39.40 (8.5 bn) 30 October 2017 0.80 40.20 (4) March/June 2018 28 March 2018 5.70 45.90 (6.7 bn) 15 June 2018 1.00 46.90 (5) March/June 2018 06 August 2018 15.00 61.90 (6.7 bn)

Table 3.20: Loans received by Greece under the 3rd programme (ESM)

164 Chapter 3