The NEMESIS Reference Manual

Coordination Erasme Team,France Prof. P. Zagamé B. Boitier,A. Fougeyrollas,P. Le Mouël Core Teams National Technical University of Athens, Greece Prof. P. Capros,N. Kouvaritakis Federal Planning Bureau, Belgium F. Bossier,F. Thierry, A. Melon The NEMESIS model had been partially funded by the research programs of the European Commission

2 Contents

Introduction to NEMESIS 8

I. The core economic Model 15 I.1. Current version of the endogenous technical change module ...... 15 I.1.1. The stock of knowledge ...... 16 I.1.2. From stock of knowledge to innovation ...... 17 I.1.3. innovation to economic performance ...... 18 I.1.4. Calibration in the model ...... 21 I.2. New production functions with embodied endogenous technical change and skills ...... 21 I.2.1. Multi-level, putty-semi-putty CES production functions ...... 22 I.2.2. The Endogeneization of Technical Change ...... 26 I.2.3. Estimation results ...... 33 I.3. Households’ ﬁnal consumption ...... 47 I.3.1. Aggregate consumption ...... 47 I.3.2. Allocation of aggregate Consumption ...... 49 I.4. External trade ...... 54 I.4.1. Intra-European trade ...... 54 I.4.2. Extra European Trade ...... 57 I.4.3. Imports and Exports prices ...... 59 I.5. Wage setting ...... 61 I.5.1. Theory ...... 62

3 Contents

I.5.2. Model ...... 63 I.5.3. Data ...... 64 I.5.4. Results ...... 68 I.6. Labour supply ...... 75 I.6.1. The data on participation rates of working-age population . . . . . 76 I.6.2. Determinants of participation rates ...... 78 I.6.3. Calibration of labour supply ...... 83 I.7. Taxation and subsidies ...... 87 I.7.1. Institutional sectors accounts ...... 87 I.7.2. Public ﬁnances ...... 87 I.7.3. Focus on most important taxations system ...... 89 I.8. Sectoral Interdependencies ...... 92 I.8.1. Demand ﬂows to products ...... 92 I.8.2. technological progress interactions...... 95 I.9. housing investments ...... 98 I.9.1. Methodology ...... 99 I.9.2. The data ...... 102 I.9.3. Model estimate and results ...... 103 I.9.4. Sensibility analysis ...... 106 I.9.5. Concluding Remarks ...... 109

4 List of Figures

.1. Basic functioning of the model ...... 10 .2. NEMESIS modularity ...... 11

I.1. From R&D expenditure to the R&D stock ...... 17 I.2. The stock of knowledge ...... 17 I.3. Two types of innovation ...... 18 I.4. Process innovation and economic performance ...... 18 I.5. Product innovation and economic performance ...... 19 I.6. CES nesting structure ...... 23 I.7. Ex-ante and ex-post isoquants ...... 25 I.8. EU National share of high skill on total employment, in 2005 (source: Eurostat) ...... 36 I.9. European high skill share in total employment for NEMESIS sectors, (source EU-KLEMS) ...... 38 I.10. European high skill share in total employment for NEMESIS sectors . . . 39 I.11. Sectoral illustration of the ﬁnal results ...... 40 I.12. Ratio of European employee unit cost between high and low skills at sectoral level ...... 41 I.13. Corrected European share of compensation of employees for high skill at sectoral level in 2005 ...... 43 I.14. Allocation of Durable Goods ...... 50 I.15. Allocation of Non Durable Goods ...... 50

5 List of Figures

I.16...... 70 I.17. results whole model ...... 71 I.18. Sectoral results P1 ...... 73 I.19. sectoral results P2 ...... 74 I.20. Participation rates to labour market of men and women aged 25 to 64, EU27 + Norway, 2005 ...... 76 I.21. Participation rates to labour market of women aged 50 and 64 by skill, EU27 + Norway, 2005 ...... 77 I.22. Social Contribution paid ...... 91 I.23. Social contribution received ...... 92 I.24. Sectoral interdependencies in NEMESIS ...... 94 I.25. Knowledge spillovers ...... 97 I.26. Rent Spillovers ...... 98 I.27. Sensibility analysis with common adjustment coeﬃcient ...... 107 I.28. Model response to 1% shock on households real disposable income: Com- parison according to adjustment coeﬃcients ...... 108

6 List of Tables

I.1. Labour compensation growth, period 1998-2005 ...... 65 I.2. Unemployment rate , period 1998-2005 ...... 66 I.3. Price growth and high skill share , period 1998-2005 ...... 67 I.4. Labour productivty growth, period 1998-2005 ...... 68 I.5. Coefficients summary ...... 69 I.6. Estimation results of participation rates ...... 82 I.7. Elasticities of activity rates in NEMESIS in 2008 ...... 85 I.8. Estimates results of households gross fixed capital formation error correction model ...... 105 I.9. Estimates results for short term model with individualised adjustment coefficients ...... 106

7 Introduction to NEMESIS

The NEMESIS model (New Econometric Model of Evaluation by Sectoral Interdepen- dency and Supply), has been partialy funded under the fifth and sixth RTD Framework Programs of European Commission General Directorate of Research1. It is a system of economic models for every European country (EU27 less Bulgaria and Cyprus, plus Norway), USA and Japan, devoted to study issues that link economic development, competitiveness, employment and public accounts to economic policies, and notably all structural policies that involve long term effects: RTD, environment and energy regulation, general fiscal reform, etc. The essential purpose of the model is to provide a framework for making forecasts, or ‘Business As Usual’ (BAU) scenarios, up to 25 to 30 years, and to assess for the implementation of all extra policies not already involved in the BAU. NEMESIS uses as main data source EUROSTAT, and specific databases for external trade (OECD, New CRONOS), technology (OECD and EPO) and land use (CORINE 2000). NEMESIS is recursive dynamic, with annual steps, and includes more than 160.000 equations. The main mechanisms of the model are based on the behaviour of representative agents: Enterprises, Households, Government and rest of the world. These mechanisms are based on econometrics works.

1The core teams of the NEMESIS model are : • ERASME (France) as coordinator • CCIP (France) • Federal Planning Bureau (Belgium) • National Technical University of Athens (Greece)

8 Introduction to NEMESIS

The main originality of the model, when compared to others used for similar policies, lies in the belief that the medium and long term of macroeconomics path is the result of strong interdependencies between sectoral activities that are very heterogeneous from a dynamic point of view, with leading activities grounded on Research and Development, and from environment and sustainable development with a huge concentration of pollutants on few activities. These interdependencies are exchanges of goods and services on markets but also of external effects, as positive technological spillovers and negative environmental externalities. Another originality of NEMESIS is that it is a “Framework model” with different possibilities on the several mechanisms involved in the functioning (see figure .2 for the different available modules). Although econometrics, the model cannot be classified as a neo-keynesian model, in the new version that built-in the new theories of growth; it escapes also to the classification of general equilibrium model, as it incorporates original mechanisms that do not refer to the strict orthodoxy of the mainstream neo-classical approach, on which was based the general equilibrium approach. We now present the main mechanisms, outputs and uses of NEMESIS.

Main NEMESIS’ mechanisms

On the supply side, NEMESIS distinguishes 32 production sectors, including Agri- culture, Forestry, Fisheries, Transportations (4), Energy (6), Intermediate Goods (5) Capital goods (5), Final Consumption Goods (3), Private (5) and Public Services. Each sector is modeled with a representative firm that takes its production decisions given its expectations on production capacity expansion and input prices. Firms behaviour includes very innovative features grounded on new growth theories, principally endoge- neous R&D decisions that allow firms improving their process productivity and product quality. Production in sectors is in this way represented with CES production functions (with the exception of Agriculture which uses Translog functions, and Forestry and Fish- eries where technology is represented with Leontief functions) with 5 production factors : capital, unskilled labour, skilled labour, energy and intermediate consumption, where also endogenous innovations of firms come modify the efficiency of the different inputs (biased technical change) and the quality of output (Hicks neutral technical change). The production function was estimated by the dual approach and estimation and calibration of links between R&D expenditures, innovations and economic performance were picked up from the abundant literature on the subject. The pricing of enterprises results from

9 Figure .1.: Basic functioning of the model

10 Introduction to NEMESIS

Figure .2.: NEMESIS modularity

an arbitrage between firms engaged in competitive behaviour and those with a pricing by mark-up (due to innovation that creates monopoly situations). Interdependencies between sectors and countries are finally caught up by a collection of convert matrices describing the exchanges of intermediary goods, of capital goods and of knowledge in terms of technological spillovers, and the description of substitutions between consumption goods by a very detailed consumption module enhance these interdependencies. On the demand side, representative households’ aggregate consumption is dependent on current income. Total earnings are a function of regional disposable income, a measure of wealth for the households, interest rates and inflation. Variables covering child and old-age dependency rates are also included in an attempt to capture any change in consumption patterns caused by an ageing population. The unemployment rate is used, in the short-term equation (only), as a proxy for the degree of uncertainty in the economy. Consistent with the other behavioural equations, the disaggregated consumption module is based on the assumption that there exists a long-run equilibrium but rigidities are present which prevent immediate adjustment to that long-term solution. Altogether, the total households aggregated consumption is indirectly affected by 27 different consumption sub-functions through their impact on relative prices and total income, to which demographic changes are added. Government public final consump-

11 tion and its repartition between Education, Health, Defence and Other Expenditures, are also influenced by demographic changes. For external trade, it is treated in NEMESIS as if it takes place through two channels: intra-EU, and trade with the rest of the world. Data availability was an important factor in this choice – it allowed an emphasis to be put on intra-EU trade flows, which are a large portion of the total trade in the EU. The intra- and extra-EU export equations can be separated into two components, income and prices. The income effect is captured by a variable representing economic activity in the rest of the EU for intra-EU trade, and a variable representing economic activity in the rest of the world for extra-EU trade. Prices are split into two sources of impacts in each of the two equations (intra- and extra-EU trade). For intra-EU trade, they are the price of exports for the exporting country and the price of exports in other EU countries. For extra-EU trade, prices impacts come through the price of exports for the exporting country, and a rest-of-the-world price variable. The stock of innovations in a country (which, in NEMESIS, is taken relative to the total innovation stock in Europe in a particular sector) is also included in the export equations in order to capture the role of innovations in trade performance and structural competitiveness. For imports, equations are identical for both intra- and extra-EU trade. The income effect is captured through domestic sales by domestic producers, while the price effects are represented in both the import price, as well as the price of domestic sales by domestic producers. The stock of innovations is again included to account for the effects of innovations on trade performance. The wage equations, which determine in NEMESIS the dynamics of prices and incomes, are based on a theory of wage-setting decisions made by utility maximising unions. The unions calculate utility from higher levels of employment and from higher real wages (relative to wages outside the sector) in the sector, subject to the labour- demand constraint imposed by firms’ profit-maximisation. The implication of this form of wage equation is that conditions in the labour market are important for determining wage and real wages in a given sector will rise if there are positive productivity shocks, changes in the unemployment rate, or changes in the real wage outside that sector. Another important NEMESIS characteristic is finally its land-use module that extended the field of policies the model could explore to the areas of Agriculture, Forestry, Bio-energies, Tourism, Transportations, Urbanization and Nature Conservation, through their implications on land-use. These six sectors are actually of significant importance for land use; they are at the origin of all possible land claims which modelling in NEME- SIS is cross sectoral to this extent that all sectors are competing for land. Land claim by each sector sum up in a common land balance for all sectors which, confrontation to

12 Introduction to NEMESIS the supply land function allows deriving the equilibrium rental price for land.

Main NEMESIS’ inputs and outputs

On the input side, NEMESIS uses for its functioning assumptions on a set of exogenous variables concerning word assumptions including interest rates, exchange rates, activity proxies for the rest of the world, prices of wholesales commodities and specially oil; demographic assumptions by country such total Population, population and participation rates to labour force by gender per 5 years cohorts; national policies assumptions and notably fiscal policies (indirect and direct taxes, social security benefits and contributions) and government expenditures (defence, health, education, infrastructures, others expenditures) and investments; and energy and environment assumptions as excises duties and other energy tax rates, CO2 taxation, etc. On the output side, NEMESIS can deliver results at EU25, country and regional NUTS2 levels for key economic indicators. The indicators the model calculates are macro-economic, as GDP (European, National or Regional) and its counterparts (final consumption, investment, exports, imports, etc.), sectoral, as production, value added and employment per NACE economic sector or sector clusters, or agent based (Gov- ernement, Non Financial Corporations, Financial Corporations, Households including NPIH, and outside). Beyond economic indicators as GDP, prices and competitiveness, employment and revenues, financial balances for the main agents, etc., NEMESIS Energy Environment Module (NEEM) gives detailed results on energy supply and demand by fuel type and technology, and on various pollutants emissions: CO2, SO2, NOX, HFC, PFC and CF6; it computes also a carbon price (Taxation or tradable permit price associated to a carbon constraint). The inclusion in the model of detailed data on population and working force, allows also the model delivering many social indicators as employment, unemployment and labour force participation rates by gender, GINI coefficient for wages and earnings, and a set of indicators dedicated to measure inequalities between European countries and regions for key variables as GDP and final consumption per capita. Additional original indicators concern land use by 6 sectors: Agriculture, Forestry, Nature Conservation, Urbanization, Transport and Energy Infrastructures and Tourism, and 8 land categories. NEMESIS calculates also, at country level, the equilibrium rental price of land, which impact strongly on housholds’ cost of living and firms’ investment price.

13 Main NEMESIS uses

With its original characteristics and great detail level results, NEMESIS can be used for many purposes as short and medium-term economic and industrial “forecasts” for business, government and local authorities; analysing Business As Usual (BAU) scenarios and economy long-term structural change, energy supply and demand, environment, land-use and more generally sustainable development; revealing the long term chal- lenges of Europe and identifying issues of central importance for all European, national, regional scale structural policies; assessing for most of Lisbon agenda related policies and especially knowledge (RTD and human capital) policies; emphasizing the RTD aspect of structural policies that allows new assessments (founded on endogenous technical change) for policies, and new policy design based on knowledge: Education, Skill and Human Capital and RTD. NEMESIS has notably been used to study BAU scenarios for European Union and reveal the implication for European growth, competitiveness and sustainable development of the Barcelona 3% GDP RTD objective, of the 7th Research Framework Program of European Commission, of National RTD Action Plans of European countries, of Euro- pean Kyoto and post-Kyoto policies, of increase in oil price, of European action plan for Renewable Energies, of European Nuclear Phasing in/out, etc. NEMESIS is currently used to assess for European Action plan for Environmental and energy technologies, for European ﬁnancial perspective (CAP reform) and for Lisbon agenda, with in deep development on the modelling of RTD, Human Capital and labour market and European regions.

14 CHAPTER I

The core economic Model

We will present in this section two versions of the supply side and of the endogenous technical change module: the ﬁrst one, the current version implemented in the model, had been used in numerous studies related to R&D and innovations, while the second one is currently tested and enhance the previous formulation.

Section I.1 Current version of the endogenous technical change module

The endogenisation of technical progress in applied models is a very recent phenomenon. It has mainly been used in overall balance models. Some of these models follow on from the work carried out by Arrow [17]. Here, the rate of technical progress is linked to expertise or experience, measured by gross accumulated investment. They therefore take up a similar viewpoint to the AK model in which the capital K variable contains information relating to the state of the technology. This approach has been adopted in

15 I.1. CURRENT VERSION OF THE ENDOGENOUS TECHNICAL CHANGE MODULE certain models relating to climate change, such as those of Goulder and Mathai [156] and Grubb [163]. In this field, the characteristics of technologies are often linked to experience curves. Making technical progress endogenous provides a more effective, immediate implementation of policies to fight against greenhouse gases as a result of the experience acquired in the field. Other overall balance models use R&D expenditure to make technical progress endogenous. This is the case in models dealing with issues relating to international trade (such as those of Diao and Roe [101], Baldwin and Forslid [24] and Diao et alii [102]) and in models applied to the environment and climate change (such as Nordhaus’ RDICE model [254] or Fougeyrollas et alii’s GEM-E3 model [147]). Endogenisation through R&D expenditure is not easy. The first difficulty comes in calculating the relationship between R&D and process or product innovations. The second comes from the possibility that there may be a number of balances. The third stems from the diversity of R&D levels of intensity and results in the sectors. Only a few sectors, such as those linked to ICT and the pharmaceutical sector, are R&D intensive. It is therefore necessary to adopt a detailed sector-based approach so that the endogenisation of technical progress is appropriate. Few models manage to overcome this difficulty. Econometric models containing technical progress mechanisms endogenised by R&D are rare. To our knowledge, only the International Monetary Fund Multimod model, which is highly agregated, includes R&D stocks at a sector level. The Nemesis model takes it place in this new family of macro-econometric models with endogenous technical progress. The special feature in Nemesis is the endogenisation of technical progress across three phases: from R&D to the stock of knowledge, from the stock of knowledge to innovation and from innovation to economic performance, the second feature is the level of disagregation.

I.1.1 The stock of knowledge

The variable that plays a vital role in the endogenisation of technical progress in Neme- sis is the variable “knowledge” (KNOW ) that arises out of the R&D stock. A sector’s R&D stock is determined by its R&D expenditure and by a constant displacement rate. It is constituted as a stock of capital, with displacement being the gradual deletion of knowledge (ﬁgure I.1). “Knowledge” is not determined only by the sector’s R&D stock but also by all the knowledge spillovers in all national and foreign sectors (ﬁgure I.2). Knowledge spillovers

16 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.1.: From R&D expenditure to the R&D stock from other sectors are dependent on their stocks of R&D, via technological flow matrices. These matrices, which are differentiated by sector and by country, are constructed according to the methodology developed by Johnson for the OECD (Johnson, 2002). This consists of identifying, for every patent registered at the European Office, the sectors producing and using the innovation described in the patent. This is then used to determine the proportion in which the knowledge accumulated in a sector will benefit others, by calculating knowledge transfer coefficients, the knowledge being, by assumption, borne by the patents. This work is done in great detail (over 100 sectors) and the results are re-agglomerated in Nemesis’ sector-based nomenclature in the form of technological flow matrices. “Knowledge” also feeds on the R&D stock in foreign sectors and on the public sector R&D stock.

KNOW

R&D Stock of the Sector

Technology Flow Matrices

R&D Stock of Other R&D Stocks of Foreign Public R&D Stock Sectors Sectors

Figure I.2.: The stock of knowledge

I.1.2 From stock of knowledge to innovation

Innovations are determined by the variant in the stock of knowledge (ﬁgure I.3). The two types of innovation are considered here:

• process innovations that increase the global productivity of factors in the speciﬁ- cation that we have chosen;

• product innovations, which, in the ﬁxed nomenclature of national accounting that

17 I.1. CURRENT VERSION OF THE ENDOGENOUS TECHNICAL CHANGE MODULE

under-pins Nemesis, are considered as quality improvements1.

These two types of innovation act very diﬀerently on economic performance:

∆KNOW

Process Innovation Product Innovation

Figure I.3.: Two types of innovation

I.1.3 innovation to economic performance

Process innovation does not lead to the same eﬀects as product innovation. Process innovation increases the global productivity of factors, thus increasing product supply and reducing the unit production cost, and therefore the price. This price reduction leads to increased demand, which is dependent on demand price elasticity (ﬁgure I.4).

Productivity Growth Increase in Supply

Process Innovation Price Fall Supply Side

Demand Side Demand Price Elasticity ε

Increase in Demand

Figure I.4.: Process innovation and economic performance

Growth in demand helps to absorb extra supply (at a constant usage level) if demand price elasticity is higher than or equal to one. However, econometric estimates in chrono-

1National accounts already take, partially, the increasing quality of goods and services into account in their calculus, however this accounting is relatively rough. In the NEMESIS model we consider the quality improvements that are additional what are very important whenever we increase R&D eﬀorts.

18 CHAPTER I. THE CORE ECONOMIC MODEL logical series reveal an elasticity generally lower than one for each sector, and thus for the whole economy. This result comes from the assumption of a representative firm per sector: we do not consider the innovative firm in competition with the other companies in its activity sector. This amounts to assuming that all firms in the sector innovate and reduce their prices. Increased demand then depends on the capacity for absorption represented by elasticity lower than one. In this case process innovation reduces the use of factors as the effects of supply outweigh the effects of demand.

Product innovation acts like an increase in efficiency per volume unit and increases demand for units of efficiency (figure I.5). Volume production is only maintained if the increase in demand for the new efficiency is just equal to the increase in efficiency due to innovation. Generally, product innovation does more than compensate for the fall in factor usage due to process innovation. R&D therefore leads simultaneously to an increase in GDP and in the use of factors.

Increase in efficiency per Fall in price of efficiency unit volume

Product Innovation Supply Side

Demand Side Increase in demand of variation of demand volume efficiency unit

Figure I.5.: Product innovation and economic performance

The ex ante eﬀects of innovation on GDP depend on the eﬀects of the increase in knowledge on the global productivity of factors and on quality and thus on demand: increased production is in fact linked to increases in demand arising from process innovation and quality innovation respectively (box 1).

19 I.1. CURRENT VERSION OF THE ENDOGENOUS TECHNICAL CHANGE MODULE

Box 1. The eﬀects of innovation on economic performance Process innovation: the accumulation of knowledge (KNOW ) generates an increase in the global productivity of factors (TFP ).

∆TFP ∆KNOW = α TFP KNOW Product innovation: the accumulation of knowledge (KNOW ) leads to an improvement in quality (QUAL).

∆QUAL ∆KNOW = α0 QUAL KNOW Economic performance: increased production (Y ) depends on increased demand due to innovation depending of two elasticities and 0.

∆Y ∆TFP ∆QUAL = + 0 Y TFP QUAL i.e.

∆Y ∆KNOW ∆KNOW = α + 0α0 = β Y KNOW KNOW Finally, economic performance, measured by increased production due to increased knowledge, is written as follows:

∆Y ∆KNOW = β Y KNOW Most of the available econometric studies link increased production with an increase in R&D stock (SRD) using the following formula :

∆Y ∆SRD = α Y SRD The diﬀerence between these two approaches is an explicit integration of all the spillovers in the ﬁrst and an implicit or nil integration in the second. Econometric studies (Mohnen [245], Mairesse and Sassenou [237], Grilliches [160], Nadiri [249], Cameron (1998)[49], . . . ) reveal a fairly broad range for parameter β of 0.05 to 0.2. The results are independent of the methods chosen. However, where β is estimated using instant cross-section series (inter-companies), it is higher than when chronological estimates are used.

I.1.4 Calibration in the model

20 CHAPTER I. THE CORE ECONOMIC MODEL

In the NEMESIS model the α and α0 had been calibrated to reproduce the desired values of the mean β parameter. This had been done by making sensitivity analysis over the historical data in order to reproduce past trends, in order to be as close as possible of the historical facts. After that, the βparameters are diﬀerentiated using sectoral R&D intensities.

Section I.2 New production functions with embodied endogenous technical change and skills

Since their conception in the late Fifties and early Sixties (see for example: Jorgenson [203], Salter [280], Solow [299]; Solow, Tobin et al. [300]), vintage models have been often adopted by applied modellers for representing the links existing between technical change and economic growth. These models gave important insights regarding the com- plementarities between productivity growth on one hand and investment on the other one, through the technical progress embodied in the new vintages. The development of new growth theories from the Eighties stated furthermore that technical change was itself an endogenous process based on R&D and innovations decisions of private and public actors. We therefore adopted for NEMESIS an embodiment approach, close from the one already implemented in GEM-E3 model, and where technical change results from investment decisions for new equipment goods and machineries on the one hand, and from investments in R&D based innovations that modify both the rate and the direction of technical change. The modelling of production technologies was inspired by the approach that was developed by Adriaan Van Zon [324] and Huub Meijers & Adri- aan Van Zon (1999) that they called RUM Putty-Semi-Putty vintage model. ’RUM’ means ’Recursive Update Model’, that allows to obtain the aggregate levels of production factor demands from a set of simple recursive update rules. This model is based on a putty-semi-putty vintage production structure, and it is precisely the possibility of limited substitution possibilities ex-post which enable to reduce to a set of few equations, the book-keeping account of all existing vintage, necessary when economic scrapping is endogenised, as for the Putty-Clay and the Clay-Clay vintage production models as they were ﬁrst introduced by Johansen [196] and Salter [280]. RUM makes positive use of the fact that it is often not necessary to know all the details of every individual vintage: from

21 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS a macro-economic point of view, only the average characteristics of the capital stock are important. In a ﬁrst subsection, we show how this vintage approach was introduced in NEMESIS, with the use of “Putty-Semi-Putty” multi-level CES production functions. A second sub- section, we describes then the modelling of technical change, endogenized on R&D, and incorporated in production vintages. The resolution of ﬁrms’ optimization program, and of the main equations that where introduced in NEMESIS for R&D, innovation and production factor demands, are available in NEMESIS reference manual, the presentation here focusing on methodological issues.

I.2.1 Multi-level, putty-semi-putty CES produc-

tion functions

The nested CES framework

The multi-level nested CES production functions, pioneered by Sato [283], have recently been widely used in macroeconomics. Its ﬂexibility, and its usefulness to implement and analyse endogenous growth makes it an attractive choice for many applications in economic theory, applied modelling and empirics (cf. Acemoglu (2002), Papageorgiou and Saam (2006) and McAdam et al. (2007)). In NEMESIS, apart for the Power sector, which has a special modelling, the other 29 production sectors were modelled with four- level nested CES production functions that diﬀer only from the values of substitution elasticities, and of share parameters.

NEMESIS production functions

NEMESIS production functions were extended to include low skilled and high skilled labour, and 5 productive inputs : Capital K, Low Skilled LabourLL, High Skilled Labour

LH , Energy E and Materials M,. The choice of factor bundles was based on the results of separability tests, abundant in the econometric literature. As it is illustrated by I.6 Source du renvoi introuvable., at a ﬁrst stage Materials are combined with a bundle regrouping all other production factors. At a second stage, Low Skilled Labour was separated from Capital, High Skilled

22 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.6.: CES nesting structure Y

M KELHSLLS

LLS KELHS

LHS KE

K E

Labour and Energy, that we supposed to be gross complements. High Skilled labour is separated from the bundle formed by Capital and Energy at a third level, and then Capital was combined with Energy at a fourth level. This grouping means that at each level, the value of the partial substitution elasticities between the factor that is separated, and each production factors in the bundle formed by other inputs, are identical. Partial substitution elasticities are noted:

• σ1for substitutions between M and K,LH LL,E;

• σ2for substitutions between LH and K,LL E ;

• σ3 for substitutions between LL and KE ;

• and σ4 for substitutions betweenK and E.

The next sub-section will details the expression of the nested production functions.

Production technology: A Putty-Semi Putty Vintage Model

For technologies of production, the underlying idea is that substitutions possibilities between production factors are greater ex-ante than ex-post, that is to say, they are greater at the moment of the investment in the new vintage than when the marginal production capacity was already installed. We have then to distinguish the ex-ante from the ex-post production functions.

23 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

The Ex-Ante Production Function

To begin with the production function for the ﬁrst level of the nesting, that combines materials and the bundle of all other inputs to produce the output in volume QY , lets consider the following production function with constant returns to scale:

a a 1 h −ρ −ρ i− ρa a 1 a 1 1 QYt,t+1 = δMt · Mt,t+1 + δKLEt · KLEt,t+1 (I.1)

Where:

• QYt,t+1 is the output in volume,

• Mt,t+1 is the amount of materials associated to vintage t at date t + 1;

• KLEt,t+1is the amount of the bundle formed by Capital, Energy and High skilled Labour associated to Low skilled labour on vintage t at date t + 1;

• δa and δa are the ex-ante distribution parameters; Mt KLEt

a • σ1 is the ex-ante partial elasticity of substitution between M and KLE with a 1 σ1 = a . (1+ρ1 )

By assumption, the capital associated to the new vintage, installed at date t, needs one year to be productive: date t + 1.

The Ex-Post Production Function

The ex-post production function has the same CES, constant returns to scale, speci- ﬁcation that the ex-ante function (I.4):

1 p − p −ρ ρ Q = δp · M 1 + δp · KLE 1 (I.2) Yt,t+1 Mt t,t+1 KLEt,t+1 t,t+1

where δp , δp and σp are the ex-post parameters of the CES function, and with Mt KLEt 1 p 1 σ1 = p . By assumption, ex-post substitution possibilities between KLE and M are (1+ρ1) p a limited and we have σ1 < σ1 .

Ex-ante and ex-post substitution possibilities

24 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.7.: Ex-ante and ex-post isoquants

To illustrate the diﬀerence between ex-ante and ex-post substitution possibilities one can express the ex-ante and ex-post production functions in term of factor coeﬃcients, respectively:

a a 1 h −ρ −ρ i− ρa a 1 a 1 1 1 = δMt · mt,t+1 + δKLEt · vt,t+1 (I.3) and

− 1 −ρp −ρp ρa 1 = δp · m 1 + δp · v 1 1 (I.4) Mt t,t+1 KLEt t,t+1

with :

• v = KLE , the coefficient for the factors inside the bundle and QY • m = M the factor coefficient for Materials. QY Figure I.7 shows that ex-post isoquants (e.p), associated with certain ex-ante technologies on the curve (e.a), have a stronger curvature than ex-ante isoquants, reflecting that the substitutions possibilities between the two categories of factors are reduced ex-post. By definition, at the date of installation of the last vintage, the ex-ante and ex- post production functions are equal and there is only the technique (m, ν), on the ex- ante isoquant, in common with the ex-post isoquants. This technique, defined as the tangential technique, allows determining the exact position of the ex-post isoquant on figure I.7.

25 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

One can also express the ex-post parameters in terms of ex-ante parameters and of the tangential technique (m, ν), by identifying equations I.3 and I.4 above :

p a δp = δa [m ]ρ −ρ (I.5) Mt Mt t

and

p a δp = δa [ν ]ρ −ρ (I.6) KLEt KLEt t

Equations I.3 and I.4, that characterize the ex-ante and ex-post production technologies can also be re-expressed in terms of the tangential technique (m, ν) and of the ex-ante parameters :

− 1 h a −ρa a −ρa i ρa gt (mt, νt) = δMt mt + δKLEt νt = 1 (I.7)

with gt (mt, νt) the ex-ante production function, in terms of factor coeﬃcients, associated with the vintage t; and :

1 p a p a − h a (ρ −ρ ) −ρp a (ρ −ρ ) −ρp i ρa ft,t+1 (mt,t+1, νt,t+1, mt, νt) = δMt mt mt,t+1 + δKLEt νt νt,t+1 = 1 (I.8)

with ft,t+1 (mt,t+1, νt,t+1, mt, νt) the ex-post production function, in terms of factor coeﬃcients, associated to the vintage t at instant t + 1. These characteristics of the ex-ante and ex-post technologies are also valid for the three other levels of the production function, that exhibit also constant returns to scale.

I.2.2 The Endogeneization of Technical Change

We show in this section how, in each production sector, the ﬁrms can increase the quality of their products, and the productivity of their inputs, by investing in R&D activities and by buying certain amounts of innovations. The underlying idea is, as for the optimal choice of production factors, that substitutions possibilities are greater ex-ante than ex-post. The representative ﬁm invests ex-ante in R&D activities to improve the quality of its products, and the productivity of its inputs, on the marginal production capacity. Ex-post, once the marginal production capacity installed, there are

26 CHAPTER I. THE CORE ECONOMIC MODEL no possibilities for modifying the quality of products or the productivity of production factors. By assumption firms run in-house R&D, with constant return to scale innovation technologies. They beneficiate of positive knowledge spillovers from R&D activities in other production sectors, but also from other countries and from public laboratories. They have also negative knowledge spillovers from their past innovations (fishing-out effect), as in Jones (1995). These knowledge externalities, by modifying the productivity of R&D, give the possibility of increasing returns to scale and endogenous growth at the industry level, even if each representative firm operates with constant returns to scale technologies. We have therefore six different sources of endogenous technical change in NEMESIS. One is Hicks-neutral, with the improvement of products quality, and the other are biased with the endogenous improvement of individual factors productivity (Capital, , High Skilled Labour, Low Skilled Labour, Energy and Materials). This general setting allows furthermore taking into account the possible crowding-in or crowding-out effects between the different innovation activities. We describe first how the quality of products is combined with the volume produced to form the output Y , and, similarly, how the input-specific innovations are combined to the volumes of inputs used to form the efficient inputs. We then present the innovation functions of firms and the formation of knowledge spillovers.

The incorporation of innovations in output

In every production sectors, product innovations are incorporated in the new production vintage. The characteristics of products, measured by the ‘marginal quality index’ , are chosen ex-ante, and once again, we must make the distinction between the ex-ante and the ex-post production technologies.

The ex-ante trade-oﬀ between improving products quality and increasing volume of output

In NEMESIS, the marginal production capacity, Y , is expressed in eﬃcient units, and it is a CES function of the production volume, QY , and of product innovations, IY :

1 a a − a a −ρ0 a −ρ0 ρ Yt,t+1 = δ Q + δ I 0 (I.9) QYt Yt,t+1 IYt Yt

where δa and δa are the distribution parameters of respectively product innovations QYt IYt

27 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

a 1 and output in volume, and σ0 = a . (1+ρ0 ) By convention, IY is set to 1 for the base year of NEMESIS (2000), where the levels of production in eﬃcient and in ordinary units (volume) are equal:Yt,t+1 = QYt,t+1 . Furthermore, product innovations are considered as ﬁxed inputs that become productive one year after their date of invention.

Products characteristics are ﬁxed ex-post and embodied in new vintages

By assumption, the technological characteristics of products are ﬁxed ex-post, and products innovations are embodied in the new production vintages from ex-ante optimal production and innovation choices, that will be described later. The marginal production capacity, measured in eﬃcient units, is given ex-post by the following linear production function:

Y = ap Q (I.10) t,t+1 Yt Yt,t+1 p with at the ex-post marginal quality index of output.

The correspondence between ex-ante and ex-post decisions for producing the eﬃcient marginal output

It is possible, similarly to marginal output in volume, QY , to re-express equations I.9 and I.10 above, in terms of factor coeﬃcients, respectively:

a a 1 h −ρ −ρ i− ρa a 0 a 0 0 1 = δ qyt,t+1 + δ iyt (I.11) QYt IYt

and

1 = ap .q (I.12) Yt yt,t+1

QY IY with qy = y and iy = y successively the factor coeﬃcients for marginal output in volume and for products marginal quality index. We can then, from I.11 and I.12 above, re-express ap in terms of the tangentiel Yt ¯ technique (q¯yt , iyt )and of the ex-ante parameters:

− 1 " −ρa # ρa q¯ 0 0 p ¯ a a yt a (¯qyt , iyt ) = δQ + δI . (I.13) Yt Yt Yt ¯ iyt

28 CHAPTER I. THE CORE ECONOMIC MODEL

which is ex-post a ﬁx parameter.

The incorporation of innovations in productive inputs

In NEMESIS, inputs used on the new production vintage are measured in physical volumes, X, and in eﬃcient units QX , with X = K,LL,LH ,E and M respectively the Capital, Low skilled labour, High Skilled Labour, Energy and Materials.

For a given marginal output in volume, QY , we have the following system of four levels nested CES functions of productive input used:

− 1 i i ρi i −ρ1 i −ρ1 1 QYt,t+1 = δMt · Mt,t+1 + δKELt · KLEt,,t+1 (I.14)

− 1 i i ρi i −ρ2 i −ρ2 2 KLEt,t+1 = δ · L + δ · KLH E (I.15) LLt Lt,t+1 KLH Et t,,t+1

− 1 −ρi −ρi ρi KL E = δi · L 3 + δi · KE 3 3 (I.16) H t,t+1 LHS,t Ht,t+1 KEt t,t+1

− 1 i i ρi i −ρ4 i −ρ4 4 KEt,t+1 = δEt · Et,t+1 + δKt · Kt (I.17)

with i = a, p for , respectively , ex-ante and ex-post production technologies. The capital used on the new vintage, Kt, is a ﬁxed factor that become productive after one year (intallation delay).

We then have ex-ante, for ,X = K,LL ,LH ,M,E:

− 1 h a −ρx a −ρx i ρx Xt,t+1 = δ Q + δ I . (I.18) QXt Xt,t+1 IXt Xt The ex-ante efficient units of inputs are CES combinations of the volume of the factor used and of factor efficiency indexes, IX , with IX = 1 for the base year of NEMESIS (2000). The factor efficiency indexes act as fixed factors, with an installation delay of one year, as for physical capital. By assumption, ex-post, efficient units of inputs are linear functions of the volume of factors used:

p Xt,t+1 = axt QXt,t+1 . (I.19)

29 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

Finally, the ex-post productivity parameters of inputs, ap , can, , similarly than for the Xt ex-post marginal quality index of output, be expressed in terms of the ex-ante parameters ¯ and of the tangential techniques (q¯xt , ixt ):

− 1 p ¯ a −ρx a ¯−ρx ρx a (¯qxt , ixt ) = δ q + δ i (I.20) xt QXt xt IXt xt

QX IX with qx = X and ix = X respectively the factor coeﬃcients for volume of inputs and factor speciﬁc innovations used on the new production vintage. p The marginal productivity of inputs in volume, axt , that depends only of the choice of the tangential technique ex-ante (date t), is consequently constant ex-post.

The innovation functions

The innovation indexes for output and inputs on last production vintage, Ij,t, are modelled is NEMESIS as innovations stocks:

Ij,t = Ij,t−1 + innovj,t (I.21)

with j = Y,K,LL ,LH ,E and M, and where innovjt are the new innovations produced at date t.

The ﬂow of innovations innovj,t, is produced with the following constant returns innovation function:

innovj,t = αj,t · RDj,t (I.22)

where RDj,t and αj,tare respectively the R&D expenditure at constant prices of the representative ﬁrm for the innovation type j, and the R&D productivity. The originality of this formulation is that research productivity in one sector and one category, j, of innovation, αj,t, is inﬂuenced by two externalities, as in Jones (1995):

KNOWj,t αj,t = αj (I.23) NEj,t

with αja constant and positive parameter, KNOWj,t the knowledge stock of the sector for innovation type j, and NEj,tthe ‘Research Diﬃculty’ index that is a positive function of all past successful innovations realized by the sector (Jones (1995), ’Fishing-out’ eﬀect):

30 CHAPTER I. THE CORE ECONOMIC MODEL

βj NEj,t = (Ij,t−1) with βj a positive parameter.

The knowledge externality, KNOWj,t, reduces R&D costs by innovation, and reﬂects the fact that if innovations are speciﬁc to sectors who produce it, the technological knowledge is to a large extend common to all sectors and all countries.

This modelling of research productivity, αj,t, is particularly important in NEMESIS for the reason that if the knowledge externality, KNOWk,j,t grows faster than the Research

Difficulty index, NEj,t, research productivity will increase in time. We will then have increasing returns to scale, at the global level, while representative firms are supposed to operate with constant return to scale production and innovation functions, compatible with pure and perfect competition on all product markets. In this case, every policies stimulating R&D expenditures will have long term positive impacts on the growth rate of the economy, while these impacts will stay limited in time in knowledge externalities do not grow fast enough to compensate the rising difficulty in time of innovating.

The modelling of knowledge spillovers

In NEMESIS knowledge externalities result from past R&D expenditures with the following generic accumulation for R&D:

SRDt = (1 − δ) · SRDt−1 + RDt−τ (I.24)

with SRDt the R&D stock at date t, δ the ’radioactive’ rate of decay and τ > 0 measuring the delay for R&D expenditures to transform into formal knowledge that will influence the productivity of R&D, αt. R&D expenditures are realized by private firms and by public universities and research centers with a repartition, in 2008, of respectively 60% and 40% for private and public R&D in EU-27. These two sources of research externalities have distinct impacts on economic performance of European firms, private R&D being more oriented toward industrial applications of inventions, and public R&D toward basic research. Econometric studies reveal to that extent a greater contribution to economic growth of basic research compared to applied research, but also greater maturation delays. In NEMESIS, τ was set to one year for private R&D expenditures, which implies that, if one takes also into account the one year delay for innovations to be introduced in production, that knowledge

31 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS externalities from private origin will influence economic performance with an average lag of two years. From knowledge externalities coming from the public sector, τ was set to 3 and it needs four years for public research to influence economic performance. Knowledge spillovers from private R&D expenditures are measured in NEMESIS with Johnson (2002) OECD Technology Concordance (OTC) that transforms patent applications data, based on the International Patent Classification (IPC), into patent counts by sector of the economy. OTC is a matrix that is used for dispatching the R&D performed by the industrial sectors (’Industries of Manufacture’) in the sectors that will the most likely use the process or product innovations that they realize (’Sectors of Use’). From this methodology, knowledge externalities flow from industrial sectors to industrial sectors themselves and toward service sectors, but there are no externalities from service sectors toward industrial sectors. This is a limitation due to IPC definitions that do not include innovations in softwares and in services. OTC matrices were calculated for NEMESIS at country level, from European Patent Office (EPO) database. Knowledge spillovers from public research are sent to sectors with a ’grandfathering’ approach consisting in a split of public knowledge stock between sectors proportionnal to their share in total private R&D expenditure. This approach was retained in NEMESIS for the reason that there do not exist precise information in existing databases on how the public R&D contribute to sectoral innovation performance. There exist data in EUROSTAT on the repartition of public R&D by socio-economic objectives, but this repartion doesn’t help much for retrieving the amount of spillovers that flow to economic sectors. Our assumption of grandfathering follows then the idea that these spillovers at sectoral as important that the specific sectors are engaged in R&D activities; also, we did supposed in NEMESIS that research For knowledge externalities from foreign sources, NEMESIS uses trade flows of goods and services. The assumption is that knowledge transfers between countries are bare by traded goods, that is to say by the imports realized by the country that receives the externality. For illustration, the intra-sectoral knowledge spillover for a country i and a sector s (source 1) is measured in the following way:

s PRODi,s,t X IMP i,c,s,t SRD = .SRDi,s,j,t + .SRDc,s,j,t i,s,j,t PROD + IMP PRODi,s,t+IMPi,s,t i,s,t i,s,t c6=i ,

where PRODi,s,t is the production of good s in country i, IMPi,s,tis the total imports of good s by country i, IMP i,c,s,tis the import of good s in country i from country c,

32 CHAPTER I. THE CORE ECONOMIC MODEL

SRDi,s,j,t−τj in the R&D stock of sector s in country i and SRDc,s,j,t−τj the R&D stock in the foreign country c. The R&D stock have the following generic equation:

SRDs,j,t = (1 − δs) · SRDs,j,t−1 + RDs,j,t−τj . (I.25)

For one sector s in a country c: In each country c, the knowledge in a sector s accumulates accordingly to the following formula:

I −I PN PF KNOWc,s,j,t = SRDc,s,j,t + SRDc,s,j,t + SRDc,s,j,t + SRDc,s,j,t

where:

I 1. SRDc,s,j,t represents the past R&D eﬀorts realized in the production sector, by national and foreign ﬁms. It is the intra-sectoral knowledge spillover;

−I 2. SRDc,s,j,t represents the past R&D eﬀorts realized in the other production sectors. It is the inter-sectoral knowledge spillover;

PN 3. SRDc,s,j,t represents R&D externalities coming from the public laboratories in the country, that beneﬁciate to the sector;

PF 4. SRDc,s,j,t represents ﬁnally R&D externalities emanating from public laboratories in foreign countries, that beneﬁciate to the sector.

Johnson D. (2002), The OECD Technology Concordance (OTC): Patents by Industry of Manufacture and Sector of Use, OECD working paper n° 2002/5.

I.2.3 Estimation results

This section presents the estimation of the new production block for the NEMESIS model. We estimated the aggregated ex-post production function for output and inpurs measured in efficiency units. We start this section by presenting the econometrical specifications of the production block and the construction of data we used. In the third part we present the econometric results. The fourth implements shock price simulations in order to test the robustness of the estimations and to obtain the direct elasticities of substitution between factors of different bundle. Finally, we discuss issues concerning the ex-ante estimates.

33 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

Estimation speciﬁcations for the eﬃcient form

In this part we estimate the production block only for the specification of input and output in efficient units. The determination of the whole block with biased technical change will be realized using calibration methods. We estimate the five factor demand equations with the FIML method (Full information maximum likehood). FIML is the asymptotically efficient estimator for linear and nonlinear simultaneous models, under the assumption that the disturbances are multivariate normal. Bundle prices and production prices used in the regression are the unit costs of production. For instance, for the last level, we consider:

1 h 1−σ4 σ4 1−σ4 σ4 i 1−σ4 PKE = PK δK + PE δE

We deﬁne the distribution parameter with the share value of the input, for example, if we consider the materials:

PM · M δM = PM · M + PKELLS LHS · KELLSLHS To avoid problems of endogeneity, we use lagged values for input and prices in the distribution parameter δ. From a technical standpoint, we add a scale parameter which takes into account independent technical progress, named A (in a non-biased form). Thus, we use an t exogenous form, which takes a deterministic linear trend, i.e. A = A0i,ce where A0i,c the scale parameter and et is the technology growth rate. Because factors do not adjust immediately we need to take into account adjustment delays. Thus, we transform the factor demand equation into the following:

" ! !# X PY ,s · δM,s log (Ms,t) = ρM,s log Y s,t − log dcA0s,c − αs · t + σ1,s log c PM,s + (1 − ρM,s) · log (Ms,t−1)

Where s is the sector index, c the country index and ρ is the time adjustment param- 1−ρ eter. Thus, ρ is the time necessary for an adjustment at a 50% level. Finally, we consider a sector for all countries, with the underlying assumption that the elasticities of substitution are common to all countries, but not between sectors. In order to measure certain country speciﬁcities, a dummy variable is introduced next to

34 CHAPTER I. THE CORE ECONOMIC MODEL the scale parameter, which therefore represents a ﬁxed eﬀect for each sector and country.

Low and high skilled labour demand data

In order to estimate the production block we have to carry out the construction of data on labour demand by skill. Speciﬁcally we construct two types of data: one related to the share of employees or total employment and the other one on the labour remuner- ations. In addition to being used for the production block estimation, the data will be incorporated into the NEMESIS model. As already mentioned, we deﬁne two categories of job skills:

• High skilled labour corresponding to the INSEAD5 to INSEAD6 classes

• Low skilled labour that corresponds to the INSEAD1 to INSEAD4 classes

The sectoral dimension of the NEMESIS model implies the availability of information on employment by skills at the sectoral level, and there are few complete data sources on this subject. According to our knowledge, only two databases provide such dataset:

• The Eurostat database which provides skilled labour data by age, gender, occupa- tional jobs, etc . . . but only at the national level

• The EU-KLEMS database in which we can ﬁnd the share of labour employment and the share of labour compensation by skills at a sectoral level.

Thus, the EU-KLEMS database seems to be well suited, regarding its sectoral dimension, to be used as a basis for the NEMESIS model. However the skills definitions are different between each country, leading to a consistency problem that could imply surprising results in the model. The Eurostat database is homogenous but provides data at the national level only. As a consequence, we choose to use the sectoral information provided by EU-KLEMS and to apply a national correction using the Eurostat database. We present in the following sections how these corrections were made to obtain the share of low and high skilled labour on total employment and employees and the share of low and high skilled labour on total labour compensation at the sectoral level, and we illustrate our corrections with statistical figures.

Employment and employees data

35 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

Figure I.8.: EU National share of high skill on total employment, in 2005 (source: Eurostat)

40%

35%

30%

25%

20%

15%

10%

0% BE EE FI NO DK ES IE UK NL LT LU SE FR DE EU GR LV SI PL HU AT SK MT IT CZ PT RO

National shares (Eurostat)

eurostat Starting from the Eurostat database, we calculate the share of high skilled (SX,HS,C ) eurostat and low skilled (SX,LS,C ) labour demand at national levels as follows:

0 eurostat HSc SX,HS,C = 0 0 0 (I.26) HSc + MSc + LSc

eurostat eurostat SX,LS,C = 1 − SX,HS,C (I.27)

Where c is the country index, X = EMP, SAL the total employment or the total employees and HS0 , MS0 and LS0 are respectively the INSEAD1-2, INSEAD3-4 and INSEAD5-6 in Eurostat. The Figure I.8 shows the share of high skilled labour in total employment according to the Eurostat database for each EU countries. We ﬁnd the share to be superior to 35% in Belgium, Estonia and Finland with respectively 36.8%, 35.9% and 30.1% whereas the lowest shares are inferior to 15% with 14.7% in Italy, 14.6% in Czech Republic, 13.4% in Portugal and 12.6% in Romania. In average, the European share of high skilled jobs in the total employment is about 25% in 2005.

Sectoral levels (EU-KLEMS)

36 CHAPTER I. THE CORE ECONOMIC MODEL

EUKLEMS EUKLEMS We also calculate the total high skilled (XHS,C,S ) and low skilled (XLS,C,S ) labour and the total high skilled and low skilled number employees, at the sectoral level using EU-KLEMS database.

EUKLEMS EUKLEMS NEMESIS XHS,C,S = SHS,C,S · XC,S (I.28)

EUKLEMS EUKLEMS NEMESIS XLS,C,S = 1 − SHS,C,S · XC,S (I.29)

EUKLEMS Where s is the sectoral index, SHS,C,S the share of hours worked by high skilled workers in the EU-KLEMS database once converted to the NEMESIS sectoral nomen- NEMESIS clature and XC,S , the total employment/employees from the NEMESIS database (Eurostat being the original source). We made here an important assumption (see equations I.28 and I.29); since we assumed that high skilled and low skilled workers work the same time by employment unit. This represents of course a relatively important hypothesis, but the lack of information and data on this subject does not allow us to overcome this issue. We present in Figure I.9 the share of high skilled workers in total employment for the year 2005 at the European sectoral level, as deduced from the previous computation. First, we can observe very low level of the high skilled labour in almost all sectors. The sector that employs the most highly skilled workers is the services sector where the share can reach between 25% and 30% of the total labour demand. Although the sectoral repartition seems relatively logical, the levels appear to be very low.

Sectoral shares and sectoral levels (Eurostat & EU-KLEMS)

We now calculate the sectoral shares of high skilled and low skilled labour ( EUKLEMS EUKLEMS δX,HS,C,S , δX,LS,C,S ) in the national total employment and in the total number of employees using the EU-KLEMS database:

EUKLEMS EUKLEMS XHS,C,S δX,HS,C,S = P EUKLEMS (I.30) C XHS,C,S

EUKLEMS EUKLEMS XLS,C,S δX,LS,C,S = P EUKLEMS (I.31) C XLS,C,S

NEW NEW We then compute the national levels for high skilled and low skilled labour (XHS,C ,XLS,C EUROST AT ) using the Eurostat national share (SX,HS,C ) and taking the national labour and

37 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

Figure I.9.: European high skill share in total employment for NEMESIS sectors, (source EU-KLEMS)

35%

30%

25%

20%

15%

10%

0% . . . . . i. il . ls . . n s v r c. ch. h. ds on sp. e up cals c o rod nuf tio i n In er Ag Dist. El S i P a c ut Serv Serv. S Cokeas Ext. m Ma Equip. M u b . ication . & t as ined O ter r. Meta e al Prod.d. . Go . t. r tri in nd G G f r t t str s d Tra rket e Ch In fice Mac ink & Tob rin . & Cater.n ir Transp.nsp F l & Re Wa Me f e th. & Footw P the on Di g A ra i F & O El O C k, MarkeMa oal a O n r. ransp & od Inla r T n r C T d, Dr L Communa on no Rubber & Plastic ea & he B he & Ag ap. S t N . Foo P Ot O rr Non Met. Min. Prod. Tex., Clo e F

NEMESIS number of employees (XC ) of the NEMESIS model.

NEW EUROST AT NEMESIS XHS,C = SX,HS,C · XC (I.32)

NEW EUROST AT NEMESIS XLS,C = 1 − SX,HS,C · XC (I.33)

NEW NEW Using these national levels (XHS,C and XLS,C ) and the share of high skilled and low EUKLEMS EUKLEMS skilled labour (δX,HS,C,S and δX,LS,C,S ), we can calculate the new sectoral levels for NEW NEW high skilled (XHS,C,S) and low skilled (XLS,C,S) labour:

NEW EUKLEMS NEW XHS,C,S = δX,HS,C,S · XHS,C (I.34)

NEW EUKLEMS NEW XLS,C,S = δX,LS,C,S · XLS,C (I.35)

Final correction

Finally, to achieve the consistency between countries, we compute the new shares by NEW NEW skills (δHS,C,S and δLS,C,S) at the sectoral level, using the employment and the number NEW NEW of employees previously calculated (XHS,C,S and XLS,C,S). These shares will be used to

38 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.10.: European high skill share in total employment for NEMESIS sectors

45%

40%

35%

30%

25%

20%

15%

10%

0% . . . . i. l . s s c p v d. h. w i on ns rv. gr Oi lec rod. c ob. od. ti ion I e A Ext. etal ro a r t ater. Serv. s E P P oods P last uc ansp.. Ser ation & S a M M G & T t. P r C r ic ined in. emicalal e t sp n. G Gas Dist.f c nk in & & ir T M Ch i Fi arket Re WaterFerr. Sup. . Met ffi Pr ther Manuf.ons Distribu land Trans A , O Elect. ransp. DrEquip. O C & Tran MarketM Oil & & bber In ank Coal and Coke on Met T Lodg. Commun er n n p. Ru her B & Agr. & Ind. Mach. Sea Non Food, Pa Ot Oth . No Tex., Cloth. & Foot Ferr compute data on high skilled and low skilled labour at the sectoral level.

XNEW δNEW = HS,C,S (I.36) HS,C,S NEW NEW XHS,C,S + XLS,C,S

NEW NEW δLS,C,S = 1 − δHS,C,S (I.37)

Figure I.10 shows the corrected European high skilled labour share in total employment at the sectoral level for the year 2005. We can see that once corrected the services sector has a high skilled labour share between 35% and 45%, whereas it was only about 25% and 30% before the correction. Figure I.11 give some illustration of the high skilled labour share in total employment in “Agriculture”, “Chemicals” and “Bank, Finance and Insurance” sectors across the EU countries. Looking at “Agriculture”, the highest share of high skilled labour is in Estonia with 27.3% followed by Finland, Norway and Latvia with 25.4%, 23.2% and 22.6% respectively. We ﬁnd the lowest share for agriculture in Italy and Austria with 1.7%. If, we look at the “Bank, Finance and Insurance” sector, the highest shares are in Sweden, Belgium and Poland with more than 50% whereas the lowest is about 16% in Italy.

39 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

Figure I.11.: Sectoral illustration of the ﬁnal results

65%

60%

55%

50%

45%

40%

35%

30%

25%

20%

15%

10%

0% AT BE DE DK ES FI FR GR IE IT LU NL PT SE UK CZ EE HU LT LV MT PL RO SI SK NO

Agriculture Chemicals Bank, Finance and insurance

Compensation of employees data

Eurostat and EU-KLEMS data

A similar calculus is realized to correct for the EU-KLEMS shares of high skill and EUKLEMS low skill in the total compensation of employees (θHS,C,S ). We start by com- puting EU-KLEMS compensation of employees for high skilled and low skilled labour NEMESIS NEMESIS (COMPHS,C,S and COMPLS,C,S ) at the sectoral level, using the EU-KLEMS NEMESIS shares and the NEMESIS sectoral compensation of employees (COMPC,S - Eu- rostat being the original source).

EUKLEMS NEMESIS EUKLEMS COMPHS,C,S = COMPC,S · θHS,C,S (I.38)

EUKLEMS NEMESIS EUKLEMS COMPLS,C,S = COMPC,S · 1 − θHS,C,S (I.39)

Thus, using the EU-KLEMS data on high skilled and low skilled employees previously computed (equations I.28 and I.29), we can ﬁnd the EUKLEMS cost per employee.

40 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.12.: Ratio of European employee unit cost between high and low skills at sectoral level

2.25

1.75

1.5

1.25

0.75

0.5

0.25

. s ...... il p ls b d ic n rv. n v ke od. od. ch ch ip o ion er. sp. sp. io r rv O etal r a a u st tio t t n n t e Agri Co Dist. d Elec. Su Pr q c Se Se e r mica M M E Pr ru ica t d as n te M . . t. t ribu Tra p. n et S e n a he d ce k & To & Footw.in & Pla s & Ca d s u k G n fi in . . n n rk l a Refi W . Min. C th on Dist a a Ferr. Metal P Of ansp Dr o Other CManuf. la r M oa n . & I Elect. rGoods In T r Mar C Oil & Gas Ext. r T d, . & Pr Lodg r e no Met o e CommBank, hFin. & onIns. n Ag o Rubber Sea & thAir Tra N & F Pap O Ot No Tex., Cl Ferr.

EUKLEMS EUKLEMS COMPHS,C,S UCHS,C,S = EUKLEMS (I.40) SALHS,C,S

EUKLEMS EUKLEMS COMPLS,C,S UCLS,C,S = EUKLEMS (I.41) SALLS,C,S The unit costs by skill at the sectoral level allow us to compute the ratio between high and low skills compensations at the sectoral level for each country.

EUKLEMS EUKLEMS UCHS,C,S µC,S = EUKLEMS (I.42) UCLS,C,S Figure I.12 presents the ratio of European employee’s unit cost between high skill and low skill at the sectoral level. In average, we can see that high skilled workers are paid between 50% and 100% more than low skilled ones. For instance, we can observe that in the “Electrical Goods” sector, a high skilled employee cost 81% more in Europe than a low skill employee.

Final correction

41 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

NEW Now, we compute the corrected compensation of employees for the high (COMPHS,C,S) NEW and low (COMPLS,C,S) skills with the corrected shares of high and low skilled NEW NEW employees,SALHS,C,S and SALLS,C,S(see equations I.34 and I.35), the NEMESIS com- NEMESIS pensation of employees (COMPC,S ) and the cost ratios computed just before.

SALNEW COMP NEW = HS,C,S · COMP NEMESIS · µEUKLEMS (I.43) HS,C,S NEW NEW C,S C,S SALHS,C,S + SALLS,C,S

SALNEW COMP NEW = LS,C,S · COMP NEMESIS (I.44) LS,C,S NEW NEW C,S SALHS,C,S + SALLS,C,S Finally, using both last results we compute the corrected share of high and low skill in the total compensation of employees that will be used for rest of the study

COMP NEW θNEW = HS,C,S (I.45) HS,C,S NEW NEW COMPHS,C,S + COMPLS,C,S

NEW NEW θLS,C,S = 1 − θHS,C,S (I.46)

Figure I.13 displays the corrected European share of employees’ compensation for high skill at the sectoral level. Thus in average, the share of high skills’ cost in the total cost of employees is between 20% and 25% on average, but it is more than 40% in “Bank, Finance and Insurance”, “Other Market Services” and “Non Market Services” sectors, with respectively 40%, 53% and 45%.

Labour input and labour cost in the production function

Labour input

Low and high skilled labour (LLS and LHS), are the product of their share in total employees and the number of hours worked:

employees LLS,C,S,t = HEMPEC,S,t · δLS,C,S,t

employees LHS,C,S,t = HEMPEC,S,t · δHS,C,S,t

42 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.13.: Corrected European share of compensation of employees for high skill at sectoral level in 2005

55%

50%

45%

40%

35%

30%

25%

20%

15%

10%

0% . l ...... t t. i p ls h. s p c r. p. n v gri. is ec. a c ti ion ion te o erv Ex D d O l c t t ns Serv. A s E Ma ood qui Tob. Prod a ati & Ins. Ser a l Prod. G E & Footw. Plas uc r as ine r. Metals emi a & & Ca T et et S nd CokeG G r nd. t. p. nk int. ir unic a & Ch I ri r nstr ark l Ref Water Su t. Min. Prod. Met lec ns P o Distribu land Transp. A i e & Office EMachra D ber & Other CManuf. mm Mark oal O , & In & r n M C gr. T Lodg. Co non Fe Rub Bank, Fin. & A Sea the No . Food ex., Cloth.Pap. Other Transp. O rr Non M T e F

Where HEMPE is the total number of hours worked by employees (millions) or the NEMESIS employees previous XC,S, and δHS,C,S,t is the corrected share of high-skilled persons engaged (share in total hours) which has been computed previously.

Labour cost

Concerning the cost of labour i.e. PLLS and PLHS, we use the same method as for the labour input:

PLHS,C,S,t = COMPC,S,t · θHS,C,S,t

PLLS,C,S,t = COMPC,S,t · θLS,C,S,t

Where COMP is the Compensation of employees (in millions of Euros) or the previous NEMESIS COMPC,S and θHS,,C,S is the high-skilled labour compensation (share in total labour compensation) which has been computed previously.

43 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS

Estimation results for the eﬃcient speciﬁcation

The data sample includes 11 countries over 12 years (from 1989 to 2000) for each of NEMESIS’ 30 sectors. The sample of countries is restricted to 11 due to the lack of data over a long period for most European countries. Countries included in the sample are: Austria, Belgium, Denmark, Germany, Finland, France, Italy, Netherland, Spain, Sweden and the United Kingdom. Due to lagged values, the sample data is restricted to 11 years, which gives 121 observations per sector. As mentioned previously, we use a pooled panel estimation method with the FIML estimator. Estimations are made sector by sector. The small sample size imposes that the parameters are common to all countries (but can vary by sector) with the exception of the scale parameters. We estimate four elasticities of substitution, five delays, one technological trend and 11x5 scale parameters for each sector. Specifically, we estimate sector parameters in three stages. First, we estimate the scale parameters and technological trends with values constraints on other parameters. In the second step, we relax the constraints on elasticities. Finally, we relax all constraints on the parameters. If some estimated values give an unrealistic result which does not allow convergence, we constrain them to an extreme value. For example, the Chemical sector (sector 5), presents an estimated value for sigma-1 close to zero. We constrain it to 0.05 so that it corresponds to our limit value, allowing some substitution between factors. The most constrained parameters appear for the delay parameter on capital. In this case, the delay seems to be infinite. The constrained value corresponds to a median delay of 20 years. Table 18??? reports estimates of the four substitution elasticities. Constrained parameters take the value n.a. for their P-Value. All elasticities are positive, highlighting a positive substitution effect between factors for all sectors. All sector elasticities lie between 0 and 1, meaning that factors and factor bundle on the same level, are gross complements (for elasticity values higher than 1 we talk about gross substitutes). Re- sults differ between sectors but some trends are well identified. The first and fourth levels present lower elasticities values than intermediate levels. Labour (both skills) is more easily substitutable than other factors. Table 19 ??? reports estimates of the technological trend and delays. It appears that only a few technological trends are significant (only 15 .T. are significant at 10%). On the 15 significant values, 9 exhibit positive estimated parameters. Since technological

44 CHAPTER I. THE CORE ECONOMIC MODEL advances are very different between countries, common technological trends are neither relevant nor significant. Delays are highly significant and there is a common trend between sectors. Materials and Labour have a short adjustment delay (a mean of 0.9 years for Materials, 0.7 for low skilled and 1 for high skilled Labour). Energy presents a medium delay of adjustment due to its high complementarity with Capital (a mean of 1.8 years). Capital exhibits a relatively long delay (a mean of 14 years). Table 20 reports the R-Squared of the estimations. The R-Squared are as a whole relatively high, partly because we use panel data with individual fixed effect. In this case an important part of the variability between observations is explained by these fixed effects. Table 27 presents all scale parameters (Ao,s,c for M, LLS, LHS, K and E). The poor quality of the data for a long period enables us only to present results for 11 countries. Besides, one shoud recall that these previous data are the results of econometric estimations and are not a necessary condition for the implementation of the 27 countries into the model. Since the scale parameters are sectoral, they are substituted with the calibrated variables (which only need data for one year).

Independent Simulations

In order to assess the robustness of our estimations and to determine the direct elasticities of substitution between all factors, we simulate the demand factor evolutions in response to diﬀerent price shocks. Table 21 to Table 26 present these simulations, which are independent of the rest of the NEMESIS model. We increase each factor price by 10% and we report the factor demand evolutions. Results are consistent with theory: direct price elasticity is negative and the sign of indirect price elasticity depend of the synergy between factors. An increase of a factor price will induce a decrease in the demand of this factor, an increase in subsidiary factors and a decrease in complementary factors. All factors are subsidiary (negative indirect elasticities), except for energy and capital which are complementary for most of the sectors.

Estimations of the Substitution Possibilities Ex-ante and Ex-post

For the vintage and ex-ante/ex-post approach, we should use the following methodology to distinguish and estimate the diﬀerent parameters of the ex-ante and ex-post

45 I.2. NEW PRODUCTION FUNCTIONS WITH EMBODIED ENDOGENOUS TECHNICAL CHANGE AND SKILLS production functions.

Ex-post, the technical level being fixed, we then just have to estimate or calibrate the parameters of the factors’ (or bundle of factor) efficiency demands. So we could use the p p previous values of substitution elasticities σ . Moreover the ex-post parameters δi are endogenously calculated with the tangential technique as describe previously. We obtain the following parameters, for example, in the first level of the nested CES:

p a ρp−ρa δM,t+1 = δM [At+1 · mt+1]

and

p a δp = δa [A · ν ]ρ −ρ KELHS LLS ,t+1 KELHS LLS t+1 t+1

In the ex-ante side we have to determine also the substitution elasticities σa that are superior to the σp previously estimated. Referring to Meijers and Van Zon (1994), we could ﬁnd σa ≈ 1.5 · σp. As achieved in the article by Meijers and Van Zon, we could be more accurate and determine the ratio σa/σp by sectors.

a Furthermore, the ex-ante parameters δi are defined like in the example corresponding to the first level of factor efficiency demand of the nested CES:

a PM,t−t · Mt−t δM,t = PM,t−t · Mt−t + PKELHS LLS ,t−t · KELHSLLS,t−t

Finally, contrarily to the ex-post functions, we need to determine the parameters relative to the biased technical change. That is to say the parameters of the following function for each factor and for the global production:

1 −ρ − I = δ · Q−ρI + δ · I I ρI t QI,t I,t It t

Physical volume I and Technological level QI equation parameters could be calibrated using patent data for the level of productivity and with the European Community In- novation Survey (CIS). Using the amount of R&D expenditures and the CIS or patent data, we could split these expenditures between factors and global productivity or product. After that, using the innovation function, we can determine the parameters δQI that represent the weight in the cost of the quality or productivity improvement. The parameters σ should be determined the same way as describe above.

46 CHAPTER I. THE CORE ECONOMIC MODEL

Section I.3 Households’ ﬁnal consumption

The consumption behaviour is divided in two stages. The ﬁrst is the agregate consumption that splits households’ incomes in consumption and global saving. At the second stage, the agragate consumption is allocated in 27 consumption functions.

I.3.1 Aggregate consumption

At the begining of the aggregate consumption equation, there is the model of David- son et al. [92] in which the consumption is linked to income and wealth by an Error Correction Model. The econometrics estimate at first the long term relationship, then the dynamics. In the first version of the model, the wealth was represented by a permanent income function that was computed as a mean of the lagged revenues 2 ; later, the cumulated investment in dwelling was used as a proxy for the housing stock of households, and was added to the wealth effect. Other significant variables on the link between wealth on different support and consumption are interest rates and inflationary pressures. The unemployment rate is used as a proxy for the degree of uncertainty in the economy. Researches on the aggregate consumption are always going on, and they are now extended for the NEMESIS model in two directions : at first the building of a genuine wealth variable in a forward looking module that would be isolated from the rest of the model ; aggregate consumption function could the be the result of two type of behaviours ; that of “liquidity constrained” households which is founded on the current revenue ; that of “neoclassical households” which can borrow or lend liquidities without restrictions and which is grounded on wealth, discounted sum of future revenues. This structure could allow to focus attention on the effects of financial liberalization on consumption. CARRUTH and HENLEY [55] ; this focus could be achieved in increasing with liberalization the past of “neoclassical households” SEFTON and VELD[288]. Co-Integrating Long term equation

2The long run elasticity of consumption in relation to incomes has been set to one to ensure that the lifecycle theory is fulﬁlled

47 I.3. HOUSEHOLDS’ FINAL CONSUMPTION

CONSNAT NQc ln = lrscnn0c POPc INCGDISPc + lrscnn1 · ln P CONSNAT T OTc POPc POPRET + lrscnn2 · ln c POPc POPCHI + lrscnn3 · ln c POPc + lrscnn4 · ln(RRLRc) + lrscnn5 · DUM97

Dynamic equation

CONSNAT NQc ∆ ln = crscnn0c POPc INCGDISPc + crscnn1 · ∆ ln P CONSNAT T OTc POPc POPRET + crscnn2 · ∆ ln c POPc POPCHI + crscnn3 · ∆ ln c POPc + crscnn4 · ∆ ln(RRLRc) P CONSNAT T OTc + crscnn5 · ∆ ln −1 P CONSNAT T OTc −1 CONSNAT NQc + crscnn6 · ∆ ln −1 POPc + crscnn7 · ERR−1 + crscnn8 · DUM97 with:

• POPc, Population

• INCGDISPc, Gross Disposable Income

• P CONSNAT T OTc, Consumers’ Price

• POPRETc Retired Population

• POPCHIc Child Population

• RRLRc Interest Rate

• ERR, the Error Term

• DUM97, a dummy variable

48 CHAPTER I. THE CORE ECONOMIC MODEL

Parameters Restrictions:

lrscnn4 < 0 crscnn1 > 0

crscnn4 < 0

crscnn5 < 0

0 < crscnn6 < 1

0 > crscnn7 > −1

I.3.2 Allocation of aggregate Consumption

We will present in this section the theoretical and empicical grounds of the system that will allow to disaggregate the macroeonomic consumption determined above. The basic The presentation thereof are from Bracke I. and Meyermans E. [38]. The only difference with respect to their work as far as the econometric analyze is concerned, is that now panel estimation is applied instead of ’individual’ OLS regressions. The econometric allocation system is derived from the theory of rational consumer and restrictions imposed by it are implemented in a flexible way thanks to a CBS version of the system. The total aggregate consumption is therefore divided into 27 components as a function of relative prices and total income (to which are added demographic changes). Furthermore, that allocation module assumes groupwise separability, meaning that the consumer faces a decision problem in several stages. In the particular, the representative consumer decides, in a first stage, how much he will spend on "durable and complementary non-durable goods" on the one hand and on "other non-durable goods" on the other hand. In a second stage, he decides how to spend the money allocated in the first stage within the group i.e. how much of the amount dedicated to the durable goods will be allocated to clothing, household utilities and transportation. Transportation includes public transportation, equipment (such as cars) and energy, divided into petrol, heavy fuel and oil. A further decision stage takes place in the non-durable goods group. It consists of the choice between "necessities" (including food, beverages, tobacco, education, rent, health, electricity and other expenditure items) and "luxuries"(including communication, tourism and domestic services).

Based on the CBS parametrization, the long-run equilibrium relationship is:

49 I.3. HOUSEHOLDS’ FINAL CONSUMPTION

Figure I.14.: Allocation of Durable Goods

04 Clothing and footwear

10 Furnitures etc 11 Households textile 12 Major appliance D Durable goods FUR Furniture and equip. 13 Hardware 14 Household operation 15 Domestic services

17 Cars etc

T Transport 18 Petrol etc 19 Rail Transports 20 Buses and coaches OT Purchased Transport 21 Air Transports 22 Other Transports

Figure I.15.: Allocation of Non Durable Goods

01 Food FB Food Bev. and tob. 02 Beverages 03 Tobacco

05 Gross rent and water

NEC Necessities 06 Electricity 07 Gas FU Fuel and power 08 Liquid Fuels 09 Other Fuels 16 Medical care

ND Non Durable goods

23 Communication 24 Equipment and accessories incl repair LUX Luxuries 25 Recreation 26 Hotel and restaurant 27 Misc. Goods and Services

CONS 27 w ln c,i = c + b ln(INCRDISP ) + X s · ln (PCONS ) c,i INCRDISP c,i i c i,j c,j c j=1

+ g1,i ln(DEMPc) + g2,i ln(DEMWc) + ϑc,i

50 CHAPTER I. THE CORE ECONOMIC MODEL

and the short-run one is:

CONS 27 26 w ∆ ln c,i = bs∆ ln(INCRDISP ) + X ss · ∆ ln (PCONS ) + X f s ϑ−1 c,i INCRDISP i c i,j c,j i,j c,i c j=1 j=1 s s + h1,i∆ ln(DEMPc) + h2,i∆ ln(DEMWc) + uc,i

where :

• i, j = 1 to 27 consumption categories

• c = 1 to 26 countries

• CONS consumption of commodity (1995m euros)

• INCRDISP regional real personal disposable income (1995m euros)

• PCONS commodity price

• DEMW share of people of working age in total population

• DEMP share of old age people in total population

More speciﬁcally, under groupwise separability, the equations that follow were estimated. They show the interactions within a group of commodities and between groups of commodities. Within a group I, the long-run equilibrium relationship is:

CONS 27 w ln c,i = cI + bI ln(QI ) + X sI · ln (PCONS ) + gI ln(DEMP ) c,i QI c,i i c i,j c,j 1,i c c j=1 I I + g2,i ln(DEMWc) + ϑc,i

for i ∈ I and where

I P I • the scale eﬀect of group I is deﬁned by ln(Qc ) = i∈I wc,i ln(CONSc,i)

I • bi : the income coeﬃcient of commodity i in group I

I • si,j: the compensated price eﬀect of commodity j on I, both elements of I

I • wc,i : the budget share of commodity i in group I,

51 I.3. HOUSEHOLDS’ FINAL CONSUMPTION and the short-run one is:

CONS 27 26 wI ∆ ln c,i = bs,I ∆ ln(QI ) + X ss,I · ∆ ln (PCONS ) + X f s,I (ϑI )−1 c,i QI i c i,j c,j i,j c,i c j=1 j=1 s,I s,I + h1,i ∆ ln(DEMPc) + h2,i ∆ ln(DEMWc) + uc,i

for i ∈ I Between groups of commodities, the long-run equilibrium relationship is:

! QI k wI ln c = cI + bI ln(INCRDISP ) + X sI · ln (PCONS ) INCRDISP c c J c,J c J=1 I I I + g1 ln(DEMPc) + g2 ln(DEMWc) + ϑc

for I = 1, ..., k groups and where

I P I • ln(Qc ) = i∈I wc,i ln(CONSc,i)

I P I • ln(PCRc ) = i∈I wc,i ln(PCONSc,i)

• wI : the budget share of group I and the short-run one is:

! QI k k−1 wI ∆ ln c = bs,I ∆ ln(Q ) + X ss,I · ∆ ln (PCONS ) + X f s,I (ϑI )−1 INCRDISP c J c,J J c c J=1 J=1 s,I s,I + h1 ∆ ln(DEMPc) + h2 ∆ ln(DEMWc) + uc,I

for I = 1, ..., k groups. From those intra- and inter-group interactions, the overall interactions, which are defined as the interactions between commodities of different groups, may be computed. For the long-run overall coefficients:

I I mi = m · mi ∀i, I I I I IJ J si,j = si,j · w · δi,j + mi · S · mj ∀i, I, j, J

with δi,j = 1 only if I, j ∈ I and = 0 elsewhere and where

52 CHAPTER I. THE CORE ECONOMIC MODEL

I • mi : the marginal propensity to spend on commodity I in group I • mI : the marginal propensity to spend on group I

• mi: the overall marginal propensity to spend on commodity i (in the case of the

CBS parametrization, the marginal propensity to consume is deﬁned as mi = bi+wi )

I • si,j : the compensated price eﬀect of commodity j on i in group I (non zero only if i, j ∈ I)

• sI,J : the compensated price eﬀect of group J on group I

• si,j : the overall compensated price eﬀect of j on i

• wI : the budget share of group I. Mutatis mutandis, those equations may also be applied to compute the short-run overall coeﬃcients. Restrictions

Pn Pn Pn Pn s • Summability : i=1 cc,i = 0, i=1 bi = 0, i=1 si,j = 0, i=1 bi = Pn s 0, i=1 si,j = 0

Pn Pn s • Homogeneity : j=1 si,j = 0, j=1 si,j = 0

s s • Symmetry: si,j = sj,i, si,j = sj,i, ∀i, j

s • Negativity : sii < 0, sii < 0 The consumption per category is then allocated to consumption by product using consumption transition matrix (mcons) with ﬁxed coeﬃcient.

27 X ADDCONSQc,s = (mconsc,co,s · CONSc,co) co=1 This transition matrix is also used for calculating consumption price per category using sectoral production and import prices prices to which VAT taxes and Excises duties are added:

30 P (mconsc,co,s · CONSc,co · P ADDDEMc,s) + V AT CPc,co + EXCIP AHc,co s=01 PCONSc,co = 1995 1995 CONSc,co + V AT CPc,co + EXCIP AHc,co

53 I.4. EXTERNAL TRADE

Section I.4 External trade

External trade is of a crucial importance in applied models such as NEMESIS, indeed, one of the most important transmission effects between the different countries in the model goes through trade in goods and services. This matter of fact is reinforced by the strong European integration that as led to an increasing degree of openness, resulting in a increasing share of external trade ratio to the final demand. External trade is modelised in the models through a three sets of equations:

1. Intra-European trade in volume

2. Extra-European trade in volume

3. Exports and imports prices equations

If it were possible to separate intra and extra European trade in volume, this is not yet possible for prices, that’s the reason why no distinction is made between intra and non prices modelisation, except the fact that rest of the world trade prices includes trade barriers such as import duties, that are not present into intra European trade.

I.4.1 Intra-European trade

The basic assumption regarding intra-European trade is that it take place into a “trade pool”, i.e. into the same distribution network, that is to say that all European countries exports to this pool and imports from it. One of the major drawback of this kind of modelisation is that as exports and imports are both econometrically estimated, nothing insure that at the global European level, total exports and total imports are equals3. As underlined by Satchi [282] it is not yet possible to estimate trade equations without bilateral data, that follows straightforwardly this constraint. However, this “adding up” problem was solved by modifying the exports equations in order to insure the equilibrium

3The modelling of bilateral trade flows insure this “adding up” constraint, we are currently studying the possibility to modelise bilateral trade flows, at least for goods, as bilateral trade flows of services data are too weak for the moment.

54 CHAPTER I. THE CORE ECONOMIC MODEL between the sums of exports and the sums of imports per sector, this implicitly signify that imports equations are better modelised than exports ones. Numerous attempt had be made for integrating in external trade equations (particularly in exports equation) the so called non price competitiveness, one convicing attempt was made using quality index build up with using made on importors by Crozet et alii [86]. In our framework however, such quality indices are not available for the 27 modelised countries, and hence we had to estimate this effect through the Knowledge variable. Of course, taking knowledge as a proxy variable for quality covers as noted by Crozet et alii [86] not exactly the same content as quality indices, and may focus on a particular dimension of quality, technological differentiation. Moreover, empirical testing shows that bilateral trade flows are more suitable for estimating such quality effects, as this allows for changes in the direction of this trade (see Hallak [171]), that one of the reason why the possibility for implementing bilateral trade flows in NEMESIS is currently studied. A great part of international trade theory nowadays concerns the so called Home Market effect (See for instance Crozet et alii [87] or Corsetti et alii [85]) explaining that big countries have an advantage for specialising their production to increasing return to scale sectors, and on the contrary, small countries are more focused on constant return to scale production. However, this effect refers largely to world trade, and the European Integration tends to largely reduce this effect. Finaly, the borders effect was not taking into account in our modeling framework the “trade pool” hypothesis does not allow for bilateral trade, moreover, one can argue that the European Market integration tends to reduce this effect (see Chen [63])

Imports equations

The three main effects integrated in the trade equations are income and prices effects and non prices effects. For imports equations, these effetcs are taken into account with the following variables

• The income eﬀect for a country is taken into account through a demand variable, represented by the demands addressed to the sector

• The price eﬀect is represented by the ratio of the import price to the domestic price.

55 I.4. EXTERNAL TRADE

• the non price eﬀect is taken into account through national knowledge stock to European knowledge stock ratio

ln(IMPEUQc,s) = limpeu0c,s

+ limpeu1s · ln(ADDDEMQc,s) ! PIMPc,s + limpeu2s · ln PPRODc,s ! KNOWc,s + limpeu3s · ln KNOWeu,s

with :

• ADDDEMQc,s Total domestic Demand by products

• PIMPc,s The Price of Imports

• PPRODc,s The production Price

• KNOWc,s the national knowledge Stock

• KNOWeu,s the European knowledge Stock

Parameters Restrictions:

limpeu1s > 0

limpeu2s < 0

limpeu3s < 0

Exports Equations

For exports equations, the incomes and prices eﬀetcs are taken into account with the following variables

• The income eﬀect for a country j is taken into account through a demand variable, resulting from the demands of partners’ countries, weighted past trade intensities (in a matrix form, for year 2000)

• The price eﬀect is represented by the ratio of the export price to a European price index, which is a weighted variable of other EU countries export prices.

56 CHAPTER I. THE CORE ECONOMIC MODEL

ln(EXP EUQc, s) = lexpeu0c,s

+ lexpeu1s · ln(INDACT EUc,s) ! PEXPc,s + lexpeu2s · ln PINDICEXPEUc,s ! KNOWc,s + lexpeu3s · ln KNOWeu,s

• INDACT EU c,s, indicator of activity

• PEXPc,s, the Export Price

• PINDICEXPEUc,s, Indicator of competing Prices

• KNOWc,s the national knowledge Stock

• KNOWeu,s the Global European knowledge Stock

Parameters Restrictions:

lexpeu1s > 0

lexpeu2s < 0

lexpeu3s > 0

I.4.2 Extra European Trade

Extra European trade vis à vis of the rest of world (divided into ten exogenous areas), follows broadly the same formalisation than intra European Trade and includes therefore the same eﬀects as described above.

Imports equations

57 I.4. EXTERNAL TRADE

• The income eﬀect for a country is taken into account through a demand variable, represented by the demands addressed to the sector

• The price eﬀect is represented by the ratio of the import price to the domestic price.

• the non price eﬀect is taken into account through national R&D stock to the extra European zone R&D stock ratio

ln(IMPROWQc,s) = limprow0c,s

+ limprow1s · ln(ADDDEMQc,s) ! PIMPROWc,s + limprow2s · ln PPRODc,s ! KNOWc,s + limprow3s · ln KNOWz,s

with :

• ADDDEMQc,s Total domestic Demand by products

• PIMPROWc,s The Price of Imports for extra European imports

• PPRODc,s The production Price

• KNOWc,s the national knowledge Stock

• KNOWz,s the extra European zone knowledge Stock

Parameters Restrictions:

limprow1s > 0

limprow2s < 0

limprow3s < 0

Exports Equations

For exports equations, the incomes and prices eﬀetcs are taken into account with the following variables

58 CHAPTER I. THE CORE ECONOMIC MODEL

• The price eﬀect is represented by the ratio of the export price to a non European price index, which is a weighted variable of other extra European zone export prices.

ln(EXP ROW Qc, s) = lexprow0c,s

+ lexprow1s · ln(INDACT ROWc,s) ! PEXPROWc,s + lexprow2s · ln PINDICEXPROWc,s ! KNOWc,s + lexprow3s · ln KNOWz,s

• INDACT ROW c,s, indicator of activity

• PEXPROWc,s, the Export Price

• PINDICEXPROWc,s, Indicator of competing Prices

• KNOWc,s the national knowledge Stock

• KNOWz,s the extra European zone knowledge Stock

Parameters Restrictions:

lexprow1s > 0

lexprow2s < 0

lexprow3s > 0

I.4.3 Imports and Exports prices

Exports and imports prices play a large role in determining trade volumes. the basic feature of trade prices in NEMESIS assume that European countries operate in oligo- polictic markets, following this assumption, importers and exporters sets mark-ups on their prices taking others partners prices into account. As noted above, the lack of data

59 I.4. EXTERNAL TRADE regarding import and exports prices differentiated per trade partners, make that the distinction between Intra-European and Rest of the World distinction was not possible, however, in order to take into account for possible trade barriers between the EU and the rest of the world, we sets two different prices, the sole difference between the two prices lay precisely in the existing trade barriers (import duties...) that multiplies the global import and exports prices. Other partners prices are weighted in the same manner than for volume equations, exchange rates are directly taken into acccount in the European (making a clear distinction between intra and extra Euro zone) and ROW price index . The majority of trade prices are treted in the same manner (with the notable exception of crude oil and gas, that are treated exogenously)

Export prices

ln(P EXP c, s) = lpexp0c,s

+ lpexp1s · ln(PINDICEXPEUc,s)

+ lpexp2s · ln(PINDICEXPRWc,s)

+ lpexp3s · ln(PPRODc,s)

• PINDICEXPEUc,s, Price Index for competing Exports in Europe

• PINDICEXPRWc,s, Price Index for competing Exports in the Rest of the World

• PPRODc,s, Production Price

Parameters Restrictions:

lpexp1s + lpexp2s + lexp3s = 1

Import prices

60 CHAPTER I. THE CORE ECONOMIC MODEL

ln(P IMP c, s) = lpimp0c,s

+ lpimp1s · ln(PINDICIMPEUc,s)

+ lpimp2s · ln(PINDICIMPRWc,s)

+ lpimp3s · ln(PPRODc,s)

• PINDICIMPEUc,s, Price Index for competing Imports in Europe

• PINDICIMPRWc,s, Price Index for competing Imports in the Rest of the World

• PPROD, Production Price

Parameters Restrictions:

lpimp1s + lpimp2s + lpimp3s = 1

Section I.5 Wage setting

In section we will present the specification, the estimation and the implementation of this modified labour market for the Nemesis model. Indeed, the integration of different labour skills in the model implies to reformulate and to extend the labour market of NEMESIS. This will allows us to revisit the latest theoretical developments and to proceed to an econometric analysis at a disaggregated level. This section is organized as follows, we first analyse the theoretical issues and the consensus that has emerged in the last years, and then in a second part we will define the formalization that will be implemented in the model. The third part is dedicated to the presentation of the data used in the econometric estimation that will be presented in the fourth part. Finally, we will present the functional form implemented in the NEMESIS model.

I.5.1 Theory

61 I.5. WAGE SETTING

The formulation of wage process suﬀers from a lack of consensus arisen from a long and stormy history. Main empirical and theoretical controversy opposes proponents of the Philips curve to those of the WS-PS model. Philips curve is an empirical relation that highlights the negative relation between nominal wage and unemployment. It could be well represented by :

4w = c + 4pe − bU (I.47)

All variables are expressed in logarithm except U which is express in level. 4w is e the variation of nominal wage (w − w−1), 4p is the expected inflation (it is equal to 4p if expectations are perfect) and is U the unemployment rate. Whereas Philips curve estimates well wage formation over the course of a business cycle, it suffers from a lack of theoretical foundation. At the opposite, the WS-PS models are theoretical based but present some unrealistics assumptions about their key concepts. Almost all these models4 are founded on assumption that real wage fluctuates around a reservation wage which representes the income opportunity of employees outside the firm. Main models explain wage reservation by unemployment benefit, labor productivity, positive trend or lagged real wage. Recent litterature (Chagny and ali (2002)[58] and Reynes (2006)[134]) rejects the theoretical underpinning of the first three explanations and retains the latter, leading to a Philips curve specification. Almost all theoretical models based of bargain model or efficiency wage can be represented as:

∼ ∼r w = w − p = w + Z − bU (I.48)

∼ ∼r w is the real wage, w is the reservation wage and Z embodied all other variables that can explain wage formation (almost institutional variables). As highlight by Manning (1993)[4], Blanchard & Katz (1999)[255], if we consider reservation wage as the lagger ∼r ∼ real wage (w = w−1 = w−1 − p−1), equation can be transformed into a Philips curve.

4w = 4p + Z − bU (I.49)

In assuming that the reservation wage is the lagged wage, the Philips curve theoretical underpinnings are as valid as those of the WS setting. Chagny and ali (2002) and Reynes (2006) go father in narrowing the empirical diﬀerence between the two approachs. Their models allows “a clear distinction between medium run of equilibrium rate of

4For instance eﬃciency wage model, matching model or competitive wage competition...

62 CHAPTER I. THE CORE ECONOMIC MODEL unemployment (ERU) and the long run ERU” which the the key diﬀerence between Philips curve and WS-PS models. Let’s assume that the medium run wage formation is directed by :

4w = Z + a4pcons − b1U − b2(U − U−1) + d4π − f4tcs (I.50)

In this speciﬁcation, wage formations may be: indexed on consumer price pcons, hys- theris or not b2, depend of labour productivity π, employer’s social contribution tcs and inﬂuenced by a pool of institutional variables Z. The long run ERU is:

0 UELR = (Z − (1 − d)4π − (1 − a)4p )/b1 (I.51)

0 where 4p is the inﬂation target of the monetary authorities. UELR diﬀers from the medium run (assume that b2 = 0) by:

d UEMR = UELR + (w − w )/b1T (I.52)

where T is the number of quarter during which authorities are implicitly assumed to correct the unemployment gap.

I.5.2 Model

We extend the model developped by Chagny and ali (2002) and Reynes (2006) in order to estimates the wage formation. We transform equation I.56 to take into account Nemesis speciﬁcities. Since Nemesis models is sectorial and intergates two kinds of labour (high skill and low skill), equation retains is as following:

4wi,l,c,t = Zi,l,c + ac,l(L)4pcl,c,t − b1,c,l(L)Ul,c,t − b2,c,l(L)(Ul,c,t − UTl,c,t) (I.53)

+dc,l(L)4πi,l,c,t + i,l,c,t

Where • t = 1; ...; 11 is a time index ranging from t = 1992 to 2005.

• c is a country index, c = 1, .., 19

63 I.5. WAGE SETTING

• l corresponds to the labour qualiﬁcation

• i is a sector index, i = 2, ..., 29

• Zi,l,c represents institutional variables.

UT correspond to the tendential unemployment rate. Due to lack in data we limit our estimated sample to 19 countries (instead of 27) and 28 sectors (instead of 30).

Institutional variables Zi,l,c are treated as country-sector ﬁxed eﬀect.

I.5.3 Data

Wage

Unfortunately there is no data available for wi,l,c (or 4wi,l,c), which make the distinction between diﬀerent kinds of labour. By deﬁnition the variation of wage equal the variation of labour compensation and the variation of employer’s social security rate, i.e.4wi,l,c =

4Compi,l,c + 4tcs,i,l,c with Compi,l,c the labour compensation. Under hypothesis that

4tcs,i,l,c = 0, wage variation equal labour compensation variation, which is available in the EUKLEMS[309] database. Compi,l,c is built as follows:

LABHS COMP ∗ 100 CompHS = HHS HEMPE ∗ 100 Where HEMPE is the Total hours worked by employees (millions), COMP the Compensation of employees (in millions of Euros), LABHS the High-skilled labour compensation (share in total labour compensation) and HHS the Hours worked by high-skilled persons engaged (share in total hours). CompHS is thus the hours labour compensation for high skill workers. Same method is used for the low skill compensation.

Unemployment

Ul,c is the unemployment rate. Because there is no data available and we assume that workers are mobile through sectors, unemployment rate is not deﬁned by sector. Data

64 CHAPTER I. THE CORE ECONOMIC MODEL

Low skil High Skill mean standard error mean standard error AT 2.07 0.84 1.79 0.74 BE 2.35 1.23 2.49 1.57 DK 3.34 0.70 3.70 0.75 GE 1.45 1.13 2.01 1.55 FI 3.04 1.60 3.85 1.37 FR 3.63 1.91 1.50 2.10 GR 5.27 1.23 3.66 2.28 IR 5.49 1.64 5.70 2.73 IT 1.76 1.86 3.42 3.07 NL 3.75 1.32 4.30 2.40 PT 3.38 2.62 3.68 3.31 SP 2.59 0.81 2.80 0.94 SW 3.65 2.18 2.87 2.08 UK 5.19 1.67 3.05 4.00 CZ 6.80 2.80 7.00 3.60 HU 10.40 5.71 11.33 5.34 PL 7.28 5.66 6.89 4.72 SN 7.92 2.62 6.99 4.89 SK 8.72 3.51 8.20 3.39

Table I.1.: Labour compensation growth, period 1998-2005

65 I.5. WAGE SETTING

Low skil High Skill mean standard error mean standard error AT 5.39 0.48 2.33 0.39 BE 9.66 1.39 3.60 0.52 DK 5.28 0.54 3.60 0.67 GE 10.60 1.55 4.86 0.57 FI 13.54 1.60 4.80 0.76 FR 11.03 1.68 6.09 0.79 GR 11.40 0.88 7.73 0.61 IR 6.02 1.72 2.33 0.47 IT 10.34 1.72 6.10 0.78 NL 4.02 1.25 2.13 0.55 PT 5.68 1.50 3.99 1.37 SP 13.76 3.15 9.90 2.92 SW 7.45 1.88 3.59 0.88 UK 6.28 0.83 2.64 0.32 CZ 8.52 1.03 2.41 0.43 HU 7.58 1.23 1.76 0.53 PL 18.80 4.16 5.59 1.93 SN 7.50 0.73 2.84 0.53 SK 18.84 2.64 4.59 0.92

Table I.2.: Unemployment rate , period 1998-2005 are taken from Eurostat, we use set called “Unemployment rates by sex, age groups and highest level of education attained (%)”. We retain as “age groups”, the 15-64 years old. We convert ISCE classification (International Standard Classification of Education) into low and high skill classification.

Price

pcl,c is taken from the OECD database: “Consumer price indices (MEI)” for all countries except for Slovenia. Since database not available for Slovenia we use Eurostat price (with lower time coverage).

66 CHAPTER I. THE CORE ECONOMIC MODEL

HHS share P CONS mean standard error mean standard error AT 12.50 0.99 1.75 0.74 BE 14.71 0.60 1.90 0.68 DK 7.50 0.69 2.14 0.53 GE 9.28 0.39 1.33 0.45 FI 33.84 1.00 1.43 0.96 FR 13.86 0.89 1.55 0.62 GR 19.91 1.63 3.45 0.64 IR 15.88 1.98 3.41 1.46 IT 10.78 1.29 2.28 0.39 NL 10.73 1.35 2.38 0.93 PT 9.41 1.42 2.98 0.72 SP 19.38 1.51 2.96 0.60 SW 16.84 2.52 1.05 0.98 UK 16.78 1.69 1.37 0.36 CZ 12.71 0.97 3.50 3.22 HU 17.62 1.82 7.91 3.51 PL 14.72 2.66 5.28 3.97 SN 16.63 2.52 6.35 2.34 SK 14.20 1.49 7.32 3.24

Table I.3.: Price growth and high skill share , period 1998-2005

67 I.5. WAGE SETTING

Low skil High Skill mean standard error mean standard error AT 3.21 1.20 -0.57 1.94 BE 1.71 1.68 -0.23 2.03 DK 2.16 1.09 -1.56 2.12 GE 2.31 1.01 0.26 4.85 FI 2.62 1.21 1.47 2.46 FR 2.68 1.60 -0.83 2.27 GR 1.68 2.56 -2.07 5.11 IR 3.69 1.99 -1.84 4.21 IT 1.08 1.06 -4.61 1.14 NL 1.97 1.84 -2.85 8.89 PT 1.37 1.40 -2.47 7.49 SP 2.05 1.01 -1.89 1.37 SW 3.44 1.85 -3.35 5.65 UK 2.72 1.33 -2.35 1.78 CZ 6.09 2.96 2.80 3.38 HU 5.69 1.87 1.02 5.53 PL 6.50 2.21 -0.61 4.96 SN 4.17 2.68 -1.61 9.68 SK 4.67 3.51 0.55 5.13

Table I.4.: Labour productivty growth, period 1998-2005

Labour productivity

πi,l,c is the labour productivity. It is calculated as usual way, i.e. the GDP divided GOQI by the number of hours worked (πHs = HHS for unskilled labour and πLs = HEMPE∗ 100 GOQI HHS for skilled labour). GOQI is also taken from the EUKLEMS database HEMPE∗(1− 100 ) (Gross output, volume indices, 1995 = 100).

I.5.4 Results

In order to highlight the diﬀerence between the macro and the sectorial view, we implement two set of regression. First set is made at the macro level. Second set is made with sectorial speciﬁcations. To deal with error autocorrelation we use an order-1

68 CHAPTER I. THE CORE ECONOMIC MODEL

Macro b2=0 Sec b2=0 Price Unemp Productivity Price Unemp Productivity LS coef 0.42 -0.46 0.41 0.76 -0.58 0.24 se -0.32 0.33 0.19 0.46 0.52 0.17

HS coef 0.69 -1.34 0.29 1.05 -0.52 0.15 se 0.43 1.09 0.37 0.54 0.50 0.08

Table I.5.: Coeﬃcients summary auto-regressive model (AR(1)).

Macro

The related equation for regressions at macro level is a transformed form of equation I.53:

4wl,c,t = Zl,c + ac,l4pcl,c,t − b1,c,lU − b2,c,l(Ul,c,t − UTl,c,t) + dc,l4πl,c,t + l,c,t (I.54)

In a first time we presume the non existence of hysteresis phenomena. In that case we constraint b2 = 0. Results of macro regression without hysteresis are given in Figure1. In a second step we test the presence of hysteresis phenomena, results are given in Figure 2. Results of first estimation present expected coefficient sign for 20 specifications on 38 (sample include 19 countries with two labour market that give 38 specifications). 6 countries: GE,FR,IR,NL,HU and SN present expected sign for both markets. In some case coefficient value exceed the unity, it’s arise only on the high skill market (except for Hungary). Table I.5 gives aggregated results from regressions which present value of the expected sign and lower than 2. Price and unemployment seem to play a more important role for high skill wage, at the opposite productivity has a stronger effect on low skill wage. Figure I.17 presents coefficient for the whole model. Most of the results are in the opposite sign than expected one. For this reason we reject this specification.

69 I.5. WAGE SETTING

Figure I.16.:

70 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.17.: results whole model

71 I.5. WAGE SETTING

Sectorials results

The related equation for sectorial regressions is transformed form of equation I.53:

4wi,l,c,t = Zi,l,c +ac,l4pcl,c,t −b1,c,lUl,c,t −b2,c,l(Ul,c,t −UTl,c,t)+dc,l4πi,l,c,t +i,l,c,t (I.55)

Main difference with equation I.54 is the sectorial specification of labour compensation and labour productivity. Results of sectorial regression are given in Figure I.18 and in Figure I.19. We proceed in the same way than in macro specification, in the first step we estimate equation I.55 with b2 = 0 then we relax the constraint. First set of results (Figure I.18) present expected coefficient sign for 24 specifications on 38. 8 equations, which present non expected value are the same than in first regression (AT:HS, BE:LS, FI, SW, PL:HS and SK:LS). Table I.5 presents aggregated results for regressions related for Figure I.18. Price seems to play a more important role for high skill wage. Unemployment has a same effect for both kind of labour and productivity has a stronger effect on low skill wage. In comparison with macro results coefficient value are higher for price and unemployment coefficient but lower on productivity coefficient. Figure I.19 present coefficient for the whole model. Most of the results are in the opposite sign than expected one. For this reason we reject this specification

ANNEX

Long run equilibrium rate of unemployment UELR

Recall the relevant wage setting equation:

4w = Z + a4pcons − b1U − b2(U − U−1) + d4π − f4tcs (I.56)

Consider 4p0 is the constraint inﬂation target of the monetary authorities. In the long run 4tcs = 0, U − U−1 = 0 it follows:

72 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.18.: Sectoral results P1

73 I.5. WAGE SETTING

Figure I.19.: sectoral results P2

74 CHAPTER I. THE CORE ECONOMIC MODEL

0 4w = Z + a4p − b1U + d4π

The unemployment rate consistent with the target inﬂation rate and growth rate of labour productivitty is:

0 0 4w − 4p = 4π = Z + (a − 1)4p − b1U + d4π

We note the long run equilibrium rate of unemployment U = UELR.

0 UELR = (Z − (1 − d)4π − (1 − a)4p )/b1 (I.57)

Section I.6 Labour supply

This section presents the modelling of labour supply in NEMESIS. The decision by individuals to participate or not to labour market relies on many socio-economic and institutional factors, such as the levels of wages, of reservation wages, of social transfers, and the dynamism of labour market. A first sub-section states the situation prevailing in the different EU member States for participation rates of working age population by categories. Then a second sub- section presents the econometric works realized for MODELS for endogenizing the labour supply for the different population categories. The data sets that were used to model participation rates in NEMESIS did not allowed distinguishing, for a given age group, the supplies by high skilled and low skilled workers, but only the supply by gender categories. The results allowed nevertheless giving parameter values to calibrate the labour supply for the different skills that were introduced in NEMESIS for MODELS, as it is explained in the third sub-section. The last sub-section concludes.

I.6.1 The data on participation rates of working-

75 I.6. LABOUR SUPPLY

Figure I.20.: Participation rates to labour market of men and women aged 25 to 64, EU27 + Norway, 2005 96%

94%

92%

90%

88%

86%

84%

82%

80%

78% BE DK DE IE GR ES FR IT LU NL AT PT FI SE UK CZ EE LV LT HU MT PL RO SI SK NO

Males Females

age population

Data on labour supply are numerous. EUROSTAT provides participation rates by sex, age groups and skills for each EU country either annually or on a quarterly basis. The Figure I.20displays the participation rate of the high skilled women and men between 25 and 64 years old in 2005. One can observe on this figure an important diversity between countries especially for women. The participation rate is about 80% in Malta and Czech Republic whereas it is more than 90% in Sweden and Portugal. The males’ participation rates differ less between countries and range between 89% in Austria and Romania and 94% in Ireland. Finally, the data confirm that the activity rates of men are, in average, higher that those of women, but relatively close in few countries as Portugal, Sweden and Romania. Figure I.21 shows the activity rates for women aged from 50 to 65 and for low skilled and high skilled populations in 2005. For this elder population category, participation rates are always superior for high skilled than for low skilled women, of up to about 8 to 10% to the population composing this age group in countries like Italy, Belgium, Spain, Greece and Luxembourg. One can state finally on this figure that there exist fewer discrepancies among EU countries for participation rates of the high skilled women that for the low skilled women population for this age group. The previous figures illustrate the contrasts existing into the behaviour of labour sup-

76 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.21.: Participation rates to labour market of women aged 50 and 64 by skill, EU27 + Norway, 2005

77 I.6. LABOUR SUPPLY ply between countries, age groups, sex, and skills. The next sub-section will then try identifying the influence of key socio-economic indicators on the decision by individuals to participate to labour market. In economic theory, individuals realize a trade-off between work that provides income for consumption and leisure. Another important determinant of the labour supply behaviour, linked to this arbitrage between consumption and leisure, is the value of the reservation wage that one can proxy by the amount of social transfers per head received by individuals. Also, a dynamic labour market, with growing employment, will encourages individuals to provide their labour force, while, on the contrary, high unemployment rates will discourage them to enter on it. There exists also strong trends affecting the behaviour of labour supply reflecting structural changes occurring in European societies, that occur for example in the context of labour market regulation, in retirement policies and in social and professional aspirations of the young generations. The use of exogenous time trends, and of proxy variables such as the share of population aged 55-64 in total working-age population, to catch the effects of retirement policies, and of school enrolment ratios, to catch changes in youngest population social aspirations, may help us to control the influence of these additional factors affecting the labour supply. We had also to deal with data scarcity problems which have constrained us to limit our analysis to age groups and gender categories, and to leave the skill dimension to future works.

I.6.2 Determinants of participation rates

The goal of this sub-section was to ﬁnd functional forms for participation rates to labour market of the diﬀerent population categories that were introduced in the model NEMESIS. As it was underlined just above, the skill dimension was leaved away, in reason of the lack of data on labour compensations and social transfers at skill level. After a presentation of the data used, we introduce the functional forms that were used and we end by a comment of the econometric results.

The data and the choice and construction of indicators

The choice of indicators for the modelling of participation rates to labour market were based of a set of empirical studies on participation rates (Mincer [244], Hughes [180],

78 CHAPTER I. THE CORE ECONOMIC MODEL

Jacobsen [192], Schrier (2000), Cutler and Turnbull [88], Conesa et al. [82], Huber [179] and Balleer et al. [25]). The data were taken from EUROSTAT. They cover all EU27 countries plus Norway, have an annual periodicity, and range from 1998 to 2004 . The S,A participation rates by sex and age groups (T xi,t ) were modelled by using the following set of explanatory variables:

wi,t/pi,t • Gwi,t = that represents the change in real wage between t − 1 and t, wi,t−1/pi,t−1

LS • GLS = i,t , the increase in jobs creation between t − 1 and t, i,t Li,t−1

Un poor Un poor farm (SB +SB )/pi,tUi,t SB +SB +SB /pi,tUi,t • GSBY = i,t i,t , GSBM = i,t i,t i,t i,t SBUn +SBpoor /p U i,t Un poor farm ( i,t−1 i,t−1) i,t−1 i,t−1 SBi,t−1+SBi,t−1+SBi,t−1 /pi,t−1Ui,t−1 Un poor old O (SBi,t +SBi,t +SBi,t )/pi,tUi,t and GSBi,t = Un poor old , that are indicators of change in (SBi,t−1+SBi,t−1+SBi,t−1)/pi,t−1Ui,t−1 social beneﬁts between t − 1 and t,

S S STi,t • SSTi,t = S,Y , the share of student in ISCED 5 and ISCED 6 in the 15 to 24 POPi,t−1 years old population,

Y M S POPi,t+POPi,t • and SPOPi,t = Y M O , the share of the young and medium age POPi,t+POPi,t +POPi,t groups in the total population (except the cohorts between 0 to 14 years)

We have:

• S = M,W the gender index,

• A = Y,M,O the index of age groups, the class Y represents population between 15 and 24 years; the class M regroups population between 25 and 64 years and the class O includes the population aged 65 years and more,

• i the index for countries,

• t the time index,

S • Li,t the total employment by gender,

S,A • POPi,t , the total population by sex and age groups,

Un P oor farm Old • SBi,t ,SBi,t ,SBi,t ,SBi,t , the social transfers for, respectively, unemployment, poverty:, family: and oldness,

• wi,t, the nominal wage rate,

• pi,t, the price index of ﬁnal consumption,

79 I.6. LABOUR SUPPLY

S • STi,t, the total number of student in ISCED 5 and 6 by sex,

S • Ui,t, the total number of unemployed by sex.

Among the explanatory variables, the change in real wages will capture the importance of the trade-off between work and leisure or between consumption and leisure. The wage rate used is identical for men and women and all age groups. The change in employment will allow measuring the flexion of labour supply to the dynamism of labour market. An alternative consists in using the evolution of unemployment rate, but this rate, at the difference of the change in employment, is partially endogenous, as it is relies by construction on the participation rate to labour market; the use of unemployment rate in the econometric specification could consequently lead to biased estimators. For social benefits, an indicator was constructed for each cohort, even if data on social benefits do not cover age groups, nor population genders. We constructed nevertheless indicators oriented toward particular age groups, by normalizing the social benefits by the respective unemployment rates of the different age groups. This allows taking into account the impact of changes in unemployment onto the evolution of social benefits per head in volume (deflated with final consumption price). For population aged 15 to 24, social benefits include unemployment allowances and poverty assistance, for population aged 25 to 64, they include also assistance to families, and for population over 64, the retirement pensions. In addition, the share of population aged between 15 and 24 engaged in tertiary education traduces the trade-off between work and studies. Finally, the share of young and medium in total working-age population aims capturing the influence of retirement policies on the labour supply behaviour of the eldest age group.

The modelling of participation rates

The participation rates were estimated econometrically from temporal trends that take the form of logistic curves, and the set of indicators that was described above. The logistic curves have the following form:

 S,A  T xi,t − µ ln  S,A  = σ · time + ρ (I.58) λ − T xi,t with µ and λ respectively the limit activity rates high and low, σ the diﬀusion speed of activity behaviours and − (ρ/σ) the year of inﬂexion of the activity rate. We obtain by developing:

80 CHAPTER I. THE CORE ECONOMIC MODEL

µ + λ · e(σ·time+ρ) T xS,A = (I.59) i,t 1 + e(σ·time+ρ) As the participation rates were estimated with pooled panel techniques, µ and λ were set respectively to 0 and 1, and the parameters σ and ρ were individualised by country. We have:

σS,A·time+ρS,A e( i i ) T xS,A = (I.60) i,t σS,A·time+ρS,A 1 + e( i i ) By introducing the other determinants of participation rates, we get ﬁnally the expressions that were estimated:

S,A S,A S,A S S,A A S,A S,A T xi,t = βw · GWi,t + βL · GLi,t + βSB · GSBi,t + βST · SSTi,t σS,A·time+ρS,A e( i i ) + βS,A · SPOP S,A + + S,A (I.61) POP i,t σS,A·time+ρS,A i,t 1 + e( i i )

The elasticity parameters βX are common to all countries and i are the i.i.d. error terms. Parameter restrictions were imposed with respect to the population groups that are considered:

S,A • βST = 0 when A = Y,M i.e. we suppose a null eﬀect of the student share (SST ) on superior age groups.

S,A • βPOP = 0 when A = Y,M, i.e. we suppose a null eﬀect of the share of 15 to 64 years old population (SPOP ) on young and medium age groups.

The model was estimated by the Full Information Maximum Likelihood procedure (FIML) with annual data and 92 independent observations. There are, depending of the population category, 3 to 4 elasticity parameters βX and 21 country speciﬁc con- S,A S,A stant ρi and time trends σi to estimate.

Estimation Results

Table I.6 displays the estimation results for the parameters , which measure the short term impacts of changes in economic conditions, in population structure and in educa-

81 I.6. LABOUR SUPPLY

Table I.6.: Estimation results of participation rates

S A S S 2 Gender age Gw GL GSB SST SPOP Adjusted R DW -0.0088 0.3154*** -0.0171 -0.7719** -- 0.98 2.04 15-24 (0.0299) (0.0514) (0.0133) (0.3136) 0.0631*** 0.3040*** -0.0285*** -- -- 0.99 2.51 Males 25-64 (0.0177) (0.3040) 0.0076) 0.0327 0.0078 -0.0226** -- -0.5028* 0.98 2.1 65-max (0.0216) (0.0646) (0.0092) (0.2899) -0.02969 0.3664*** -0.0266* -0.9362*** -- 0.99 2.23 15-24 (0.0477) (0.05571) (0.0142) (0.1283) -0.0142 0.2531*** -0.0136* -- -- 0.97 1.98 Females 25-64 (0.0142) (0.0478) (0.0079) -0.0002 -0.0214 -0.0067 -- -0.4758*** 0.97 1.61 65-max (0.01461) (0.0474) (0.0046) (0.0540) *,**,***: parameter signiﬁcantly diﬀerent to zero at 10%, 5% and 1% respectively tion, on the evolution of activity rates5.

The parameters βw can be interpreted as the semi-elasticity of activity rates to vari- ations in the corresponding explanatory variable.

For real wages, one can see in the first column of the table that the parameter βw is significant only for men aged between 25 and 64. For this population category, an increase of 1% of the growth of real wages increases labour supply of about 0.06 point, which represents an augmentation of 0.64%. For employment effect, that measures the sensitivity of labour supply to an ameliora- tion, or a deterioration of labour market, the second column of the table indicates that the parameters βLare significant at 1% level for every categories, except persons aged 65 and over, with anyway very low participation rates to labour market. The parameters values are important and range between 0.25, for women aged between 25 and 64, and 0.37 for those aged between 15 and 24. These values mean that in average, an increase of 1% of employment will lead to a rise of about 0.3 point of activity rate of population in working-age, that is to say to an increase of about 0.35% of the labour supply. These results indicate, in other words, that, in average, if 3 new jobs are created, it will reduce the number of unemployed of 2. For reservation wage, or replacement revenue, the third column of Table I.6 shows that the growth of these revenues represents a discouragement to offer its labour on the market, for the major part of population in working-age, and for men aged over 64. The importance of this effect is nevertheless quite small, but generally significant with parameter values that range between -0.0136 and -0.0285, for women and men aged 25

5For information, Table I.6 shows also adjusted R-squared, that exhibit values superior to 0.97, and Durbin-Watson statistics that do not reveal serious autocorrelation of error terms, with values close from 2 generally

82 CHAPTER I. THE CORE ECONOMIC MODEL to 64 respectively. Finally, as expected, the share of young aged between 15 and 24 engaged in tertiary education reduces very signiﬁcantly, and mechanically, the activity rate of this population category, for both men and women, while the share of working –age population in population aged 15 and over reduces also signiﬁcantly, and mechanically, the labour supply by population at (or close from) retirement age.

I.6.3 Calibration of labour supply

This last sub-section describes the calibration of labour supply behaviour for the diﬀerent population categories that the model NEMESIS distinguishes, presented in 0?? The expressions of activity rates that were used are identical to those presented above, with as main explanatory variables the variation of real wages, the variation of social beneﬁts and the variation in employment over two periods:

S,A S,A S,A S S,A A S,A S,A T xi,t = βw · GWi,t + βL · GLi,t + βSB · GSBi,t + βST · SSTi,t (I.62) σS,A·time+ρS,A e( i i ) + βS,A · SPOP S,A + (I.63) POP i,t σS,A·time+ρS,A 1 + e( i i )

As we did not dispose of data by skill to perform the econometric estimations, we then used, for the different age-groups, the same value of elasticity parameters for the two skill categories. The value of elasticity parameters were retrieved from the estimation results presented in Table I.6. Also, for the age groups of NEMESIS 25-54 and 55-64, the same elasticity parameters were used, that correspond to the econometric estimations realized for the global category 25-64. For wages, the elasticity parameter estimated for women aged between 25 and 64 was not significative, and had the wrong sign, and we consequently used for women the elasticity parameter estimated for men. For the categories aged between 15 and 24, we kept the assumption, revealed by econometric estimations, that changes in real wages are not a determinant of activity rates. For social benefits, the parameters estimated where globally significative, or close from significativity level, and we kept the values estimated, that are quite homogenous between population categories. For employment also, the estimated values, are always significative and homogenous

83 I.6. LABOUR SUPPLY between population categories were kept. For the youngest categories, aged between 15 and 24, we retained also as explanatory variable the share of students, with the elasticity parameters that were estimated. For women, an increase of one point of school enrolment ratio, decreases of 0.94 point the participation rate, and of 0.77 point for men. The impact is less than proportional for the reason that students may cumulate studies with a professional activity. The participation rates for the population aged 65 and over, that present very low values, and for which the econometric estimation dir not provided good results, were not modelled, and were kept them exogenous at this stage in NEMESIS. The other determinant of the participation rates are finally the logistic trends, tra- ducing the influence of other factors, exogenous in NEMESIS. For illustration of our calibration procedure, the Table I.7 sums-up the elasticities of activity rates with respect to one point change in their explanatory variables. For wages, social benefits and employment, the values represent the percentage change of activity rate, that is to say of labour supply, of the category, with respect to a 1% increase in the growth rate of the explanatory variables. For the share of students and the categories aged between 15 and 24, the values are semi-elasticities measuring the percentage change of activity rates with respect to a one point variation of school enrolment ratio of the category. One can see on Table I.7 that even if we use identical parameters for low skilled and high skilled populations, the values of elasticities are different for the two categories, and that they differ also between countries. For a given age and gender category, the elasticities of labour supply are inversely proportional to the initial value of the activity rates. There are consequently in every country stronger for women, who have activity rates inferior to men, and for low skilled persons, that have participation rates inferior to high skilled ones. We’ll have also superior values of elasticities for the age categories 15-24 and 54-65, the youngest and eldest populations having inferior activity rates than population aged between 25 and 54. For wages, an increase of 1% in labour remuneration will rise between 0.07% and 0.09% the labour supply of population aged between 25 and 54, which represents the major part of the labour force, with few discrepancies between skills, genders and countries. In other words, a permanent increase of 1% in the growth rate of real wage will increase by less that 0.1% the labour supply of the medium aged population, confirming the very weak impact of wages onto the labour supply behaviour. There is no impact at all for population aged between 15 and 24. For the eldest category, the impact of wages range between 0.09% for high skilled men and 0.25% for low skilled women, and 0.13% for the

84 CHAPTER I. THE CORE ECONOMIC MODEL

Table I.7.: Elasticities of activity rates in NEMESIS in 2008 Female Male Low High Low High Wages: 15-24 ------25-54 0.09 0.07 0.07 0.07 (CZ: 0.07; PT: 0.09) (RO: 0.07; BE: 0.08) (IE: 0.07; LU: 0.08) (GR: 0.06; FI: 0.08) 55-64 0.25 0.13 0.13 0.09 (SK: 0.11; AT: 0.35) (UK: 0.08; AT: 0.15) (EE: 0.09; LT: 0.18) (UK: 0.07; LU: 0.11) Social benefits: 15-24 -0.08 -0.04 -0.04 -0.03 (CZ: -0.04; LU: -0.17) (SK: -0.03; LT: -0.06) (CZ: -0.02; LV: -0.06) (SK: -0.02; FR: -0.06) 25-54 -0.02 -0.02 -0.03 -0.03 (CZ: -0.02; PT: -0.02) (RO: -0.02; BE: -0.02)) (IE: -0.03; LU: -0.04) (GR: -0.03; FI: -0.03) 55-64 -0.05 -0.03 -0.06 -0.04 (SK: -0.02; AT: -0.08) (UK: -0.03; AT: -0.08) (EE: -0.04; LT: -0.08) (UK: -0.03; LU: -0.05) Employment: 15-24 1.12 0.52 0.74 0.49 (CZ: 0.5; LU: 2.29) (SK: 0.4; LT: 0.89) (CZ: 0.42; LV: 1.37 (SK: 0.34; FR: 1.05) 25-54 0.31 0.25 0.30 0.28 (CZ: 0.29; PT: 0.44) (RO: 0.26; BE: 0.31)) (IE: 0.32; LU: 0.38) (GR: 0.31; FI: 0.33) 55-64 1.00 0.50 0.60 0.43 (SK: 0.45; AT: 1.41) (UK: 0.36; AT: 0.42) (EE: 0.43; LT: 0.95) (UK: 0.35; LU: 0.54) Share students 15-24 -2.87 -1.33 -1.80 -1.20 (CZ: -1.27; LU: -5.88) (SK: -1.03; LT: -2.29) (CZ: -1.01; LV: -2.57) (SK: -0.82; FR: -2.57) Values are non weighted European averages In brackets: minimum and maximum values

85 I.6. LABOUR SUPPLY two other categories. The strongest impact is found in Austria for high skilled women: 0.35%.

For social beneﬁts, the impacts are again weaker than for wages. A permanent increase of 1% of the growth rate of social beneﬁts will reduce labour supply between 0.02% for high skilled women aged between 25 and 54 and 0.08% for low skilled women aged between 15 and 24.

The main endogenous determinant of participation rates to labour market are finally jobs creation. A 1% increase of jobs creations raises labour supply of 0.25% for high skilled women aged between 25 and 54 and of 1.12% for low skilled women aged between 15 and 24. This last figure underlines the difficulty of reducing unemployment for young women with low school attainment level and for which an increase of 1% of job opportunities could provoke an augmentation more than proportional of their labour supply in the short term. It it nevertheless not the case for every country, the impact depending on the initial participation rate to labour market of this category of women. The smallest value of the flexion coefficient of labour supply to job opportutnities is found for Czech Republic, with 0.5 and the highest value for Luxembourg where it reaches 2.29. We find also high values of flexion coefficients for low skilled women aged between 55 and 64, with 1 in average, a minimum value of 0.45 in Slovakia and a maximum value of 1.41 in Austria. For the greater population category, aged between 25 and 54, the value is 0.28 in average, with few differences between skill and gender groups and between countries.

For the youngest population category, aged between 15 and 24, one can see also than school enrolment ratios, than can be endogenized on expenditures in eduction in NEMESIS, could inﬂuence very importantly the activity rates, and therefore the unemployment rates of both women and men, and of both low skilled and high skilled persons. In Europe, in average, an increase of 1 point of the school enrolment ratio of low skilled women reduces their labour supply of 2.87%, with a minimum of 1.27% in Czech republic and a maximum of 5.88% in Luxembourg. We have the more limited impacts of changes in school enrolment ratios on activity rates for high skilled men, with an elasticity of 1.2% in Europe in average, a minimum value of 0.82 in slovakia and a maximum value of 2.57 in France.

86 CHAPTER I. THE CORE ECONOMIC MODEL

Section I.7 Taxation and subsidies

I.7.1 Institutional sectors accounts

The main data source was the Eurostat database, completed if necessary by national sources (mainly for Luxemburg, Denmark and Norway). The data availability for some countries (Ireland, Luxemburg, Hungary, Malta,Slovenia and Romania) were too weak for building agents account for them, but main taxes and subsidies were integrated. All data and the sequence of accounts follows the European accounting framework ESA956. The different institutional sectors that are represented are the General Government(GG), Households and Non-Profit Institutions Serving Households (HNPISH), Financial Cor- porations (FC), Non-Financial Corporations (NFC), all of which are of course linked to the sectoral nomenclature of the model. The split of households and NPISH’s was not possible for most countries, so it had been decided not to separate them for the moment, this will be done as soon as data will be available. This huge database has been checked (agregations, paid/received...) completely and corrected if errors were encountered. Agents accounts are implemented from the production account up to the Acquisition of non financial assets account (i.e. up to the b9 Net lending (+) /net borrowing (-)).

I.7.2 Public ﬁnances

The main taxes and subsidies considered are see [126] for more information:

Taxes on production and imports (D.2)

• Taxes on products (D.21)

6ESA: European System of Accounts

87 I.7. TAXATION AND SUBSIDIES

– value added type taxes (D.211) – Taxes and duties on imports excluding VAT (D.212) – Taxes on products, except VAT and import taxes (D.214) – Excises duties and consumption taxes (D.214a)

∗ Mineral oil ∗ Alcoholic beverage ∗ Tobacco ∗ Electricity ∗ Non alcoholic beverages ∗ ...

– Other taxes on products (D214-D214a)

• Other taxes on production (D.29)

Subsidies (D.3)

• Subsidies on products (D.31)

• Other subsidies on production (D.39)

Current taxes on income, wealth, etc. (D.5)

• Taxes on income (D.51)

• Other current taxes (D.59)

Social Contributions (D.61)

• Actual social contributions (D.611)

– Employer’s actual social contributions (D.6111) – Employees’ social contributions (D.6112)

88 CHAPTER I. THE CORE ECONOMIC MODEL

– Social contributions by self and non-employed persons (D.6113)

• Imputed social contributions (D.612)

Capital Transfers (D.9)

• Capital Taxes (D.91)

• Investment Grants (D.92)

I.7.3 Focus on most important taxations system

We will focus here on the main important taxation part of the model. The main diﬃculties in sectoral applied modelling is to apply the right taxation rate and/or subsidy to the right sector. For a part of the taxation system, some information are availlable in the EUROSTAT datasets, while for others some assumptions had been made.

Value added type taxes (D.211)

The VAT is probably the most difficult tax to be implemented in a model such as NEMESIS. Firstly if we consider the datasets needed the available information on VAT particularities is often too detailed for modeling as on one side the sectoral disaggregation of the model allow a strict differentiation of the different VAT rates applicable to the different products and services, but on the other side, calculating VAT rates applicable to one sector based on actual rates is fastidious and need numerous assumptions concerning the sharing of each rate in the same sector, taking into account existing exemptions and therefore complicate more the linking of the sectoral taxation system up to the macro-econoomic one. Secondly, considering the formalisation in itself, the traditional framework in applied modeling for integrating VAT, is to calculate implicit tax rates for each sector, with the drawback that for analysing the consequences of the modification of one VAT rate, the implicit rate has to be recalculated ex-ante with all errors that it may imply. In the NEMESIS model the implicit VAT rate is fully modelised flowing from the actual VAT rates up to the product/sector implicit rate. The implicit rate is thus the

89 I.7. TAXATION AND SUBSIDIES result of linear combination of the different rates (0 rate, super reduced rate, reduced rate 1, reduced rate 2, normal rate and the parking rate) and of the different shares of each products in the consumption. The main information sources for building the data neede are furnished by the commission and the taxation and customs DG. For calculating final consumption VAT rates as precisely as possible, the most dis- agregated data of the COICOP nomenclature were used. Starting from this the VAT bloc is composed of three components:

1. The actual VAT rates series for the period 1980-2007, for all european countries.

2. The share for each COICOP three digit category of the diﬀerent rates applied

3. Finaly coeﬃcients allowing to ﬂow from the COICOP three digit nomenclature to the NEMESIS one

Hence our ﬁnal implicit VAT rate is the linear combination of these three datasets (the exemple shown below is for the NEMESIS Medical Care category):

medcar X pmed T V AIMP = shpmedc · αT,c · T T X econs + sheconsc · αT,c · T T X hosp + shhospc · αT,c · T T

Avec:

• shpmed, sheconset shhosp, respectively the share of the COICOP three digit «medical products and apparel», «external consultation???» et «hospital services» categories in the Medical care category of NEMESIS

• T = T 0,TSR,TR1,TR2,TN,TP , the diﬀernet existing rates, 0 rate, super reduced rate, reduced rate(s) (sometimes two rates), normal rate and the parking rate.

pmed • αT,c , the share of the pmed category to which we apply the rate T in country c.

Taxes on products, except VAT and import taxes (D.214)

90 CHAPTER I. THE CORE ECONOMIC MODEL

These taxes were splited into two broad taxes, Excises duties an consumption taxes (D214a) on one part, and other taxes on products (D214-D214a) on the other part. The distinction between the diﬀerent excises duties and there allocation between the sectors were made possible using DG taxation and custom Union informations (see [122]) , the same database was used for allocating the rest of Taxes on products, except VAT and import taxes. Hence, aside the three main Excises duties (alcoholic beverages, tobacco and mineral oil), some countries have other excises duties (electricity, non alcoholic beverage...), all of which had been incorporated in the model.

Social contributions

Employers’ social contribution (D6111) are splited into sectors using the sectoral data on D11 wage and salaries and D1 Compensation of employees that are available on Eurostat, employees’ social contribution (D6112) as well as imputed social contribution (D612) are splited between sectors depending on relative compensation of employees as no other data were available, while Social contributions by self and non-employed persons (D6113) are calculated only at the macroeconomic level, the figures I.22 and I.23 sums up the functiuning of the social contribution bloc. Then each type of social contribution is allocated to institutional sectors account ( Gov: general governement, FC: financial corporations, NFC, Non Financial corporations, H&NPISH: Households and non profit institutions serving households) through fixed shares.

Figure I.22.: Social Contribution paid

91 I.8. SECTORAL INTERDEPENDENCIES

Figure I.23.: Social contribution received

Section I.8 Sectoral Interdependencies

In sectorally detailled models, macroeconomic dynamics is driven by sectoral ones, the mix of the 32 sectors evolutions will descibe the strength and weaknesses of each european economy modeled, and hence describe their respective macoeconomic results in terms of economic growth, employment,etc.... Therefore, interlinkages between sectors are thus an important part of the model scheme as they will reﬂect the diﬀerent sectoral tendencies either in the short/medium term or in the long term.

I.8.1 Demand ﬂows to products

Each sector, in order to produce a certain quantity of its product (supposed to be homogenous), needs production factors: the five factors described in NEMESIS are employment, intermediate energy demands, final energy demands, materials demands and investments. A sixth factor could be added, even if it is not directly treatened as a pure production factor, this is the research and development expenditures. In the NEMESIS model, sectoral interdependencies are handled through energy demands (intermediate and final), materials demands, investment demands, and through R&D rent and knowledge spillovers (that will be explained separately). Each of this factor

92 CHAPTER I. THE CORE ECONOMIC MODEL demands is addressed to one or more sectors, other sectors but also the demanding sector (reflecting the intra-branch cousumption). These interactions are presented in figure I.24 below. These interactions between the sectors are threatened in two ways in the NEMESIS model depending on the simulations runs term. In the short/medium term, one can consider that substitutions between products are rather weak, as input substitution often requires changes in the production process (employees’ formation, capital structure,...) thus the coefficients of the different matrices are considered to be fixed and the demands are formulated as:

j j DEMC,i = βc,i · F ACT DC,i (I.64)

With this formulation a sector can not shift from a product to an other. If the sector that produces product j improve its productivity (that is produce the same product but with a lower price), every sectors i that uses the product j will face a lower investment price (ceteris paribus). By using¯fixed coefficient matrices, this will only leed for the i to a smaller global investment price, but this sector can not choose to buy more of the j’s good instead of other ones. Consequently, we can easily see that the j has no gain to make TFP in order to lower its price. Theoretically, if the i’s sector lowers its price, this must lead to improve its market share. The f¯ixed coefficient matrices are therefore not compatible with the developments proposed. Consequently, in order to keep the global theoretical coherency of the model, we have to endogeneise these coefficients. We choose to endogeneise the share of each product j in the total factor demand of sector i as a cost minimisation on a CES function. Firms determine their global factor demand using its production function, then minimises the cost of approvisioning this global demand from a C.E.S function of elasticity of substitution i. The substitution elasticity of that C.E.S function ought to be sufficiently slack to not to conduct to too sharp fluctuations for technical coefficients. In order to lowers as more as possible quick shifting between the different sources of supply, we add in this derived shares adjustment delays, the formulation we choose for each type of firms matrices (Intermediate Consumption, final Energy demand and Investment) is the following:

!i j j PDC,i j COEFC,i = λi · coefmatC,i · j + (1 − λi) · coefmatC,i (I.65) PVC with

93 I.8. SECTORAL INTERDEPENDENCIES

Figure I.24.: Sectoral interdependencies in NEMESIS

94 CHAPTER I. THE CORE ECONOMIC MODEL

• λi the adjustment delay,

• PDC,i the global factor demand price of sector i

j • PVC the sale price of the product j and

• i the price elasticity.

Using this formulation, if the sector j decrease its price (other sectors unchanged) the share of demand of the sector i asked to sector j will increase depending on the price elasticity and the adjustment delay. The price elasticities can not be estimated and was selected from other studies between 0.05 and 0.1, and the delays taken between 4 years and 10 years. Of course, endogeneising all these coeﬃcient increases dramatically the number of equations of the model (around 60 000 equation added). Some remarks must be here formulated:

1. The adoption of an optimisation procedure for the choice of these coeﬃcients, grounded on a re-agregation function of a C.E.S. type, allows to easily explicitate the products components of investment, intermediate consumption and energy sectoral demands and is moreover fully coherent with the framework we choose for closing- up the NEMESIS supply side to grounded microeconomic behaviour. Nev- ertheless, coeﬃcients so calculated are not those determined by national accounts statisticians.

2. But in the baseline projections, coeﬃcients evolution must be exogeneised, the endogenous determination of thousands of coeﬃcients complicate the model resolution.

This formalisation of matrices’ coeﬃcients had been tested using several economic and environmental policies and is operational.

I.8.2 technological progress interactions.

Endogenous technical change in NEMESIS needs to takes into account technologicals interactions between sector. T.C needs three kind of interactions: knowledge spillover, rent spillover and technology ﬂows.

95 I.8. SECTORAL INTERDEPENDENCIES

Knowledge spillovers

Knowledge spillover represent the case that one sector could beneﬁt from R&D activities of another sector without pay monetary compensation. An example of knowledge spillovers is when one invention might lead to a new ideas for diﬀerent inventor. In Nemesis model knowledge spillover relies positively R&D expenditure of one sector to the knowledge stock of another. We distinguish national and international knowledge spillover.

Knows,c,t = f(RDs,c,t0 ,SKNs,c,t00 ,SKIs,c,t00 ,RD/Y )

• s:sector, c: country and t: time

• Know is the knowledge stock of sector

• RD is the R&D expenditure of sector

• SKN is the national Knowledge spillovers

• SKI is the international Knowledge spillovers

Measure of knowledge spillover follow the seminal work of Jaffe [194] and Verspagen [317] which develop methods to take into account non-incorporated or disembodied R&D spillovers. This concept of technological link is called technological proximity because it is derived from the relative position of sector in a technological space. Concretely technological proximity matrix assume that the main IPC code into which a patent is classified provides a good proxy of the producing sector of the knowledge, and the listed supplementary IPC codes given an indication for technology spillovers to other industrial sectors. The more two sector are close to each other, the higher is the effect of R&D expenditure. For national knowledge spillover we assume:

X SKNi,c,t = θij · R&Dj,c,t0 j6=i For international knowledge spillover we assume :

XX SKIi,c,t = βcdθij · R&Dj,d,t0 · d j6=i

96 CHAPTER I. THE CORE ECONOMIC MODEL

Figure I.25.: Knowledge spillovers

where θijis the technology proximity between sector i and j and βcdis the economy distance between c and d Matrix used is Verspagen matrix transformed to Nemesis sectorial classiﬁcation.

Rent Spillovers

Second kind of technological interaction is rent spillovers. Rent spillovers refer to the case where R&D intensive input are purshased from other industries at less than their fully adjusted price. This failure to embody correctly a higher quality into output price is the consequence of imperfectly monopolistic pricing arising from competitive pressure on innovating industry. For Griliches [159] rent spillover is a problem of measuring capital equipement, materials and their price correctly and not a case of pure knowledge spillovers. If innovation are sold at prices that entirely reflect quality improvment i.e. on hedonic price index, problem does not arise. In Nemesis model, we assume that prices do not reflect totaly quality improvment. Importance of rent spillovers relatively to the adjustment of price will depend on the degree of competition. Low degree induce more importance on rent spillovers effect than on price adjsutment. In Nemesis model, rent spillover originate exclusively from economic transaction. We assume that rent spillover diffuse proportionaly to the level of intermediate input flows between sectors. This level is simply measured by Input-Output matrices. It reslut that factor productivity is not only affected by its own R&D but also by productivity improvment in another sector to the extend of its purchase.

97 I.9. HOUSING INVESTMENTS

Figure I.26.: Rent Spillovers

X RentSi,t = δijInnovj,c,t0 j6=i

Where δij is the I-O matrix coeﬃcient and Innov is product innovation of sector j.

Technology ﬂows

Nemesis model allows some innovation to be produce in one sector and implemented in another. Innovation which improve productivity could be made in own sector or purchase to another throught patent transaction. To link sector innovation in user- producer principle we use the so-called “Yale matrices”. This matrix is constructed on the basis of data from the Canadian patent Oﬃce. This last (exclusively in the world) assigns principal user and producing sectors to each patent. We use matrices made by Johnson [197] and extent it to other country.

Section I.9 housing investments

I.9.1 Methodology7

98 CHAPTER I. THE CORE ECONOMIC MODEL

Households investment was already modelled in the NEMESIS model but in a very roughly way, and the implementation of the land use module implies a better modelling of it. We will present in this section the new formalisation and estimate of households investment in the NEMESIS model. Either at theoretical or empirical point of view, interactions between residential market and macroeconomic are not very analysed (Leung, 2004[230]), this explains that the modelling of households investment in large applied economic model is not very developed, or at least is not highlighted compared to others macroeconomic variables. This fact is reinforced by the lack of consensus regarding households investments formalisation, mainly due to national regular regimes but also because of real estate bubbles (Baghli et al., 2004[23]). Furthermore, there are two aspects on the housing investments. The first one is associated with the services provides by the housing which can be view as a consumption and the second one concerns the wealth effect related with the own- ership of housing8. We analyse how some large applied economic models and especially econometrics ones represent the housing investments9. In the INTERLINK model developed by the OECD (Richardson 1988[275]), Egebo and Lienert (1988[117]) estimate housing stocks for six main OECD countries with a stock adjustment model. In their modelling, the variation of the housing stock is a function of the households real disposable income per capita, the real interest rates (in moving average), the housing relative price (either relative price of housing services or relative price of housing investment), the existing housing stock per capita at the previous period, the variation of unemployment rate and finally a partial adjustment term10. For the MIMOSA model11, Chauffour and Fourmann (1990[59]) formalise the investment rate (i.e. the ratio between housing investment and housing stock) as a function of households income per capita (smoothed variable), real housing investment price (smoothed variable), real interest rate, previous housing stock per capita and unemployment rate change. A very similar version of housing investment model is developed for the french economy (Bonnet et al. 1994[33]) in the AMADEUS model (INSEE 1998[185]), they also estimate the investment rate but they replace households income per capita by

7The section depends for a part on a study realised in the ERASME laboratory (Lécina, 2008[224]) and especially for the literature survey. 8We do not treat the trade-oﬀ between buying a housing or renting it (see e.g. Arrondel and Lefebvre 2001[16] or Henderson and Ioannides 1983[173]). Furthermore, from macroeconomic point of view, the housing investment only concerns new housing purchase, the second hand market is not considered even if it follows the new housing purchase market, as demonstrated by Demers (2005[98]) for Canada. 9We do not present the estimate results of these studies but we will use them to compare our estimate results in the following section. 10All variables are expressed in logarithm expect for real interest rate and unemployment rate. 11See Delessy et al. 1996[97] for a description of the MIMOSA model.

99 I.9. HOUSING INVESTMENTS gross households income and unemployment change by employment variation. Another interesting model is the European HERMES model (1993[191]) composed by seven individual macro-sectoral models for Belgium, Netherlands, France, Germany, Ire- land, Italy and United-Kingdom. Only four models introduce the housing investment in their respective presentation but in one of them, the Dutch model (Mot et al. 1993[247]), housing investment is described as exogenous. According to the French HERMES model (Assouline and Epaulard 1993[20]), housing investment is described as a model based on the desired housing stock adjusted with the help of an error correction model (Engle and Granger 1987[119]). This desired housing stock depends on households income, relative price, active population and interest rates. In the Irish HERMES model (Bradley et al. 1993[39]), the housing investment is modelled by the housing investment per capita that is a linear function of real per capita personal disposable income, government transfers for housing, interest rates and inflation12. Finally, Bosi et al. (1993[35]) use similar housing investment model for the Italian HERMES model. They relate, in logarithmic form, housing investment per capita with income per capita and relative price of housing investment and they include also a dynamic partial adjustment. More recently, Chauvin et al. (2002[60]) develop a error correction model for the Emod.fr model in which they model housing investment rates (the ratio between housing investment and households real disposable income) with real disposable income, unemployment rate and interest rate as explanatory variables. In this case, the use of housing investment rates, in a error correction model, imposes a long term elasticity between households investment and real disposable income equals to one13. Finally, a very recent description of the MESANGE model (see Klein and Simon 2010[214]) housing investment is also modelled with an error correction model in which they link for short term, housing investment variation with previous variation, housing investment price variation, real short term interest rate variation (3 month) and unemployment rate variation while in the long term equation, they only keep the link between real disposable income and real long term interest rate (10 years)14. As one can see in the previous descriptions, there are few differences in the explanatory variable used to describe households investment in the models, even if the endogenous variable are slightly different (investments rates, investment in level and stock of housing, ...). This can be summarised as follows:

• Firstly, the real disposable income allows taking into account the purchase ability

12All variables are expressed in logarithm. 13We will specify these properties in the following section. 14All variables are expressed in logarithm except interest rates and unemployment rate.

100 CHAPTER I. THE CORE ECONOMIC MODEL

as well as the borrowing power of households.

• The purchase of housing, in most case, requires a long term loan. This aspect is included with the help of interest rates, the payback power being reduce when interest rates increase.

• In addition, socio-economic aspect can also act on the housing investment, and particularly, the demography can play a non negligible role, it is why some models use per capita variables as explanatory variables.

• The relative housing investment price, generally the ratio between housing investment price and consumption price, allow the modelling of the traditional substitution effect. But, in the case of housing, which is a asset for the households, the investment price acts also on the expected wealth - insomuch as it follows housing stock price - and this wealth effect can be effectively very important as illustrated by the recent real estate bubble. Thus, housing investment price plays a double role, it increases purchase cost but it also raises housing expected value.

• The general economic context is generally represented through unemployment rate or employment which are relatively important for the households expectations on economic futures and then for their conﬁdence on their payback capacities.

• Finally, other variables such as government subsidies for housing, like in Irish HERMES model (Bradley et al. 1993[39]), can act on households investment decision. Some of them are already included in the real disposable income, this is the case for instance of transfers. It could also be interesting to include speciﬁc variables that could reﬂect change in national regular regimes but as we use a panel of 12 countries, the time required to get good and reliable information constrains us to exclude this option.

According to the modelling, we can see that the most recent studies (Chauvin et al. 2002[60] and Klein and Simon 2010[214]) use the error correction model that we also choose for NEMESIS because error correction model allow the distinction between two models: one for short term and a second for long term (equilibrium). Nevertheless, the error correcting model requires a deep examination of variables with numerous econometric time series tests, and especially, it requires unit roots and cointegration tests. Thus, we present in the following sections the data used for the modelling, the unit roots and cointegration tests, following by the error correction model estimate and ﬁnally we display some sensibility analysis on the estimated model.

101 I.9. HOUSING INVESTMENTS

I.9.2 The data

All economic data used for the estimate come from the Annual Macro-ECOnomic database (AMECO 2008[13]) of the European Commission’s Directorate General for Economic and Financial Aﬀairs (DG ECFIN) which provides structured and coherent data on national account and especially times series for prices. Population data come from Eurostat Population database (Eurostat 2008[130]). Thus we have the following variables for 12 European countries15 from 1995 to 2008:

• Households and Non-Profit Organisation (NPO) real gross fixed capital formation16 (GF CF ) which is the GF CF in value divided by the total economy gross fixed 17 capital formation price (PGF CF ) ,

R • The real total economy gross ﬁxed capital formation price (PGF CF ) which is the ratio between PGF CF and the consumption price (PCONS)

• The households real disposable income (REV Q) which is the ratio between households disposable income and consumption price,

• The long term and short term real interest rates (TXLT and TXCT ) which are the ratio between interest rates and consumption price,

• The number of unemployed persons (UNEMP ),

• And the population divided in 5 age groups, the “very young” (POP YY ) between 0 and 4 years, the “young” (POP Y ) between 0 and 19 years old, the “medium” (POP M ) between 20 and 39 years, the “medium-old” (POP MO) between 40 and 59 years old and the “old” (POP O) more than 60 years.

All these variables are transformed in logarithm except for the real interest rates. We present in the following section the unit root tests and cointegration tests realised on these variables.

15Only EU-15 countries: Belgium, Denmark, Germany, Spain, France, Italy, Netherlands, Austria, Portugal, Finland, Sweden and United-Kingdom. 16Households gross fixed capital formation and households and NPO gross fixed capital formation in residential are unavailable. 17Price of households and NPO gross fixed capital formation is unavailable.

102 CHAPTER I. THE CORE ECONOMIC MODEL

I.9.3 Model estimate and results

We estimate a error correction model (Engle and Granger, 1987[119]) with the coin- tegrated variables presented above. Thus, we have two models, one for long term relationship and a second for short term relationship. The long term relationship is deﬁned by the general model I.66 whereas short term equation is deﬁned by general model I.67.

R revQ Q pgfcf R pop gfcfi,t = αi + δiti + βi revi,t + βi pgfcf,i,t + βi popi,t (I.66) unemp TXLT LT + βi unempi,t + βi TXi,t + εi,t

R revQ Q pgfcf R pop 4gfcfi,t = µi + θi 4revi,t + θi 4pgfcf,i,t + θi 4popi,t (I.67) unemp TXLT LT res + θi 4unempi,t + θi 4TXi,t + θi εˆi,t−1 + ui,t

As, we use panel data, we impose identical parameters for each country for long term and short term equations (see equation I.68) except for intercept (αi and µi) and trend

(δi). εˆ are estimated residuals from long term relationship.

βU = βU ∀i i (I.68) Z Z θi = θ ∀i

Q R LT Q R LT Where U = rev , pgfcf , pop, unemp, T X and Z = rev , pgfcf , pop, unemp, T X , εˆ. We also estimate the models I.66 and I.67, either keeping free the long term relationship between households gross fixed capital formation and households real disposable Q pR income (βrev ) and gross fixed capital formation real price (β gfcf ) or constraining these relationships. Table I.8 displays the estimated results of both models, with and without constrained parameters. Looking at the unconstrained models, we can see that parameters of long term model are all significantly different to zero except for population. The long term elasticities of households gross fixed capital formation (GF CF ) with respect to households real disposable income (REV Q) is equal to 0.52. If this elasticity can appear relatively good, this result supposes a progressive decrease of the ratio between households investment in level and their income in level i.e. the share of households investment in their budget tends to decline. Thus, we can not keep this results, and we must impose, as in the studies

103 I.9. HOUSING INVESTMENTS presented above (for instance in Chauvin et al. 2002[60]), the long term relationship for households real income. In addition, the elasticity of households gross fixed capital formation with respect to R households gross fixed capital formation real price (Pgfcf ) is estimated to 1.47. This positive and superior to unity value of price elasticity can be quite surprising, however, as we mentioned above, housing is a spending for households but it is also an asset, as a consequence, an increase of investment price increases purchase cost but raises also the anticipated value of the asset. Thus, as our data cover the 1995 to 2008 period, it includes the recent real estates bubbles that occurs in most of the EU-15 countries, and an elasticity of 1.47 can traduce the bubble effect of the housing price. Nevertheless, the introduction of such a parameter value in a economic model such as NEMESIS would lead to misleading results, therefore we constrained this elasticity at -0.5%, value in adequacy with those estimated in the literature. For instance, Egebo and Lienert (1988[117]) find elasticities of -0.45% for France, -0.56% for United Kingdom and -0.44 for Italy while Chauffour and Fourmann (1990[59]) find elasticities about -0.4% for France, -0.4% for Italy and -0.3% for West Germany. More recent studies, imposed an elasticity or exclude the real price of housing investment of their models in order to avoid such results. In the constrained models, all parameters are significantly different to zero, at least at 10% level except for long term interest rate in the short run relationship and population in the long run one. Regarding the effect of the long term interest rate, the null parameter seems not so surprising, and even using the short term interest rate does not provide better parameters estimates, consequently we keep the hypothesis that long term interest rate has no impact at short term. The parameter estimates for the population appears to be strong in the short run (+3%)18, but does not influence households investment in the long run. The unemployed persons elasticity is negatively related to housing investment with a short term elasticity of -0.28% and a long term elasticity a little bit lower with -0.13%. For households investment prices, the positive short term elasticity (1.2%) represent the bubble effect where households anticipate the increase of the value of their assets, while in the long run the more traditional behaviour dominates and explains the negative parameter value (-0.5%). Finally, an increase of 1% of households real disposable income raises the households gross fixed capital formation about 0.66% at short term and 1% at long term. We tried to individualise some coefficients either in long term or short term model, but due to our limited sample (168 obs.) and the increasing number of parameters, the

18We will limit the short term eﬀect at 1.5% in the implemented version of housing investment in NEMESIS to keep a certain stability even at short term.

104 CHAPTER I. THE CORE ECONOMIC MODEL

Table I.8.: Estimates results of households gross ﬁxed capital formation error correction model

Model I.66 (Long Term) Model I.67 (Short Term)

Unconstrained Constrained Unconstrained Constrained

βrevQ 0.5196∗∗∗ 1(a) θrevQ 0.7575∗∗∗ 0.6633∗∗

(0.3574) – (0.2896) (0.2809)

pR pR β gfcf 1.4731∗∗∗ -0.5(a) θ gfcf 1.6091∗∗∗ 1.2401∗∗∗

(0.3465) – (0.3707) (0.3523)

βpop -0.2841 -0.4065 θpop 3.0925∗∗ 3.2721∗∗∗

(0.5338) (0.5736) (1.2115) (1.1696)

βunemp -0.1398∗∗∗ -0.1309∗∗∗ θunemp -0.2502∗∗∗ -0.2849∗∗∗

(0.0475) (0.0411) (0.0502) (0.0494)

βTXLT -0.0178∗∗∗ -0.0143∗ θTXLT -0.0074 -0.0067

(0.0068) (0.0074) (0.005) (0.0048)

θres -0.5418∗∗∗ -0.5302∗∗∗

(0.0882) (0.0739) ∗, ∗∗, ∗ ∗ ∗: parameter significantly different to zero at 10%, 5%, 1% respectively. (a): fixed parameters. results are globally disappointing, few coefficients being significant. The only parameter providing relatively good results when individualised, it is the adjustment parameter res (θi ), which estimates are given for short term equation in Table I.9. We observe that the other coefficients are close to their value with common adjustment parameters. The short term elasticity is a little bit lower for real disposable income and unemployed persons, stronger for population and still not significant for long term interest rate. Now looking at the individualised adjustment coefficients, we first see that all are negative. But the coefficients for Germany, France, Italy, Austria and Portugal are not significant at 10% level. And we can also see that the range of significant coefficients is confined between a minimum of -0.5 in Belgium and a maximum of -0.77 in Denmark. To analyse the effect of adjustment parameters as well as the effects of short and long

105 I.9. HOUSING INVESTMENTS

Table I.9.: Estimates results for short term model with individualised adjustment coef- ﬁcients

res Individualised adjustment parameter: θi

BE -0.5005∗∗

DK -0.7661∗∗∗

DE -0.2594 Fixed parameters ES -0.7418∗∗ θrevQ 0.523∗ FR -0.3308 pR θ gfcf 1.2497∗∗∗ IT -0.3787 θpop 4.0173∗∗∗ NL -0.5545∗ θunemp -0.228∗∗∗ AT -0.1028 θTXLT -0.006 PT -0.4417

FI -0.5214∗∗

SE -0.7269∗∗∗

UK -0.6318∗∗∗ ∗, ∗∗, ∗ ∗ ∗: parameter signiﬁcantly diﬀerent to zero at 10%, 5%, 1% respectively. term parameters, we realise a sensibility analysis by introducing standard stocks in the error correcting model.

I.9.4 Sensibility analysis

We make a sensibility analysis of housing investment error correction model estimated in the previous section by introducing shocks on one variable and keeping the other ones ﬁxed. Figure I.27 and I.28 display the model responses to a permanent shock of 1% on each variable with the exception of long term interest rates that had been raised by 1 point of percentage permanently. Figure I.27 presents the responses for common adjustment parameters (-0.53) whereas ﬁgure I.28 compares results with individualised adjustment parameter of Denmark (-0.77) and Austria (-0.10) with common parameters

106 CHAPTER I. THE CORE ECONOMIC MODEL case for 1% shock on real disposable income.

Figure I.27.: Sensibility analysis with common adjustment coeﬃcient

3.5%

3.0%

2.5%

2.0%

1.5%

1.0%

0.5%

0.0%

-0.5% 12345678910 Rev P Tx U Pop

Q R LT With Rev = rev , P = Pgfcf , T x = TX , U = unemp and P op = pop.

We can see in figure I.27, that, in accordance with estimate results, the short term effect of population is very strong but decreases progressively to reach zero at long term. Thus, population rise has only a transitory effect on households gross fixed capital formation. At the opposite, the real disposable income shows a moderate short run effect on housing investment, the first year, households investment increases of about 0.6% and tends to 1% (as imposed) at long term. Furthermore, the real price of housing investment has a particular dynamic. In a first time, an raise of 1% of investment price pushes housing investment up to 1.25% what can be view as a transitory bubble effect. And in a second time, the short term positive effect declines to reach its long term equilibrium of -0.5% (as constrained). Thus a perpetual shock on housing investment real price plays, at short term, as a bubble effect but this bubble effect progressively declines to finally reduces housing investment. Looking at unemployment effect, we observe a bigger short term shock (-0.3%) than the long term with -0.13%. Finally, as demonstrated by estimated parameters, short term effect of the long term interest rates

107 I.9. HOUSING INVESTMENTS is null, and long term eﬀect starts one year after the introduction of the shock, to reach -0.15% at long term.

Figure I.28.: Model response to 1% shock on households real disposable income: Com- parison according to adjustment coeﬃcients

1.1%

1.0%

0.9% 0.8%

0.7%

0.6%

0.5%

0.4%

0.3% 0.2%

0.1%

0.0% 12345678910 Denmark Austria Common

Figure I.28 shows the effect of different adjustment parameter on the response dynamic. Denmark, where the adjustment parameter is the higher with -0.77, tends more rapidly to its long term equilibrium, therefore the Danish average adjustment decay is about 0.3 years i.e. 50% of the adjustment to the long term equilibrium is done in 4 months. At the opposite, the average adjustment is about 9 years for Austria, where the adjustment coefficient equals -0.1 thereby, full adjustment is not still realised at t + 10. For the common adjustment parameter (-0.53), the average adjustment decay is 0.9 year. We have displayed the responses of the error correction model for different variables and for different adjustment parameters and we showed their respective importance. We keep, for the implementation in the NEMESIS model, the estimated coefficients except for the short term effect of population. In fact, the estimate value of this parameter seems to strong and we decide to reduce it at 1.5%, i.e. a little bit higher than the unity and we still suppose that its long term elasticity is null. Similarly, we impose a null short term effect of long term interest rate and finally we take the individualised adjustment parameters for estimated countries and we use the common adjustment parameter for

108 CHAPTER I. THE CORE ECONOMIC MODEL not estimated European countries.

I.9.5 Concluding Remarks

We have now an endogenous model for households investments. This model is error correcting model that determine housing investment in each European countries according to its prices, households real disposable income, total population, unemployed persons and long term interest rate. With the housing investment for each European country, we can calculate their housing stock using perpetual inventory method. And ﬁnally, we can determine the national land used by housing by using the density coeﬃ- cients that link national housing stock with its land use.

109 Bibliography

[1] Global mitigation of non-co2 greenhouse gases. Technical report, United States Environmental Protection Agency, 2006.

[2] Technical guidelines voluntary reporting of greenhouse gases (1605(b)) program. Technical report, US Energy Information Administration, 2007.

[3] Greenhouse gas emission trends and projections in europe 2009. Technical report, European Environment Agency, 2009.

[4] Manning A. Wage bargaining and the phillips curve: The identiﬁcation and spec- iﬁcation of aggregate wage equation. The economic journal, 103, 1993.

[5] Daron Acemoglu. Directed technical change. Review of Economic Studies, 69:781– 809, 2002.

[6] Daron Acemoglu and Fabrizio Zilibotti. Productivity diﬀerences. Quarterly Jour- nal of Economics, 116:563–606, 2001.

[7] P. Aghion and P. Howitt. "A Model of Growth Through Creative Destruction". Econometrica, 60(2):323–351, 1992.

[8] P. Aghion and P. Howitt. Capital accumulation and innovations as complementary factors in long-run growth. Journal of Economic Growth, 3:111, 1998.

[9] P. Aghion and P. Howitt. Endogenous Growth Theory. MIT Press, Cambridge, 1998.

110 Bibliography

[10] AGRESTE - Ministère de l’alimentation, de l’agriculture et de la pêche. AGRESTE - Enquêtes sur la valeur vénale des terres agricoles, 2007.

[11] H. Akaike. A new look at the statistical model identiﬁcation. IEEE Transactions on Automatic Control, 19(6):716–723, 1974.

[12] W. Alonso. Location and Land Use. Harvard University Press, Cambridge, USA, 1964.

[13] AMECO. Annual Macro-ECOnomic database. Database, 2008.

[14] J.R. Anderson, E.E. Hardy, J.T. Roach, and R.E. Witmer. A Land Use and Land Cover Classiﬁcation System for Use with Remote Sensor Data. Geological Survey Professional Paper 964, United States Department of the Interior, 1976.

[15] P.. Armington. A Theory of Demand for Products Distinguished by Place of Production. Staﬀ paper, International Monetary Fund, 1969.

[16] L. Arrondel and B. Lefebvre. Consumption and Investment Motives in Housing Wealth Accumulation: A French Study. Journal of Urban Economics, 50(1):112– 137, 2001.

[17] K. Arrow. The economic implications of learning-by-doing. Review of Economic Studies, 26:155–173, 1962.

[18] K. J. Arrow, H. B. Chenery, B. S. Minhas, and R. M. Solow. Capital-labor substitution and economic eﬃciency. The Review of Economics and Statistics, 43(3):225– 250, 1961.

[19] K.J. Arrow, H.B. Chenery, B.S. Minhas, and R.M. Solow. Capital-Labor Substitu- tion and Economic Eﬃciency. The Review of Economics and Statistics, 43(3):225– 250, Aug. 1961.

[20] M. Assouline and A. Epaulard. HERMES: A Macrosectoral Model for the French Economy, chapter 6, pages 205–276. In Commission of the European Communities [191], 1993.

[21] T.O. Awokuse and J.M. Duke. The Causal Structure of Land Price Determinants. Canadian Journal of Agricultural Economics/Revue canadienne d’agroeconomie, 54(2):227–245, 06 2006.

111 Bibliography

[22] P. Bagnoli. Developments in growth literature and its relevance for simulation models : annex 1. Technical report, Working Party on Global and Structural policies, OCDE, ENV/EPOC/GSP(2002)7/ANN1, 2002.

[23] M. Bahgli, V. Brunhes-Lesage, O. De Brandt, H. Fraisse, and J.P. Villetelle. MAS- COTTE : Modèle d’AnalySe et de prévision de la COnjoncture TrimesTriElle. Notes d’Etudes et de Recherche 106, Banque de France, Fev. 2004.

[24] R. Baldwin and R. Forslid. Dynamic Issues in Applied Commercial Policy Analysis, chapter Putting growth eﬀects in computable equilibrium trade models, pages 44– 85. Cambridge University Press., 1999.

[25] Almut Balleer, Ramon Gomez-Salvador, and Jarkko Turunen. Labour force participation in the euro area - a cohort based analysis. Working Paper Series 1049, European Central Bank, May 2009.

[26] R. J. Barro and X. Sala-I-Martin. Economic Growth. MCGraw-Hill, inc, 1995.

[27] C. Bean. Endogenous growth and the procyclical behaviour of productivity. Eu- ropean Economic Review, 34:355–363, 1990.

[28] Gary S. Becker and Barry R. Chiswick. Education and the distribution of earnings. The American Economic Review, 56(1/2):358–369, 1966.

[29] Jess Benhabib and Mark Spiegel. Human capital and technology diﬀusion. Pro- ceedings, (Nov), 2002.

[30] E.R. Berndt and L.R. Christensen.

[31] Ernst R. Berndt and Jack E. Triplett. Fifty Years of Economic Measurement: The Jubilee of the Conference on Research in Income and Wealth, volume 24. The National Bureau of Economic Research, The University of Chicago Press, Chicago, 1991.

[32] Olivier Jean Blanchard and Lawrence Katz. Wage dynamics: Reconciling theory and evidence. NBER Working Papers 6924, National Bureau of Economic Research, Inc, February 1999.

[33] X. Bonnet, G. Forgeot, and Maire P. L’investissement logement des ménages dans le modèle AMADEUS. Notes de Conjonctures, INSEE - Institut National de la Statistique et des Etudes Economiques, Dec. 1994.

112 Bibliography

[34] Valentina Bosetti, Carlo Carraro, Marzio Galeotti, Emanuele Massetti, and Mas- simo Tavoni. A world induced technical change hybrid model. The Energy Journal, 0(Special I):13–38, 2006.

[35] P. Bosi, R. Golinelli, and A. Stagni. HERMES: A Macrosectoral Model for the Italian Economy, chapter 9, pages 453–531. In Commission of the European Com- munities [191], 1993.

[36] A.F. Bouwman, T. Kram, and K.K. Goldewijk. Integrated Modelling of Global Environmental Change. Netherlands Environmental Assessment Agency, 2006.

[37] R. Boyer and M. Didier. Innovation et croissance. Rapport du conseil d’analyse Économique, La documentation Française, 1998.

[38] I. Bracke and E. Meyermans. Speciﬁcation and estimation of an allocation system for private consumption in europe. Working-Paper 3-97, Federal Planing Bureau, Belgium., 1997.

[39] J. Bradley, J. Fitz Gerald, D. Hurley, L. O’Sullivan, and A. Storey. HERMES: A Macrosectoral Model for the Irish Economy, chapter 8, pages 327–452. In Com- mission of the European Communities [191], 1993.

[40] D. Brécard, C. Chevalier, A. Fougeyrollas, P. Le Mouël, L. Lemiale, and P. Zagamé. "3 l’aide du modèle Némésis". Rapport, Commission Européene, 2004.

[41] D. Brécard, A. Fougeyrollas, P. Le Mouël, L. Lemiale, and P. Zagamé. Evalua- tion pour la France des conséquences de l’augmentation de l’eﬀort de R&D. In Incidences économiques d?une augmentation des dépenses de recherche en Europe, number 391 in Rapport d’information du Sénat, 2004.

[42] D. Brécard, F. Fougeyrollas, P. Le Mouël, L. Lemiale, and P. Zagamé. Macro- economic consequences of European research policy: Prospects of the Nemesis model in the year 2030. Research Policy, 35(7):910–924, September 2006.

[43] Dorothee Brecard, Arnaud Fougeyrollas, Pierre Le Mouel, Lionel Lemiale, and Paul Zagame. Macro-economic consequences of European research policy: Prospects of the Nemesis model in the year 2030. Research Policy, 35:910–924, 2006.

[44] Paolo Buonanno, Carlo Carraro, and Marzio Galeotti. Endogenous induced technical change and the costs of Kyoto. Resource and Energy Economics, 25:11–34, 2003.

113 Bibliography

[45] J.-M. Burniaux and H.-L. Lee. Modelling Land Use Changes in GTAP. Presented at the 6th Annual Conference on Global Economic Analysis, The Hague, The Netherlands, 2003.

[46] J-M Burniaux and T. Truong. GTAP-E: An Energy-Environmental Version of the GTAP Model. GTAP Technical Papers 16, Center for Global Trade Analysis, Department of Agricultural Economics, Purdue University, 2002.

[47] O.R. Burt. Econometric Modeling of the Capitalization Formula for Farmland Prices. American Journal of Agricultural Economics, 68(1):10–26, 1986.

[48] Ricardo J. Caballero and Adam B. Jaﬀe. How high are the giants’ shoulders: An empirical assessment of knowledge spillovers and creative destruction in a model of economic growth. NBER Working Papers 4370, National Bureau of Economic Research, Inc, May 1993.

[49] G. Cameron. Innovation and growth: a survey of the empirical evidence. Nuﬃeld College, Working Paper, 1998.

[50] J.B. Campbell. Spatial correlation-eﬀects upon accuracy of supervised classiﬁca- tion of land cover. Photogrametric Engineering And Remote Sensing, 47(3):355– 363, 1981.

[51] D.R. Capozza and R.W. Helsley. The Fundamentals of Land Prices and Urban Growth. Journal of Urban Economics, 26(3):295–306, Nov. 1989.

[52] D.R. Capozza and R.W. Helsley. The Stochastic City. Journal of Urban Eco- nomics, 28(2):187–203, Sep. 1990.

[53] C. Carraro. Climate modelling and policy strategies. the role of technical change and uncertainty. Energy Economics, 20:463–471, 1998.

[54] C. Carraro and M. Galeotti. Economic growth, international competitiveness and environmental protection: R&d and innovation strategies with the warm model. Energy Economics, 9:2.28, 1997.

[55] Alan Carruth and Andrew Henley. Housing assets and consumer spending: A regional analysis. Regional Studies: The Journal of the Regional Studies Association, 27(11):611–621, 1993.

[56] J. Cavailhès and P. Wavresky. Urban Inﬂuences on Periurban Farmland Prices. European Review of Agricultural Economics, 30(3):333–357, Sept. 2003.

114 Bibliography

[57] Odile Chagny, Frédéric Reynès, and Henri Sterdyniak. Le taux de chômage d’équilibre. discussion théorique et évaluation empirique. Revue de l’OFCE, 81(2):205–244, 2002.

[58] Reynes F. Chagny O. and Sterdyniak H. Le taux de chomage d’équilibre: Discus- sion théorique et évaluation empirique. Revue de l’OFCE, 81, 2002.

[59] J.P. Chauﬀour and E. Fourmann. L’Investissement-Logement des Ménages dans MIMOSA. Document de Travail 90-01, CEPII (Centre d’Etudes Prospectives et d’Informations Internationales) and OFCE (Observatoire Français des Conjonc- tures Economiques), Jan. 1990.

[60] V. Chauvin, G. Dupont, H. Heyer, M. Plane, and X. Timbeau. Le modèle France de l’OFCE - La Nouvelle Verion: emod.fr. Revue de l’OFCE, 81(2):245–300, 2002.

[61] Valérie Chauvin, Gaël Dupont, Éric Heyer, Mathieu Plane, and Xavier Timbeau. Le modèle france de l’ofce. la nouvelle version : e-mod.fr. Revue de l’OFCE, 81(2):245–300, 2002.

[62] J-P. Chavas and A. Thomas. A Dynamic Analysis of Land Prices. American Journal of Agricultural Economics, 81(4):772–784, Nov. 1999.

[63] Nathalie Chen. Intra-national versus international trade in the european union: why do national borders matter. Journal of Internaional Economics, 63(1):93–118, May 2004.

[64] Carole Chevallier, Arnaud Fougeyrollas, Pierre Le Mouël, and Paul Zagamé. A time to sow, a time to reap for the European countries : a macro-econometric glance at the RTD national action plans. Revue de l’OFCE, 97bis:235–257, 2006.

[65] D.L. Chicoine. Farmland Values at the Urban Fringe: An Analysis of Sale Prices. Land Economics, 57(3):353–362, Aug. 1981.

[66] I. Choi. Unit root tests for panel data. Manuscript, Kookmin University, Corée, 1999.

[67] I. Choi. Unit root tests for panel data. Journal of International Money and Finance, 20(2):249–272, 2001.

[68] W. Christaller. Die zentralen Orte in Süddeutschland. Gustav Fisher, Jena, Ger- many, 1933.

115 Bibliography

[69] L. Christensen, D. Jorgenson, and L. Lau. Conjugate duality and the transcendental logarithmic functions. Econometrica, 39:255–256, 1971.

[70] L. Christensen, D. Jorgenson, and L. Lau. Transcendental logarithmic production frontiers. Review of Economics and Statistics, 55:28–45, 1973.

[71] L.R. Christensen and D.W. Jorgenson. The Measurement of U.S. Real Capital Input, 1929-1967. Review of Income and Wealth, 15(4):293–320, Dec. 1969.

[72] L.R. Christensen and D.W. Jorgenson. Measuring Economic Performance in the Private Sector. In The Measurement of Economic and Social Performance, NBER Chapters, pages 233–352. National Bureau of Economic Research, Inc, 1973.

[73] L.R. Christensen, D.W. Jorgenson, and L.J. Lau. Transcendental Logarithmic Production Frontiers. The Review of Economics and Statistics, 55(1):28–45, Feb. 1973.

[74] M. Cincera and B. van Pottelsberghe. International r&d spillovers: a survey. Cahiers Economiques de Bruxelles, 2001.

[75] Leon Clarke, John Weyant, and Alicia Birky. On the sources of technological change: Assessing the evidence. Energy Economics, 28:579–595, 2006.

[76] Leon Clarke, John Weyant, and Jae Edmonds. On the sources of technological change: What do the models assume. Energy Economics, 30:409–424, 2008.

[77] Wesley M Cohen and Daniel A Levinthal. Innovation and learning: The two faces of r&d. Economic Journal, 99(397):569–96, September 1989.

[78] European Commission. Vat rates applied in the member states of the european community, situation at september 2006. Document DOC/1829/2006, Taxation and custom Union, 2006.

[79] European Commission. « une politique de l’énergie pour l’europe », communication de la commission au conseil européen et au parlement européen. COM(2007) 1 Final, Janvier 2007.

[80] European Commission. Vat rates applied in the member states of the european community, situation at 1st may 2007. Document DOC/2137/2007, Taxation and custom Union, 2007.

116 Bibliography

[81] European Commission. Directive of the european parliament and of the council, on the promotion of the use of energy from renewable sources. COM(2008) 19 ﬁnal, 01 2008.

[82] Juan Carlos Conesa, Carlos Díaz-Moreno, and José Enrique Galdón-Sánchez. Ex- plaining cross-country diﬀerences in participation rates and aggregate ﬂuctuations. Journal of Economic Dynamics and Control, 26(2):333 – 345, 2002.

[83] GEM-E3 Consortium. Gem-e3 model manual. Technical report, E3M-Lab, July 2006.

[84] C. Cornwell, P. Schmidt, and D. Wyhowski. Simultaneous equations and panel data. Journal of Econometrics, 51:151–181, 1992.

[85] Giancarlo Corsetti, Philippe Martin, and Paolo Pesenti. Productivity, terms of trade and the home market eﬀect. Journal of International Economics, Forthcom- ing, 2007.

[86] Matthieu Crozet and Hélène Erkel Rousse. Trade performances, product quality perceptions and the estimation of trade price-elasticities. Review of International Economics, 12(1):108–129, 2004.

[87] Matthieu Crozet and Federico Trionfetti. Trade costs and the home market eﬀect. CEPII Working Paper N° 2007-05, March 2007.

[88] Joanne Cutler and Kenny Turnbull. A disaggregated approach to modelling uk labour force participation. Discussion Papers 04, Monetary Policy Committee Unit, Bank of England, 2001.

[89] P. Dasgupta and E Maskin. The simple economics of research portfolios. The Economic Journal, 97:581–595, 1987.

[90] P. David. Clio and the economics of qwerty. American Economic Review, 75:332– 337, 1985.

[91] P. A. David, B. H. Hall, and A. A. Toole. Is public r&d complement or substitute or private r&d? a review of the econometric evidence. Research Policy, 29:497–529, 2000.

[92] J. Davidson, D. Hendry, F. Srba, and S. Yeo. Econometric modelling of the aggregate time-series relationship between consumers’ expenditure and income in the united-kingdom. The Economic Journal, (88):661–692, 1978.

117 Bibliography

[93] P. de Fontnouvelle and S.H. Lence. Transaction Costs and the Present Value. Southern Economic Journal, 68:549–565, 2002.

[94] Angel de la Fuente and Rafael Doménech. Human capital in growth regressions: How much diﬀerence does data quality make? an update and further results. CEPR Discussion Papers 3587, C.E.P.R. Discussion Papers, October 2002.

[95] B. Van Pottelsberghe de la potterie. Les politiques de science et technologie et l’objectif de barcelone. Université Libre de Bruxelles Working Paper WP-CEB 04/11, 2004.

[96] A. Deaton and J. Muellbauer. An almost ideal demand system. American Eco- nomic Review, 70(3):312–326, 1980.

[97] H. Delessy, H. Sterdyniak, B. Coquet, P. Cour, H. Le Bihan, F. Lerais, F. Busson, J. Baude, B. Châtelain, and M. Fiole. La nouvelle version de MIMOSA, modèle de l’économie mondiale. Revue de l’OFCE, 58:103–155, July 1996.

[98] F. Demers. Modelling and Forecasting Housing Investment: The Case of Canada. Working Papers 05-41, Bank of Canada, 2005.

[99] F.A.G. den Butter and F.J. Wollmer. An empirical model for endogenous technology in the Netherlands. Economic Modelling, 13:15–40, 1996.

[100] A. Di Gregorio and L.J.M. Jansen. Land Cover Classiﬁcation System (LCCS): Classiﬁcation concepts and user manual. FAo Corporate Document Repository, SDRN (Environment and Natural Resources Service) - FAO (Food and Agriculture Organization of the United Nations), 2000.

[101] X. Diao and T. Roe. Implications of the new growth theory to agricultural trade research and trade policy, chapter A dynamic CGE model of R&D based growth in the US economy: an experiment using the new growth theory. 1997.

[102] X. Diao, T. Roe, and E. Yeldan. Strategic policies and growth : an applied model of r&d driven endogenous growth. Journal of Development Economics, 60:343–380, 1999.

[103] Xinshen Diao, Terry Roe, and Erinc Yeldan. Strategic policies and growth: an applied model of R&D-driven endogenous growth. Journal of Development Eco- nomics, 60:343–380, 1999.

118 Bibliography

[104] D.A. Dickey and A.W. Fuller. Distribution of the Estimators for Autoregressive Time Series With a Unit Root. Journal of the American Statistical Association, 74(366):427–431, Jun. 1979.

[105] D.A. Dickey and A.W. Fuller. Likelihood Ratio Statistics for Autoregressive Time Series with a Unit Root. Econometrica, 49(4):1057–1072, Jul. 1981.

[106] W. Diewert. An application of the shephard duality theorem: A generalized leontief production function. Journal of Political Economy, 79:481–507, 1971.

[107] W. Diewert. Frontiers of Quantitative Economics, chapter Application of duality theory vol. II, pages 106–171. North Holland, Amsterdam, 1974.

[108] W. Diewert. Handbook of Mathematical Economics, chapter Duality approaches to microeconomic theory, pages 535–599. North Holland, Amsterdam, 1982.

[109] W.E. Diewert. Functional forms for proﬁt and transformation functions. Journal of Economic Theory, 6(3):284–316, June 1973.

[110] W.E. Diewert. Applications of Duality Theory, chapter 3, pages 106–171. Volume II of D. A. Kendrick and M. D. Intriligator [187], 1974.

[111] W.E. Diewert. Intertemporal Consumer Theory and the Demand for Durables. Econometrica, 42(3):497–516, May 1974.

[112] W.E. Diewert and T.J. Wales. Flexible functional forms and global curvature conditions. Econometrica, 55:43–68, 1987.

[113] E. Duguet. Les subventions à la recherché et développement : complement ou substitut au ﬁnancement privé ? Etude pour le Ministère de l’industrie (DIGITIP- SESSI), 2002.

[114] EC - European Commission. FADN - Farm Accounting Data Network, 2009.

[115] EEA - European Environmetal Agency. Corine Land Cover 2000 - Mapping a decade of Change, 2004.

[116] EEA - European Environmetal Agency. Urban sprawl in Europe: The ignored challenge. Report 10/2006, EEA, 2006.

[117] T. Egebo and I. Lienert. Modelling Housing Investment for Seven Major OECD Countries. OECD Economics Department Working Papers 63, OECD, Economics Department, Dec. 1988.

119 Bibliography

[118] W. Enders. RATS Programming Manual. Department of Economics, Finance & Legal Studies, University of Alabama, US, 2003.

[119] R.F. Engle and C.W.J. Granger. Co-Integration and Error Correction: Represen- tation, Estimation, and Testing. Econometrica, 55(2):251–276, March 1987.

[120] T. Engsted. Do Farmland Prices Reﬂect Rationally Expected Future Rents? Ap- plied Economics Letters, 5(2):75–79, February 1998.

[121] ESPON Project 2.3.2. Governance of Territorial and Urban Policies from EU to Local Level. Final report, ESPON Project, 2006.

[122] European Commission, taxation and custom Union. Taxes in europe database.

[123] Commission Européenne. Gouvernance européenne : Un livre blanc. Technical Report COM(2001) 428 ﬁnal, Commission Européenne, 2001.

[124] Commission Européenne. Mieux légiférer à l’aide d’analyses d’impact de meilleure qualité. SEC(2007) 926, 2007.

[125] EuroRAPE - European Road Assessment Programme. European Road Assess- ment Programme: Pilot phase Technical Report. Pilot phase technical report, EuroRAPE, 2003.

[126] Eurostat. European system of accounts ESA 1995.

[127] Eurostat. Trends in households in the European Union: 1995-2025. Statistics in Focus 24/2003, Eurostat - Statistical Oﬃce of the European Communities, 2003.

[128] Eurostat. Common Questionnaire on Transport Statistics. Database, September 2004.

[129] Eurostat. Land Use Statistics. Database, September 2004.

[130] Eurostat. Population. Database, July 2008.

[131] Eurostat. Selling Prices of Agricultural Products (absolute prices), Land prices and Rents. Database, 2008.

[132] Eurostat. Annual National Accounts. Database, February 2009.

[133] Eurostat. Structure of Agricultural Holdings. Database, July 2009.

120 Bibliography

[134] Reynes F. La formation des salaires et le concept de taux de chomage d’equilibre: Une tentative de réconciliation des courbes Wage Setting et de Phillips. PhD thesis, Institut d’Etudes Politiques de Paris, 2006.

[135] D. Mc Fadden. Production Economics: A Dual Approach to Theory and Appli- cation, chapter Cost, revenue and proﬁt functions, VOL I, pages 3–109. North Holland, New York, 1978.

[136] B. Falk. Formally testing the present value model of farmland prices. American Journal of Agricultural Economics, 73:1–10, 1991.

[137] B. Falk and B.S. Lee. Fads versus Fundamentals in Farmland Prices. American Journal of Agricultural Economics, 80(4):696 – 707, 1998.

[138] F.A.O. - Food and Agriculture Organization. Global Forest Ressources Assessment 2000. FAO Forestry Paper 140, FAO, 2001.

[139] F.A.O. - Food and Agriculture Organization. FAOSTAT - Land Database, 2008.

[140] A.M. Featherstone and T.G. Baker. An Examination of Farm Sector Real Asset Dynamics: 1910-85. American Journal of Agricultural Economics, 69(3):532–546, August 1987.

[141] Franklin M Fisher, Robert M Solow, and James M Kearl. Aggregate production functions: Some ces experiments. Review of Economic Studies, 44(2):305–20, June 1977.

[142] P. Fisher, A.J. Comber, and R. Wadsworth. Land Use and Land Cover: Contra- dition or Complement, chapter 6, pages 85–98. In Fisher and Unwin [143], 2005.

[143] P. Fisher and D. Unwin. Re-Presenting GIS. John Wiley & Sons, Ltd, 2005.

[144] R.A. Fisher. Statistical Methods for Research Workers. Oliver and Boyd, Edinburg, 1932.

[145] Karen Fisher-Vanden and Ian Sue Wing. Accounting for quality: Issues with modeling the impact of R&D on economic growth and carbon emissions in developing economies. Energy Economics, (forthcoming).

[146] A. Fougeyrollas, G. Koléda, and P. Le Mouël. Quelques tests des propriétés des modèles de croissance endogène avec micro-économie de l’innovation sur données françaises,. contribution au 2ème programme cnrs « les enjeux économique de l’innovation », CNRS, 2000.

121 Bibliography

[147] A. Fougeyrollas, P. Le Mouël, and P. Zagamé. Endogenous growth and environmental regulation: the case of the kyoto protocol. In 7th congress of the Society of Computational Economics, Yale University, June 28-30 2001.

[148] A. Fougeyrollas, P. Le Mouël, and P. Zagamé. Economy wide employment impacts resulting from the implementation of the kyoto protocol, a simulation using the nemesis model. OECD, ENV/EPOC/WPNET(2003)11/FINAL, 2003.

[149] A. Fougeyrollas, P. Le Mouël, and P. Zagamé. Croissance endogène et politique structurelle : le cas des politiques de lutte contre le changement climatique et des politiques de r&d. Revue Economique, 56(5), Septembre 2005.

[150] Arnaud Fougeyrollas, Pierre Le Mouël, and Paul Zagamé. Les nouvelles théories de la croissance en application : L’évaluation des politiques structurelles, le cas du protocole de Kyoto. Revue économique, 56(5):1089–1105, 2005.

[151] M. Fujita. Urban Economic Theory - Land use and city size. Cambridge University Press, 1989.

[152] Reyer Gerlagh and Bob van der Zwaan. Gross world product and consumption in a global warming model with endogenous technological change. Resource and Energy Economics, 25:35–57, 2003.

[153] A. Golub, T.W. Hertel, and B. Sohngen. Land Use Modeling in Recursively Dy- namic GTAP Framework, chapter 10, page 235. In Hertel et al. [176], 2009.

[154] A. Golub, W.T. Hertel, and H-L Lee. Modeling Land Supply and Demand in Long Run. In Presented at the Ninth Annual Conference on Global Economic Analysis, Addis Ababa, Ethipia, June 15- 17 2006.

[155] C. Gouel and T.W. Hertel. Introducing Forest Access Cost Functions into a General Equilibrium Model. GTAP Research Memorandum 8, Center for Global Trade Analysis, Department of Agricultural Economics, Purdue University, 2006.

[156] Lawrence H. Goulder and Koshy Mathai. Optimal CO2 Abatement in the Presence of Induced Technological Change, Journal of Environmental Economics and Man- agementVolume 39, Issue 1, , January 2000, Pages 1-38. Journal of Environmental Economics and Management, 39:1–38, 2000.

[157] Lawrence H. Goulder and Stephen H. Schneider. Induced technological change and the attractiveness of co2 abatement policies. Resource and Energy Economics, 21(3-4):211–253, August 1999.

122 Bibliography

[158] Lawrence H. Goulder and Stephen H. Schneider. Induced technological change and the attractiveness of CO2 abatement policies. Resource and Energy Economics, 21:211–253, 1999.

[159] Z. Griliches. The search for r&d spillovers. NBER, 1991.

[160] Z. Griliches. The search for r&d spillovers. Scandinavian Journal of Economics, 94:29–47, 1992.

[161] G. Grossman and E. Helpman. Innovation and Growth in the Global Economy. Cambridge: MIT Press, 1991.

[162] G.M. Grossman and E. Helpman. Quality ladders in the theory of growth. Review of Economic Studies, 58:43–61, 1991.

[163] M. Grubb. Economic dimensions of technological and global responses to the kyoto protocol. Journal of Econonic Studies, 27:111–125, 2000.

[164] Arnulf Grubler and Sabine Messner. Technological change and the timing of mitigation measures. Energy Economics, 20:495–512, 1998.

[165] D. Guellec and B. van Pottelsberg de la Potterie. Recherche-développement et croissance de la productivité : analyse des données d’un panel de 16 pays de l’ocde. Revue Économique de l’OCDE, 33:111–136, 2001.

[166] D. Guellec and B. van Pottelsberghe. The impact of public r&d expenditure on business r&d. Economic of Innovation and New Technology, 12(3):225–244, 2003.

[167] D. Guellec and B. van Pottelsberghe de la Potterie. R&D and Productivity Growth: Panel Data Analysis of 16 OECD Countries. OECD Science, Technology and Industry Working Papers 2001/3, OECD, Directorate for Science, Technology and Industry, June 2001.

[168] L. Gutierrez, J. Westerlund, and K. Erickson. Farmland Prices, Structural Breaks and Panel Data. European Review of Agricultural Economics, 34(2):161–179, June 2007.

[169] B.H. Hall and C. Cummins. TSP program for IPS Panel unit root test - panuit.tsp. TSP Program, May 1999.

[170] B.H. Hall and C. Cummins. TSP 5.1 - Reference Manual. TSP International, 2009.

123 Bibliography

[171] Juan Carlos Hallak. Product quality and the direction of trade. Journal of Inter- national Economics, 68(1):238–265, January 2006.

[172] K. Helming, M. Pérez-Soba, and P. Tabbush. Sustainability Impact Assessment of Land Use Changes. Springer, 2008.

[173] J.V. Henderson and Y.M. Ioannides. A Model of Housing Tenure Choice. American Economic Review, 73(1):98–113, March 1983.

[174] F. Henin. Les statuts du fermage en Europe - « Nous ne pouvons comparer que ce qui est comparable », Feb. 2010.

[175] T.W. Hertel. Global Trade Analysis: modelling and Applications. Cambridge University Press, 1997.

[176] T.W. Hertel, S.K. Rose, and R.S.Tol. Economic Analysis of Land Use in Global Climate Change Policy. Routledge, 2009.

[177] W. Hogan. A dynamic putty–semi-putty model of aggregate energy demand. En- ergy Economics, 11(1):53–69, January 1989.

[178] K. Hubacek and J.C.J.M. van den Bergh. The Role of Land in Economic Theory. Interim report IR-02-037, IIASA, May 2002.

[179] Peter Huber. Intra-national labor market adjustment in the candidate countries. Journal of Comparative Economics, 32(2):248 – 264, 2004.

[180] Barry Hughes. Direct income and substitution eﬀects in participation decisions. The Journal of Political Economy, 80(4):793–795, 1972.

[181] Charles R. Hulten. The Measurement of Capital, chapter 4, pages 119–158. Vol- ume 24 of Berndt and Triplett [31], Jan. 1991.

[182] C. Hurlin and V. Mignon. Une synthèse des tests de racine unitaire sur données de Panel. Economie & Prévision, 169(3-4-5):253–294, 2005.

[183] C. Hurlin and V. Mignon. Une synthèse des tests de cointégration sur données de Panel. Economie & Prévision, 180-181(4):241–265, 2007.

[184] K.S. Im, M.H. Pesaran, and Y. Shin. Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115(1):53–74, July 2003.

124 Bibliography

[185] INSEE - Institut National de la Statistique et des Etudes Economiques. Présenta- tion du modèle Annuel MAcroéconomique à DEUx Secteurs (AMADEUS). Série des documents de travail de la Direction des Etudes et Synthèses Économiques G 9801, INSEE, Fev. 1998.

[186] INSEE - Institut National de la Statistique et des Etudes Economiques. Ménages selon la structure familiale. Tableau Economique de la France (TEF), 2010 , INSEE, 2010.

[187] M.D. Intriligator and D.A. Kendrick. Frontiers of Quantitative Economics, volume 2. North-Holland Publishing Company, 1974.

[188] Ireland. The new industrial economics: recent developments in industrial organisation, oligopoly and game theory, chapter Product diﬀerentiations and quality, pages 84–106. 1992.

[189] W. Isard. Location and Space-economy: A General Theory Relating to Indus- trial Location, Market Areas, Land Use, Trade, and Urban Structure. MIT Press, Cambridge, 1956.

[190] K. Itakura and T. Hertel. A Note On Changes Since GTAP Book Model (Version 2.2a / GTAP94 ). Technical report, GTAP, 2000.

[191] A. Italianer, G. d’Alcantara, and P. Zagamé. HERMES: Harmonised Econometric Research for Modelling Economic Systems. North-Holland Publishing Company, 1993.

[192] Joyce P. Jacobsen. Labor force participation. The Quarterly Review of Economics and Finance, 39(5):597–610, 1999.

[193] Henry D. Jacoby, John M. Reilly, James R. McFarland, and Sergey Paltsev. Tech- nology and technical change in the MIT EPPA model. Energy Economics, 28:610– 631, 2006.

[194] A. Jaﬀe. Technological opportunities and the spillover of r&d. American Economic Review, 76:984–1001, 1986.

[195] T. Jansson, Bakker M., Boitier B., Fougeyrollas A., Helming J., van Meijl H., and Verkerk P. J. Cross Sector Land Use Modelling Framework, pages 159–180. In Helming et al. [172], 2008.

125 Bibliography

[196] L. Johansen. Substitution versus Fixed Production Coeﬃcients in the Theory of Economic Growth. Econometrica, 27(2):157–176, 1959.

[197] D. K. N. Johnson. The oecd technology concordance (otc): Patents by industry of manufactoring and sector of use. OECD STI Working Paper 2002/5, 2002.

[198] Charles I Jones. R&d-based models of economic growth. Journal of Political Economy, 103(4):759–84, August 1995.

[199] C.I Jones. R&d based models of economic growth. Journal of Political Economy, 103:759–784, 1995.

[200] C.I Jones. Times series tests of endogenous growth models. Quarterly Journal of Economics, 110:495–527, 1995.

[201] C.I. Jones. Théorie de la croissance endogène. De Boeck University, 2000.

[202] C.I. Jones and C. Williams. Measuring the social return to r&d. Quarterly Journal of Economics, 113:1119–1135, November 1998.

[203] D. Jorgenson. The embodiment hypothesis. Journal of Political Economy, 1(74):1– 17, 1966.

[204] D.W. Jorgenson. Capital as a Factor of Production, chapter 1, pages 1–35. In Jorgenson and Landau [205], 1989.

[205] D.W. Jorgenson and R. Landau. Technology and Capital Formation. MIT Press, Cambridge, MA, 1989.

[206] R.E. Just and J.A. Miranowski. Understading the Farmland Price Changes. Amer- ican Journal of Agricultural Economics, 75(1):156–168, Feb. 1993.

[207] N. Kaldor and J. Mirrlees. A new model of economic growth. Review of Economic Studies, 29:174–192, 1962.

[208] M. Katz and C.Shapiro. Systems competition and network eﬀects. Journal of Economic Perspectives, 8:93–115, 1994.

[209] M. Katz and C. Shapiro. Network externalities, competition and compatibility. American Economic Review, 75:424–440, 1985.

[210] M. Katz and C. Shapiro. Technology adoption in the presence of network externalities. Journal of Political Economy, 94:822–841, 1986.

126 Bibliography

[211] Claudia Kemfert. Estimated substitution elasticities of a nested ces production function approach for germany. Energy Economics, 20(3):249–264, June 1998.

[212] P. Kennedy. A Guide to Econometrics - Fifth Edition. The MIT Press, Cambridge, MA, 2003.

[213] J. Kim and L. Lau. The sources of economic growth of the east asian newly industrialized countries. Journal of the Japanese and International Economies, pages 235–271, 1994.

[214] C. Klein and O. Simon. Le modèle MÉSANGE nouvelle version réestimée en base 2000. Document de Travail de la DGTPE 10/02, DGTPE (Direction Générale du Trésor) - MINEFE (Ministère de l’Economie de l’Industrie et de l’Emploi), March 2010.

[215] Rainer Klump, Peter McAdam, and Alpo Willman. Factor substitution and factor- augmenting technical progress in the united states: A normalized supply-side system approach. The Review of Economics and Statistics, 89(1):183–192, 2007.

[216] N.R. Kocherlakota and K. Yi. Can convergence regressions distinguish between exogenous and endogenous growth models. Economic Letters, 49:211–215, 1996.

[217] S. Kortum. "Research, Patenting, and Technological Change". Econometrica, 65(6):1389–1419, November 1997.

[218] M. Kremer. Population growth and technological change: one million b.c. to 1990. Quarterly Journal of Economics, 108(3):681–716, 1993.

[219] P. Krugman. Increasing Returns and Economic Geography. The Journal of Polit- ical Economy, 99(3):483–499, June 1991.

[220] Socrates Kypreos. A MERGE Model with Endogenous Technological Change and the Cost of Carbon Stabilization. FEEM Working Paper No. 123.05, October 2005.

[221] E3M Lab. General equilibrium model for economy energy environment. Technical report, Institute of Computers and Communications Systems-National Technical University of Athens, 2008.

[222] L. Latruﬀe and C. Le Mouël. Capitalization Of Government Support In Agricul- tural Land Prices: What Do We Know? Journal of Economic Surveys, 23(4):659– 691, 09 2009.

127 Bibliography

[223] L. Lau. A characterization of the normalized restricted proﬁt function. Journal of Economic Theory, pages 131–163, 1976.

[224] O. Lécina. Une fonction d’investissement en logements des ménages pour le modèles NEMESIS. Master Thesis (1st year), Université de Paris 1, July 2008.

[225] P. Le Mouël, Zagamé P., Ortiz R.A., Nielsen T.S., Kaas B.C., and Hasler B. Modelling Multi-Sectoral Land Use in Macro-Economic Model for EU-27. In Pre- sentation at IALUC Conference Berlin, April 2008.

[226] H. Lee, R.F. Owen, and D. van der Mensbrugghe. Regional integration in Asia and its eﬀects on the EU and North America. Journal of Asian Economics, 20(3):240– 254, 2009.

[227] H.-L. Lee, T.W. Hertel, S. Rose, and M. Avetisyan. An Integrated Global Land Use Database for CGE Analysis of Climate Policy Options, chapter 4, page 72. In Hertel et al. [176], 2009.

[228] Jong-Wha Lee and Robert J. Barro. Schooling quality in a cross section of countries. NBER Working Papers 6198, National Bureau of Economic Research, Inc, September 1997.

[229] S.H. Lence and D.J. Miller. Transaction Costs and the Present Value Model of Farmland. American Journal of Agricultural Economics, 81:257–278, 1999.

[230] C. Leung. Macroeconomics and Housing: a review of the literature. Journal of Housing Economics, 13(4):249–267, Dec. 2004.

[231] G. Livanis, C.B. Moss, V.E. Breneman, and R.F. Nehring. Urban Sprawl and Farmland Prices. American Journal of Agricultural Economics, 88(4):915–929, Oct. 2006.

[232] Andreas Löschel. Technological change in economic models of environmental policy: a survey. Ecological Economics, 43:105–126, 2002.

[233] R.N. Lubowski. Determinants of Land-Use Transitions in the United States: Econometric Analysis of Changes among the Major Land-Use Categories. PhD thesis, Havard University, 2002.

[234] D.L. Mac Fadden. Cost, Revenue, and Proﬁt Functions. Department of Economics, University of California, Berkeley, mimeographed, 1970.

128 Bibliography

[235] D.B. Madan and I.R. Prucha. A note on the estimation of nonsymmetric dynamic factor demand models. Journal of Econometrics, 42(2):275–283, Octobre 1989.

[236] G.S. Maddala and S. Wu. A Comparative Study of Unit Root Tests with Panel Data and a New Simple Test. Oxford Bulletin of Economics and Statistics, 61(Spe- cial I):631–52, 1999.

[237] J. Mairesse and M. Sassenou. Recherche-développement et productivité, un panorama des études économétriques sur données d’entreprises. STI Revue, n°8, pp. 9-45, 1991.

[238] Alan Manning. Wage bargaining and the phillips curve: The identiﬁcation and speciﬁcation of aggregate wage equations. The Economic Journal, 103(416):98– 118, 1993.

[239] D.W. Marcouiller, J.G. Clendenning, and Kedzior R. Natural Amenity-led De- velopment and Rural Planning. Journal of Planning Literature, 16(4):515–542, 2002.

[240] A. MATHIS. Vat indicators. Working-Paper 2/2004, European Commission, DG Taxation and customs.

[241] E. McGrattan. A defense of ak growth models. Federal Reserve Bank of Min- neapolis Quarterly Review, 22:13–27, 1998.

[242] H. Meijers and A. van Zon. RUM, a Recursive Update Putty-Semi-Putty Vintage Production Model : Sectoral Estimation Results For Germany and the Nether- lands. MERIT Research Memoranda 1994-010, UNU-MERIT (United Nations University - Masstricht Economic and social Research and training center on In- novation and Technology), 1994.

[243] Huub Meijers and Adriaan Van Zon. Endogenous technological change in a putty- semi-putty vintage world. GEM-E3 Workin-Paper, 1999.

[244] Jacob Mincer. Labor force participation of married women. In Aspects of Labor Economics, NBER Chapters, pages 63–106. National Bureau of Economic Re- search, Inc, February 1962.

[245] P. Mohnen. R&d and productivity growth: a survey of the literature. Université du Québec, Cahier de recherche n.57, 1990.

129 Bibliography

[246] C.B. Moss and A.L. Katchova. Farmland Valuation and Asset Performance. Agri- cultural Finance Review, 65(2):119–130, July 2005.

[247] E.S. Mot, P.J. van den Noord, D.D. van der Stelt-Scheele, and M.A. Koning. HERMES: A Macrosectoral Model for the Duch Economy, chapter 5, pages 113– 204. In Commission of the European Communities [191], 1993.

[248] V. Mukerji. A generalized s.m.a.c. function with constant ratios of elasticity of substitution. The Review of Economic Studies, 30(3):233–236, 1963.

[249] M. I. Nadiri. Innovations and technological spillovers. NBER Working Paper, n°4423, 1993.

[250] M. I. Nadiri and I.R. Prucha. Dynamic factor demand models and productivity analysis. NBER Working Paper No. W7079, Avril 1999.

[251] Richard R. Nelson and Edmund S. Phelps. Investment in humans, technological diﬀusion, and economic growth. The American Economic Review, 56(1/2):69–75, 1966.

[252] A. Ngwa Zang. Modèle dynamique de demande de facteurs fondé sur la frontière stochastique de coût implicite intégrant l’innovation de produit : une application à l’analyse et mesure de la productivité du secteur agricole de l’Union européenne. PhD thesis, Ecole Centrale Paris, France, December 2008.

[253] A. Ngwa Zang and P. Le Mouël. Estimation of a dynamic system of factors inputs with multiple outputs for agriculture using a dynamic panel for 9 Eu countries. In Presented at Journées INRA-IFER de recherches en sciences sociales, Société Française d Economie Rurale, December 2007, 2007.

[254] W. D. Nordhaus. "Modeling Induced Innovation in Climate Change Policy". paper presented at the IIASA-Yale Conference on Induced Innovation, June 1999.

[255] Blanchard O. and Katz L.F. Wage dynamics: Reconciling theory and evidence. The American Economic Review, 89, 1989.

[256] OCDE. Une nouvelle économie ? transformation du rôle de l’innovation et des technologies de l’information dans la croissance. Technical report, Série "Société de l’information". Paris, 2000.

130 Bibliography

[257] OECD - Organisation for Economic Co-operation and Development. Measuring Capital - Measurement of Capital Stocks, Consumption of Fixed Capital and Cap- ital Services. OECD Manual, OECD, 2001.

[258] S.G. Pantula, G. Gonzalez-Farias, and W.A. Fuller. A Comparison of Unit-Root Test Criteria. Journal of Business & Economic Statistics, 12(4):449–59, October 1994.

[259] Chris Papageorgiou and Marianne Saam. Two-level ces production technology in the solow and diamond growth models. Departmental Working Papers 2005-07, Department of Economics, Louisiana State University, 2005.

[260] P. Pedroni. Critical Values for Cointegration Tests in Heterogeneous Panels with Multiple Regressors. Oxford Bulletin of Economics and Statistics, 61(Special Issue):653–70, 1999.

[261] P. Pedroni. Panel Cointegration: Asymptotic and Finite Sample Properties of Pooled Time Series Tests with an Application to the PPP Hypothesis. Econometric Theory, 20(3):597–625, 2004.

[262] P. Pedroni. WinRATS program for Pedroni Tests - pancoint.prg. WinRATS Pro- gram, July 2007.

[263] Patrick Perrier. La fonction de production et les données canadiennes. Working Papers 05-20, Bank of Canada, 2005.

[264] P.C.B. Phillips and P. Perron. Testing for a Unit Root in Time Series Regression. Biometrika, 75(2):335–346, June 1988.

[265] A.J. Plantinga, R.N. Lubowski, and R.N. Stavins. The eﬀects of potential land development on agricultural land prices. Journal of Urban Economics, 52(3):561– 581, November 2002.

[266] A.J. Plantinga and D.J. Miller. Agricultural Land Values and the Value of Rights to Future Land Development. Land Economics, 77(1):56–67, Feb. 2001.

[267] David Popp. ENTICE: endogenous technological change in the DICE model of global warming. Journal of Environmental Economics and Management, 48:742– 768, 2004.

[268] David Popp. ENTICE-BR: The eﬀects of backstop technology R&D on climate policy models. Energy Economics, 28:188–222, 2006.

131 Bibliography

[269] David Popp. Innovation in climate policy models: Implementing lessons from the economics of R&D. Energy Economics, 28:596–609, 2006.

[270] A.A. Powell and F.H.G. Gruen. The Constant Elasticity of Transformation Produc- tion Frontier and Linear Supply System. International Economic Review, 9(3):315– 328, October 1968.

[271] I.R. Prucha and M.I. Nadiri. A comparison of alternative methods for the estimation of dynamic factor demand models under non-static expectations. Journal of Econometrics, 33(1-2):187–211, 1986.

[272] P. Pyykkönen. Factors Aﬀecting Farmland Prices in Finland. Adacemic disserta- tion, Faculty of Agriculture and Forestry of the University of Helsinki, May 2006.

[273] D. Ricardo. On the Principles of Political Economy and Taxation. London: John Murray, 1817.

[274] Gordon R. Richards. Endogenous technological advance in an econometric model: implications for productivity and potential output in the United States. Economic Modelling, 17:13–34, 2000.

[275] P. Richardson. The Structure and Simulation Properties of OECD’s Interlink Model. OECD Economic Studies 10, OECD, 1988.

[276] Paul M Romer. Increasing returns and long-run growth. Journal of Political Economy, 94(5):1002–37, October 1986.

[277] Paul M Romer. Endogenous technological change. Journal of Political Economy, 98(5):S71–102, October 1990.

[278] Paul M Romer. The origins of endogenous growth. Journal of Economic Perspec- tives, 8(1):3–22, Winter 1994.

[279] Thomas F. Rutherford and David G. Tarr. Trade Liberalization and Endogenous Growth In a Small Open Economy: A Quantitative Assessment. FEEM Working Paper No. 60.98, May 1998.

[280] W. Salter. Productivity and Technical Change. Cambridge University Press., 1960.

[281] P. A. Samuelson. Prices of Factors and Good in General Equilibrium. The Review of Economic Studies, 21(1):1–20, 1953.

132 Bibliography

[282] Mathan Satchi. A note on adding-up restriction when modelling trade ﬂows. Eco- nomic Modelling, 21(6):999–2002, December 2004.

[283] K. Sato. A two-level constant-elasticity-of-substitution production function. The Review of Economic Studies, 34(2):201–218, 1967.

[284] F.M. Scherer. R&D, Patents and Productivity, chapter Using linked patent and R&D data to measure interindustry technology ﬂows. University of Chicago Press, 1984.

[285] U.A. Schneider, J. Balkovic, O. de Cara S.and Franklin, S. Fritz, P. Havlik, I. Huck, K. Jantke, A.M.I. Kallio, F. Klaxner, A. Moiseyev, M. Obersteiner, C. Ivie Ramos, C. Schleupner, E. Schmid, D. Schwab, and R. Skalsky. The European Forest and Agriculture Optimisation Model - EUFASOM. Working Papers FNU-156, Research Unit Sustainability and Global Change, Hamburg University, February 2008.

[286] P. Schreyer. User costs and bubbles in land markets. Journal of Housing Eco- nomics, 18(3):267–272, Sep. 2009.

[287] Theodore W. Schultz. Investment in human capital. The American Economic Review, 51(1):1–17, 1961.

[288] J.A. Sefton and J.W in’t Veld. Consumption and wealth: An international comparison. Manchester School, 1998.

[289] P. Segerstrom. "Endogenous Growth Without Scale Eﬀects". American Economic Review, (5):1290–1310, December 1998.

[290] P. Segerstrom. "The Long Run Growth Eﬀects of R&D Subsidies". Michigan State University and IUI Working Paper, 1998.

[291] Henry D. Jacoby Richard S. Eckaus James McFarland Marcus Saroﬁm Mal- colm Asadoorian Sergey Paltsev, John M. Reilly and Mustafa Babiker. The mit emissions prediction and policy analysis (eppa) model: version 4. Technical report, MIT Joint Program on the Science and Policy of Global Change Report No. 125., 2005.

[292] SETRA - Sercices d’Etudes Techniques des Routes et Autoroutes. Comprendre les principaux paramètres de conception géométrique des routes. Technical report, SETRA, January 2006.

133 Bibliography

[293] R. Shephard. Theory of Cost and Production Functions. Princeton University Press, 1953.

[294] Y.J. Shi, T.T. Phipps, and D. Colyer. Agricultural Land Values under Urbanizing Inﬂuences. Land Economics, 73(1):90–100, 1997.

[295] C. R. Shumway and A. A. Powell. A Critique of the Constant Elasticity of Transfor- mation (CET) Linear Supply System. Western Journal of Agricultural Economics, 9(2), December 1984.

[296] Sjak Smulders and Michiel de Nooij. The impact of energy conservation on technology and economic growth. Resource and Energy Economics, 25(1):59–79, February 2003.

[297] B. Sohngen and S. Brown. The inﬂuence of conversion of forest types on carbon sequestration and other ecosystem services in the South Central United States. Ecological Economics, 57(4):698–708, 2006.

[298] R. Solow. A contribution to the theory of economic growth. Quarterly Journal of Economics, 70:65–94, 1956.

[299] R. M Solow. Mathematical Methods in the Social Sciences, chapter Investment and Technical Progress., pages 89–104. Stanford University Press, Stanford CA, 1960.

[300] R. M. Solow, J. Tobin, C. C. von Weizsacker, and M. Yaari. Neoclassical growth with ﬁxed factor proportions. Review of Economic Studies, 33(2), 1966.

[301] T.W. Swan. Economic growth and capital accumu. Economic Record, 32:334–361, 1956.

[302] A. Tabeau, B. Eickhout, and H. von Meijl. Endogenous Agricultural Land Supply: Estimation and Implementation the GTAP model. In Presented at the Ninth Annual Conference on Global Economic Analysis, Addis Ababa, Ethipia, June 15- 17 2006.

[303] A. Tegene and F. Kuchler. A Regression Test of the Present Value Model of US Farmland Prices. Journal of Agricultural Economics, 44(1):135–143, 1993.

[304] J. Temple. Robustness test of the augmented solow model. Journal of Applied Econometrics, 13:361–375, 1998.

134 Bibliography

[305] J. Temple. New growth evidence. Journal of Economic Literature, 37(1):112–156, Mars 1999.

[306] N.E. Terleckyj. Eﬀects of r&d on the productivity growth of industries: an ex- ploratory study. Washington: National Planning Association, 1974.

[307] Peter Thompson and Timothy G Taylor. The capital-energy substitutability de- bate: A new look. The Review of Economics and Statistics, 77(3):565–69, August 1995.

[308] Marcel P. Timmer, Mary O’Mahony, and Bart van Ark. Eu klems growth and productivity accounts: An overview. International Productivity Monitor, 14:71– 85, Spring 2007.

[309] O’Mahony M. & Van Ark B. Timmer M. The eu klems growth and productivity accounts: An overview. University of Groningen & University of Birmingham, 2008.

[310] C.G. Turvey, K. Meilke, A. Weersink, S. Clark, R. Klein, and R. Sarker. Measuring the Capitalization of Income Transfer Programs into Agricultural Asset Values. Working paper 95-465, Agricultural Canada, Policy Branch, 1995.

[311] Peter A.G. van Bergeijk, Gilbert H.A. van Hagen, Ruud A. de Mooij, and Jarig van Sinderen. Endogenizing technological progress: The MESEMET Model. Economic Modelling, 14:341–367, 1997.

[312] D. van der Mensbrugghe. LINKAGE Technical Reference Document - Version 6.0. Technical report, The World Bank - Development Prospects Group (DECPG), 2005.

[313] B. C. C. van der Zwaan, R. Gerlagh, G. Klaassen, and L. Schrattenholzer. En- dogenous technological change in climate change modelling. Energy Economics, 24:1–19, 2002.

[314] H. van Meijl, T. van Rheenen, A. Tabeau, and B. Eickhout. The impact of diﬀerent policy environments on agricultural land use in Europe. Agriculture, Ecosystems & Environment, 114(1):21 – 38, 2006.

[315] van Pottelsberghe de la Potterie. The eﬃciency of science and technology policies inside the triad. PhD thesis, Université libre de Bruxelles, 1998.

135 Bibliography

[316] Jérôme Vandenbussche, Philippe Aghion, and Costas Meghir. Growth, distance to frontier and composition of human capital. Journal of Economic Growth, 11(2):97– 127, June 2006.

[317] Bart Verspagen. Measuring intersectoral technology spillovers: Estimates from the european and us patent oﬃce databases. Economic Systems Research, 9(1):47–65, 1997.

[318] J.H. von Thünen. Der isolierte Staat in Beziehung auf Nationalökonomie und Landwirtschaft. Gustav Fisher, Jena, Germany, 1826.

[319] I. Sue Wing. Induced Technical Change in Computable General Equilibrium Mod- els for Climate Change Policy Analysis. PhD thesis, Massachusetts Institute of Technology, Mai 2001.

[320] Ian Sue Wing. Representing induced technological change in models for climate policy analysis. Energy Economics, 28:539–562, 2006.

[321] World Bank. Global Economic Prospects - Commodities at the Crossroads - 2009. Report, The World Bank, 2009.

[322] J. Wu. Environmental Amenities, Urban Sprawl, and the Economic Landscape. In Analysis of Land Markets and the Impact of Land Market Regulation, a Lincoln Institute course, held from July 10-12, 2002, 2002.

[323] A. Zellner. An Eﬃcient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias. Journal of the American Statistical Association, 57(298):348–368, June 1962.

[324] Adriaan van Zon. Rum: a simple recursive update model providing a condensed representation of a putty-semi-putty vintage model. MERIT Research Memoranda 1994-002, 1994.

[325] van Zon, Adriaan and I. Hakan Yetkiner. Endogenous energy saving technical change in the romer model. Research Memoranda 022, Maastricht : MERIT, Maastricht Economic Research Institute on Innovation and Technology, 1999.

136