Populist Policies and Balance-of-Payment Crises in Emerging Economies
Alejandro Nakab∗ UC San Diego
March 2018
Abstract
Many emerging economies experience recurrent macroeconomic fluctuations where ex- pansions – characterized by increasing consumption and government spending but de- creasing foreign reserves – are followed by balance-of-payment crises. We develop a heterogeneous agent GE small open-economy model calibrated for a semi-industrialized economy where governments tend to avoid currency depreciations due to its distributive effects following a populist approach of macroeconomic management. In the model, the government transfers resources from capitalists to workers through fiscal policy and exchange market interventions, generating short-term expansions and redistribution – even in the presence of negative external shocks– but with balance-of-payment crises in the medium-term, which transfers resources back to capitalists and exporters. More- over, we show that the model’s generated pattern is consistent with 25% of all large depreciation episodes in developing countries in the period between 1950-2016.
∗Email: [email protected]. I am grateful to David Lagakos, Valerie Ramey and Natalia Ramondo for invaluable discussions and advice. I am also grateful to Marc Muendler, Tommaso Porzio and seminar participants at UCSD for helpful comments. 1 Introduction
Since the 1930s, many emerging economies have experienced macroeconomic fluctuations with some common features. Particularly, there were expansions with decreasing foreign reserves associated with populist policies where governments tended to avoid currency de- preciations, which ultimately led to a balance of payment (BOP) crisis and a recession (Braun and Joy(1983), Sachs (1990) Dornbusch and Edwards(1990) and Edwards (2003)). After some absent years, there has been a resurgence of populist politicians in many devel- oping economies as is particularly the case in some Latin American countries (Acemoglu, Egorov and Sonin 2013). As Dornbusch and Edwards (1990) defined: “Macroeconomic populism is an approach to economics that emphasizes growth and income distribution and deemphasizes the risks of inflation and deficit finance, external constraints and the reaction of economic agents to aggressive non-market policies.” In this paper, we propose a model to formalize these ideas discussed in this older literature in a modern heterogeneous agents GE model. We study the impact of these populist policies and analyze their distributive effects, focusing on the interaction between re-distributive policies based on exchange market interventions and the external constraint (the economy’s foreign currency availability). Fixed exchange rate regimes required constant central bank interventions to prevent the exchange rate from moving. Although in the last decades many countries tended to “offi- cially” migrate to floating regimes, in practice, central banks systematically intervened in exchange markets (Calvo and Reinhart 2002). In fact, record accumulation of reserves post 2002 has much to do with countries’ willingness to stabilize exchange rates in a world with greater capital mobility (Ilzetzki, Reinhart and Rogoff 2017). In most cases, interventions try to limit appreciations rather than depreciations, following a mercantilist view of de- fending a devalued exchange rate to protect domestic industries (Levy-Yeyati, Sturzenegger and Gluzmann 2013). Nonetheless, the constant loss of foreign reserves before these crises suggest the opposite, meaning exchange market interventions preventing real exchange rate depreciations. These populist cycles in which expansions were followed by BOP crises were vastly doc- umented for South American countries but it was not documented for other countries.1 Therefore, by conducting a case study analysis and using updated data, we first document these patterns for large devaluation episodes for South America in more recent years and for many others African and Asian countries as well. Using yearly data from the World Bank In-
1Among others Braun and Joy(1983), Dornbusch and Edwards(1990) and Edwards (2003) study the macroeconomic fluctuations in Latin American countries.
1 dicators for every country between 1960 and 2015, we first define a BOP crises as a situation where the real exchange rate depreciates more than 30% within a year. We document that in 25% of the 145 large devaluations episodes, countries suffer a decrease in international reserves for at least 2 years (3 years for most cases) before the BOP crises. Then, we restrict the analysis to those devaluation episodes and find that previous years to the BOP crises are associated with (1) large and increasing government spending and consumption which drop sharply at and after the crises and/or (2) negative term of trade shocks several years before the crises. We interpret these dynamics as populist policies, in which policy makers tend to avoid currency devaluations by intervening in the exchange market boosting consumption at the expense of international reserves. We start by constructing a GE small open-economy model with a representative agent (RA) based on Burstein, Neves and Rebelo (2003) and Burstein, Eichenbaum and Rebelo (2007), but extend it by adding foreign reserves and analyzing its behavior when specific fiscal and monetary policies are undertaken. Following the observation that the described pattern is mostly observed in middle-income countries that are net exporters of primary goods and low-skilled intensive goods and net importers of capital goods, we calibrate the model to a semi-industrialized economy with those characteristics. Then, we analyze the features the model must have to generate growth with decreasing foreign reserves. The impulse response functions show that if there is a positive shock to productivity or terms of trade, the economy will experience an increase in output and foreign reserves, generating a positive correlation between these two. We find that we need one of two things to reproduce output growth with decreasing foreign reserves in this model. The first option is a households’ preference shock coming from a decrease in the household discount factor or a decrease in the willingness of households to hold money, which does not seem to be a plausible story behind these recurrent macroeconomic fluctuations. The second is to have an increase in money supply jointly with exchange market interventions to avoid a real exchange rate depreciation. Nonetheless, although the model matches the sign of the correlations from the data, the impulse response functions from these shocks and policies cannot match the timing of the BOP crises. In the model, we find a first expansion and a slow convergence back to the steady state which do not correspond to the expansion followed by a sharp exchange rate depreciation. Given that the RA model fails to match the timing of the BOP crises, we then proceed to include heterogeneous agents. We model an economy with two types of agents: workers with no access to bond markets but who can hold money and capitalists who have access to foreign bond market and who can hold real balances. With this heterogeneous agents model we aim to understand the economic dynamics coming from populist policies, analyzing its income
2 distribution implications at each part of the cycle. We model populist policies as (1) transfers from capitalist to workers driven by increases in taxes and lump-sum transfers between groups and as (2) direct transfers to workers financed by a central bank that issues money to cover a fiscal deficit, both with and without exchange market interventions.2 We find that populist policies provoke changes in mean propensities of consumption and money holding in the aggregate. Particularly, the policies generate a decrease in aggregate savings due to agents having different borrowing constraints and different preferences over real balances.3 Therefore, we provide a deeper explanation for preference changes in the aggregate caused by government policies that redistribute resources. Moreover, when the government intervenes in the exchange market, preventing the exchange rate to devalue while running a fiscal deficit, the model more accurately matches the data pattern, showing one large devaluation episode and recession once the exchange market intervention ends. We then redirect our attention towards populist reaction to external shocks. We show the impulse response function to a negative terms of trade shock followed by different exchange market intervention policies. We show that with no intervention, the terms of trade shock generates a decrease in production with a sharp devaluation and a real wage decrease. Fur- thermore, we find that the more aggressive the intervention in the exchange market is, the lower the real depreciation and the fall in GDP are. In the extreme case, the economy could even experience an expansion with a small real depreciation at the shock if the government spends all its reserves in a short time span, but with a recession once the government stops the intervention. Populist governments can postpone BOP crises to protect workers, even in the presence of negative terms of trade shocks, but for only a short period of time. Finally, as we analyze the income distribution dynamics, we find that we can represent class conflict with a particular measure of real wage (nominal wage over nominal exchange rate, which co-moves with the real exchange rate), which will be a sufficient statistic to know how the real income of each class behaves for each policy chosen. We conclude that populist policies tend to increase the real income of workers and decrease the real income of capitalists and that devaluations increase the real income of capitalist and decrease the real wage of workers. Thus, that cycle alters the wealth and income distributions, increasing the labor share in the presence of populist expansions and decreasing it in BOP crises.
2We sustain the idea that an increase in the size of the public sector, which is one of the most common choices of populist governments, tends to create the same dynamic that an increase in direct transfers to workers financed by taxes or an increase in money supply to cover the fiscal deficit. 3The existence of different consumption bundles of agents across the income distribution was documented a long time ago by Engel (1857, 1895) and also recently by Almas (2012) using household surveys for many countries. To calibrate the model we use an Argentinian survey of household’s consumption and we see that the consumption bundle are in deed different for each decile of the income distribution.
3 Related literature. Due to evidence on the resurgence of macroeconomic populism, this paper studies the role of external restrictions as redistributive policies are implemented. This paper relates to old literature on populist policies, macroeconomic fluctuations, fi- nancial instability and contractionary devaluations (Diaz Alejandro (1963), Krugman and Taylor (1978), Braun and Joy(1983), Dornbusch and Edwards(1990) and Edwards (2003) among others). These studies show the dynamics of populist cycles in terms of income re- distribution dynamics in expansions and devaluations (mainly empirically) but they do not provide general equilibrium models where all the macroeconomic variables of interest can be jointly analyzed. This paper contributes first by documenting some features of the so-called populist cycle (focusing on exchange rate policies) in emerging economies around the world. More importantly, the main contribution of this paper is to build a modern HA-GE small open-economy model to formally study these policies. In doing so, our contribution relies theoretically on including the role and dynamics of foreign reserves as different policies and exchange market interventions are implemented and studying the distributional effects of these policies. In its focus on exchange market interventions and exchange rate dynamics, this paper relates to the literature on the effects of devaluations on relative prices and the real exchange rate (e.g, Burstein, Eichebaum and Rebelo 2005, 2007). Within this strand of the litera- ture, our paper is closely related to the studies analyzing the distributional effects of large devaluation episodes. For instance, Cravino and Levchenko (2017) show that the impact of devaluations on the cost of living is higher at the bottom compared to top of the income distribution. We focus on workers, firm owners, and exporters and analyze the income dis- tribution dynamics between these groups. Furthermore, our paper relates to the strand of the literature studying exchange market interventions. Levy-Yeyati, Sturzenegger and Gluz- mann (2013) show that in fact many of these interventions tend to act to avoid appreciations although populist policies tend to act in the opposite way. There are no papers that develop the idea of fear of depreciation with the exception of Dutta and Leon (2002) who empirically show that governments tend to avoid nominal exchange rate depreciations, consistently with our notion on populist policies. In its focus on international reserves, our paper is related to the strand of the literature that studies its optimal level and historical dynamics (e.g, Rodrik (2006), Alfaro and Kanczuk (2009), Olivier and Ranciere (2011), Ilzetzki, Reinhart and Rogoff (2017)) although the focus of these studies contrast sharply with ours. These studies focus on interventions to stabilize exchange rates and macroeconomic volatility while we focus on interventions and policies that redistributes resources among groups to benefit workers at the expense of firm owners and exporters. Finally, an important model input is the pattern of specialization and the general struc-
4 ture of the economy, as we think it is a key aspect that actively generates very specific dynamics in the economy. Few prior papers analyze the role and effects of the pattern of specialization when different fiscal and monetary policies are undertaken. Two exceptions are Gavin and Perotti (1997) who show how certain fiscal policies tend to be more disruptive in Latin American economies compared to other industrialized economies and Ilzetzki, Men- doza and Vegh (2013) who show the responses of fiscal shocks in different economies focusing on the differences on exchange rate policy, openness, and external debt levels. Nevertheless, they do not focus on other features as production structure and input-output linkages, which are a key aspect of this paper. The paper is organized as follows: In section 2, we show some regularities for the identified populist BOP crises in the data. In section 3, we construct a representative agent model and analyze what the model must have to match the data correlations and how it fails in successfully describing the BOP crises timing. In section 4, we build an HA-GE model and show how we can match the populist-driven BOP crises described in the presence of expansionary monetary policies and exchange market interventions. In section 5, we show the scenario with negative terms of trade shocks and different exchange market interventions. In section 6, we provide alternative policies on exchange rate regimes; and finally, in section 7, we conclude.
2 Empirical results
We are interested in the link between populist policies and external constraints and, therefore, we focus the analysis on exchange market interventions and policies that end in a BOP crisis. For this purpose, we use yearly data from the World Bank Indicators for all countries between 1960 to 2015. We first define a BOP crises as situations where the real and nominal exchange rates depreciates more than 30% in a year.4 While BOP crises may have different causes, we only focus on those driven by populist exchange market interventions which imply governments must be constantly intervening selling foreign reserves before the large devaluation. We select all BOP crises in which international reserves decreased for at least two consecutive previous years. We find large devaluation episodes all around the world where BOP crises took place after years of falling foreign reserves and we see that this pattern is particularly strong in South America. From 1960 to 2015 there were 145 large depreciation episodes, from which 34 are episodes where countries had falling foreign reserves for more than 2 years previous to the year of the episode (which represents around 25% of
4We restricted the sample to 50%, 30% and 10% and chose 30% to have enough observations but to have depreciation episodes that were also large enough to generate a sizable shock.
5 all devaluation episodes).5 Our restricted sample consists in all emerging economies.6 Note, that macroeconomic dynamics in a developed economies are not usually similar to that of emerging economies; thus, certain policies such as currency depreciations could affect them differently. Moreover, populist policies of income redistribution, as described by Dornbusch and Edwards (1990), are not usually observed in high-income countries which also have floating exchange rate regimes. We show next that virtually all episodes we keep after restricting the sample are devaluations preceded (1) by terms of trade shocks several years before the crisis in which governments try to avoid devaluations intervening in the exchange market as much as their foreign reserves allows them, or (2) by increases in government expenditure, consumption, and real exchange rate appreciations at the expense of a loss in international reserves. Both dynamics are inside our definition of populist policies. Note that even after a term of trade shock which would be exogenous, the choice of intervening in the exchange market for such long periods of time may not be optimal. Figure 2 shows the dynamics on the mean nominal exchange rate, real exchange rate, GDP, foreign reserves, exports and imports, and the dynamics on the components of GDP (C, I, G and CA) from all the large devaluations selected.7 The first feature to point out is the clear contractionary effect from devaluations. As the literature vastly documented currency devaluations tend to be recessionary in developing countries. These sub-figures ∗ show there is a constant real exchange rate appreciation, defined as the change in etPt /Pt, before the large devaluation. At the same time, evidently due to the episodes selection, there is a large decrease in foreign reserves in the year of the large devaluation and in most cases a permanent decrease in the previous 3 years as well. In our definition of populist policies, we focus on exchange market interventions in which governments tend to avoid real depreciations, but note that these interventions have limits as at some point central banks cannot keep on intervening and the economy crashes into an external constraint. We see that after the depreciation international reserves not only stop falling but start increasing, which means the central bank is buying reserves in exchange of domestic currency.
5In the appendix we graph 2 different sets of large devaluations. First, we graph all South American devaluations without any restrictions as they all experience international reserves decreases for more than 2 years (3 years in average) previous to the BOP crisis except for the case of Argentina in 1989, which shows a different pattern. Second, we graph all the rest that includes African, Asian, and a few Central American countries with the restriction of having at least 2 years of falling reserves. Finally, note that our restricted sample does not include the Asian currency crises of the late 1990s which also shows that this currency crisis was not driven by governments intervening in the exchange market, but for other reasons, as sharp and sudden stops in capital inflows. 6From the restricted sample we observe there are no developed economies except Iceland in 1968. 7In the appendix we show every devaluation for South America and the rest of the world with a separate line as well as the means for these two groups.
6 Figure 1: Large Devaluation Episodes
(a) Log Nominal Ex Rate (b) rer change (c) Log GDP
(d) Log Int. Reserves (e) Log Exports (f) Log Imports
(g) Log Gov Expenditure (h) (Gov Exp)/Cons (i) (Gov Exp)/GDP
(j) Consumption/GDP (k) Investment/GDP (l) CA/GDP The graphs show years with a real depreciation of more than 30% jointly with a nominal depreciation of more than 30%. The lines represent the means of all the variables for large devaluations with decreasing reserves in the 2 previous years of the episode.
7 Moreover, in the external front, the dynamics of exports and imports are very clear (there is small variance around the mean) and what we expect to find in any small open economy model. There is a large imports decrease when the devaluation takes place as the recession and depreciation will provoke a lower demand of final goods and a lower demand for imported inputs, as capital goods. After the episode, imports tend to increase once again (as the economy starts growing, demand for imported final goods and foreign inputs increases). In the case of exports, they tend to decrease before the devaluation and they increase after the episode. Furthermore, there is a constant increase in government expenditure up to the year of the devaluation where it decreases substantially. Government expenditure in terms of GDP increases before the devaluation and drops in the same year and after the large devaluation. Consumption (as share of GDP) tends to be higher before than after the devaluation. These general tendencies are consistent with an increasing government spending and a country keeping high levels of consumption that may not be sustainable as the external front tends to deteriorate. Given these patterns, we will focus on exchange market interventions jointly with different government spending policies.
3 Representative Agent Model
We first develop a GE small open-economy model in the spirit of Burstein, Eichenbaum and Rebelo (2007) and Burstein, Neves and Rebelo (2003), and we extend it by adding foreign reserves and analyzing its behavior when specific fiscal and monetary policies are undertaken. The economy is modeled as a partially-industrialized small open economy with comparative advantage in agricultural goods and low-skilled intensive goods. There are 2 sectors in the model: (1) a tradable sector (agricultural goods), and (2) a sector that produces a good that is not internationally traded. The agricultural sector uses labor and land as inputs, and the non traded sector uses an imported input (i.e, high-quality machines) and labor.
Representative agent (RA). For the first specification of the model there is a repre- sentative agent in this economy that consumes both final goods and holds money. There is mass one of households [0,1] and the expected utility and budget constraint are defined as:
( ∞ 1+φ ) X Mt L E βt α log(T ) + α log(N ) + α log − t T t N t M P 1 + φ t=0 t
T N ∗ ∗ ∗ m Pt Tt + Pt Nt + Mt + etRt−1Dt−1 + Rt−1Dt−1 = Ωt + Mt−1 + etDt + Dt + ∆t
8 where Tt is the consumption of the tradable, Nt is the consumption of the non-traded, m Mt is money holding, Ωt is income, ∆t is transfers from the government and Lt is labor. ∗ Moreover, Dt−1 is the stock of domestic bonds, Dt−1 is the stock of international bonds, ∗ Rt is the international interest rate, Rt is the domestic interest rate and et is the nominal 8 N T exchange rate defined as local currency per 1 unit of foreign currency. Finally, Pt ,Pt and
Pt are the prices of the non-tradable, the tradable and the consumer price index respectively. Moreover, there are an infinite amount of varieties of the non-traded goods that agents demand. Therefore, Nt is the aggregate amount of the internationally non-traded goods and we use a Dixit-Stiglitz aggregator defined as: σ σ − 1 Z 1 σ − 1 L Nt = yit σ di 0
L where yit represents each one of the varieties in the market and σ is the elasticity of substi- tution between varieties.9
Locally produced good, manufacturing sector. The manufactured goods, being an emerging economy for reasons of low quality or high transportation cost, are not traded in the international markets. There are an infinite number of producers producing a unique variety with a CRS production function of the form:
N N N N γL 1−γL yit = At(Lit ) (Iit)
N where Lit is labor and Iit are imported goods used in firm i and period t, and At is productiv- ity. The intuition for this production function is that the economy needs high quality capital goods that are not produced domestically, and therefore, the economy needs to import in m∗ order to grow. Moreover, the prices of the inputs are wt for labor and Pt et for the imported good (assuming the law of one price holds), thus, the price of a good in each country will be equal to the price in international markets multiplied by each country exchange rate. In the aggregate we have a CRS production function given by:
N N N N γL 1−γL Yt = At(Lt ) (It)
8 In the case of emerging economies et is the number of units of domestic currency needed to buy 1 USD. 9Note that we use monopolistic competition instead of a simpler perfect competitive market to have positive benefits for the firms and an easy closed form of the price of the non-traded which will be important to analyze the redistributive effects later on.
9 and that the price index given by:
N N m∗ (1−γL ) γL N σ 1 Pt et wt Pt = N N σ − 1 At 1 − γL γL σ where is a constant markup and the rest of the expression is the marginal cost of the σ − 1 good that in this case, as the production function is CRS, is the price in perfect competition with a weighted combination of the input prices.
Tradable sector, agricultural goods. We are modeling an emerging economy that ex- ports commodities and agricultural goods. Thus, two inputs are used (labor and land) to produce the tradable good and the production function is:
T T T T γL T 1−γL Yt = At(Lt ) (At Lat)
T where Lt is labor in sector T in period t and Lat is land that, as it is fixed, is normalized T to 1. In addition, At is country specific productivity and At is land productivity in sector T ∗ T. As the law of one price holds, the domestic price of the tradable good is: Pt = Pt et.
Income Sources. Households’ have three sources of income. First, there are profits de- KT rived from the tradable sector, denoted by Ωt :
KT T T ∗ Ωt = (1 − γL )Yt Pt et
Note that the production function in this sector is CRS in labor and land, therefore the owners of the land will have profits from the production of the tradable good and that is T going to be exactly the share (1 − γL ) of output. Second, there are profits derived from the KN non traded sector, denoted by Ωt :
N N ∗ (1−γL ) γL KN 1 1 Pt et wt N Ωt = N N Yt σ − 1 At 1 − γL γL
σ Firms charge a constant markup σ−1 over the marginal cost (the price index from the CRS N cobb douglas production function), and Yt is the amount of units sold. Third, there is W income derived from wages, denoted by Ωt :
W T N Ωt = Lt wt + Lt wt
Thus, households’ total income, denoted by Ωt, is the sum of the three sources of income.
10 External sector and aggregation. In this model, domestic agents face an interest rate that is increasing in the country’s net foreign debt:10
∗ ∗ W Dt Rt = Rt exp − d GDPt
From this condition is clear that a higher debt to GDP ratio will generate an increase in the interest rate. Moreover, in this economy the market clearing condition for the non-tradable good implies that domestic consumption must be equal to total production:
N Yt = Nt
On the tradable sector, there are agricultural goods which are exported and non-traded inputs that are imported, and thus, the trade balance (in foreign currency) is defined by:
∗ T m∗ TBt = Pt (Yt − Tt) − Pt It
T where TBt is trade balance, Yt and Tt are production and domestic consumption of the tradable good produced locally, respectively, and It is imported goods used as inputs in the N ∗ m∗ production of the local good Yt . Finally, Pt and Pt are the prices in foreign currency of the agricultural good and the imported good. Moreover, nominal GDP is:
N N N m∗ T T GDPt = Yt Pt − Pt etIt + Pt Yt
Nominal GDP is the total value of production in the tradable sector plus total value of production in the non traded sector net of imported goods used as inputs. The CPI is then:
N N T T Yt Pt N Yt Pt T N T Pt = N N T T Pt + N N T T Pt = γPt + (1 − γ)Pt (Yt Pt + Yt Pt ) (Yt Pt + Yt Pt )
The index is defined as a weighted average of the prices in the tradable and non-tradable sector that are the goods that agents consume. Note that due to the assumption on func- tional form of the utility function, the shares of consumption of each good relative to total consumption will be constant (γ denotes the share of N that households spend in terms of total consumption goods). Finally, we define GDP in foreign currency and real GDP as:
N N e GDPt R GDPt GDPt = and GDPt = et Pt 10Following the approach of previous papers as Senhadji (1994), Mendoza and Uribe (2000), Schmitt- Grohe and Uribe (2001), and Schmitt-Grohe and Uribe (2003)
11 where GDP in foreign currency is nominal GDP in domestic currency, divided by the ex- change rate and real GDP is nominal GDP divided by the CPI.
Resource Constraint and Balance of Payment. Note that summing all the budget constraints of agents and using the market clearing conditions, the consolidated budget constraint is defined by:
m ∗ ∗ ∗ ∆Mt − ∆t = TBtet + Dt − Rt−1Dt−1 et where the left-hand side represents the excess demand of money (∆Mt is the change in m money demand and ∆t is the change in money supply) and the right-hand side represents net inflow of foreign currency to the economy in domestic currency. A positive trade balance means exports are greater than imports and, therefore, foreign currency is flowing into the economy and a positive second term means the new debt (foreign currency flowing into the economy) is greater than what is being paid of interest and capital for past debt (foreign currency flowing out of the economy). Thus, the behavior of the external sector variables reflects on movements in the monetary base. Therefore, the fundamental BOP identity is: