HERMIN-S4 A four-sector structural model of the Spanish Economy for the analysis of COMMUNITY SUPPORT FRAMEWORKS by Simón Sosvilla-Rivero* and José A. Herce* DOCUMENTO DE TRABAJO 94-08

June, 1994

* FEDEA (Foundation for Applied Economics Studies) and Universidad Complutense of Madrid.

Acknowledgements: This research has benefited from research assistance by Sonsoles Castillo. We are also in debt to John Bradley, for his collaboration through this project, David Taguas, Juan F. Jimeno, Pedro Abad and the participants at a presentation held at the Ministry of Finance. Financial Support by the EC Commission under contract JOU2-CT92-D257 is gratefully acknowledged. Remaining shortcomings are the authors responsibility.

This paper describes the main features of HERMIN-S4, a macroecometric model of the Spanish economy. HERMIN-S4 is a model with four sectors: the public sector (G), the exposed tradable sector (T) that coincides with the manufacturing industry, the protected non-tradable sector (N) that includes marketed services, building and construction and energy, and the agricultural sector (A). The basic philosophy underlying the design of HERMIN-S4 is twofold. On the one hand, we try to construct a fairly simple and flexible core model of our economy, which could, nevertheless, replicate its main relationships. On the other hand, it requires strict comparability with other similar models built for Greece, Ireland and Portugal, in order to use them in a joint research project to assess the effects of the Community Support Frameworks on growth in the European periphery.

Abstract

I. INTRODUCTION II. THE HERMIN-S4 MODEL FRAMEWORK 2.1. The Sectoral Breakdown 2.2. General Overview of the model 2.3. The Theoretical Framework 2.4. The HERMIN-S4 Data III. ESTIMATING HERMIN-S4 3.1. The Supply Side 3.2. Absorption 3.3. Income Distribution IV. POLICY ANALYSIS WITH HERMIN-S4 4.1. A Baseline for 1990-2010 4.2. Domestic Policy Shocks 4.3. External World Shocks 4.4. A Shock in Working Age Population V. CONCLUDING REMARKS REFERENCES TECHNICAL APPENDICES APPENDIX 1: Model Listing in XTSP language for SIMPC APPENDIX 2: Variables Description APPENDIX 3: Computation of Final Demand Measures

I. INTRODUCTION

This paper describes the results of on going research aimed at the construction, testing and regular use for policy analysis of a long-run structural econometric model of the Spanish economy. It contains thus the full model specification at four sector disaggregation, its estimation and a series of simulations of policy-induced or exogenous shocks. The results here presented are preliminary in the sense that both estimation and simulation have been kept as simple as possible although they will be refined in the near future. A comprehensive document describing a three sector version of the model is also available (Herce an Sosvilla (1994)).

The primary use of the model will be the analysis of Community Support Frameworks. For this, the model demanded strict comparability between the four peripheral countries of the E.U.: Greece, Spain, Ireland and Portugal.

There was also needed a very simple model favoring progressive sectoral disaggregation at no excessive data cost, focused in the long-run properties of the economy and suited to the analysis of both the supply and demand side effects of the structural funds from the E.U. to its less developed member states. We are thus mainly interested in the medium to long-run properties of the model, as well as comparability with the models built by the other teams from Greece, Ireland and Portugal, rather than the analysis of short-term fluctuations and forecasting.

HERMIN-S4 is thus the model that will serve that purpose. HERMIN owes its name to the HERMES model, a nine sector model developed in ten EC countries, during the 1980s, at the request of the EC Commission. In the case of the Irish and Portuguese models, it results actually from a contraction to four sectors and many other simplifications to obtain a mini-HERMES, from whence its name. As Spain lacked an operational HERMES model, we developed HERMIN-S4 which is the HERMIN version, four sectors, for the Spanish economy. The next step will be to develop its ability to fully assess Community Support Frameworks (CSF) effects.

These effects are wider than the conventional demand and current account implications. CSF grants to member States increase considerably the expenditure capacity of the receiving public administrations. For certain regions, in Spain, annualized European Fund for Regional Development (EFRD) transfers plus the additional funds to which national administrations are obliged under the CSF conventions amounted to almost 3% of their GDP. Such a shock to aggregate demand is several times larger than the conventional ones policy analysts usually consider in their models.

Public capital formation is greatly encouraged by this shock. The current account also registers positively the impact of CSF capital transfers. Part of the total grants can be directly transferred to firms for targeted expenditures by them. Finally, a fairly significant portion of CSF associated expenditure finances human capital enhancing activities.

It is well known that expansionary demand shocks will temporarily raise both output and prices. The long-term persistence of these changes will depend on the characteristics of the economy: expectations formation, types of rationing suffered by firms, degree of competition in their factor and product markets, etc.

CSF effects will also depend on how the extra expenditure will affect the supply side of the economy. Transport and telecommunications infrastructure, for instance, is vital for the efficiency of all types of operators. The research on the supply side effects of productive public expenditure is very active at present [see Draper and Herce, (1993) for a survey on this literature] and some of their best established results can be very easily incorporated in a model like HERMIN-S4.

Of course, these supply side effects are not as certain as the demand ones. They will only come about if the projects on which the money is spent are actually needed and efficiently operated. The microeconomic evaluation of these projects is of paramount importance to ensure this.

One specificity of the CSF in Spain, not present in the other three countries mentioned above, concerns the "north-south" divide of our economy. CSF funds are mostly directed towards nine less developed "Comunidades Autónomas" or regions of the seventeen that make the whole territory. This fact will force us to further develop HERMIN-Spain to capture differentiated growth in different groups of regions. This is a stimulating perspective from the modelling point of view for the purpose of CSF funds is actually to promote the catching-up of these regions.

One question remains that concerns the whole exercise of model building. Is it worthwhile to do this kind of exercise today? The reader is directed to Fair (1993), Helliwell (1993) and Taylor (1993). These authors emphasize how econometric models could be adapted in order to continue to be unavoidable policy analysis tools: through testing of their properties, if possible against alternative structural models (rather than against VAR or Autoregressive Components models); developing the linkages between national models to capture globalisation trends; or to take account of rational expectations based behaviour, credibility issues, growth, etc. In sum, the research programme before macroeconomic modellers seems to be intense and attractive.

One final question refers to the transparency of model building as argued by Wallis, (1993). An important requirement in modelling work is that it should be fully documented and completely open, so that the results can be replicated and compared with those from other macroeconometric models. When economists disagree, it is usually because they are making different assumptions or placing relative emphasis on parts of the system. If the assumptions are not made public, sensible discussion cannot proceed. The Spanish case is particularly lucky in this respect. We already mentioned the availability of the MOISEES model for comparison of results.

The rest of the paper is organized as follows. The sectoral desaggregation and the theoretical foundations of the behavioural equations of the model are discussed in Section 2 where also the data problems are mentioned. Section 3 is devoted to the estimation of the behavioral equations of the model. In Section 4 several policy simulations help to illustrate the properties of HERMIN-S4 in its present formulation and to explore the likely long-term growth patterns that would follow the Spanish economy under changing circumstances. Concluding comments on the whole exercise and future research agenda of the HERMIN-Spain team are the object of Section 5. A series of technical appendices contains the full model description, variables listing, etc.

II. THE HERMIN-S4 MODEL FRAMEWORK

2.1. The Sectoral Breakdown

The HERMIN-S4 model has been conceived as a four sector model of the Spanish economy. The progressive sectoral breakdown for HERMIN-Spain is as indicated in the Table 2.1.

The choice of the sectoral disaggregation is justified by the desire of keeping the model as small and simple as possible while separating sectors with different behaviour and driven by different forces:

i) the public sector (G) is dependent on Government policy decisions, with expenditure and tax rates as instruments.

ii) the exposed tradable sector (T) is driven by both domestic and foreign demand, and by international cost competitiveness.

iii) the protected non-tradeable sector (N) is driven by domestic demand, and

iv) the agricultural sector (A) is treated as mainly exogenous.

Table 2.1 Sectoral breakdown in HERMIN-SPAIN

2.2. General overview of the model

Table 2.2 contains the basic structure of the model that is detailed below. Only behavioural equations are listed in the table. The rest of the equations of the model, most of them identities, can be found in the model listing in Appendix 1.

2.3. The Theoretical Framework

In building the HERMIN-S4 model, we have opted for simple and encompassing theoretical foundations sufficiently established in other modelling exercises similar to this one. The way output, employment wages, prices, unemployment and aggregate demand are determined by the corresponding agents in the different sectors of the economy is explained in this chapter.

Taking into account that the model is going to be used for long-run policy analysis, the behavior of the agents we are considering is not going to be affected by cyclical factors. This allows us to symplify further the specifications by just considering the fundamental determinants of the agents' decisions in agreement with the underlying theories. In what follows we will concentrate on the most important behavioural equations in HERMIN-S4. The complete listing of the model can be found in Appendix 1.

2.3.1. The CES-based joint factor demand system (cost minimization)

Concerning output and factors demand decisions, we will opt for a CES production function that, under the assumption of cost minimization, will allow us to obtain demands for the services of labour and capital as functions of their relative prices, wages in the case of labour and user cost in the case of capital. In order to apply the cost minimization assumption, it will be further assumed that output is determined by final domestic demand, world output and the competitiveness of the economy. Once output is determined, cost minimizing employment per unit of output will be just a function of relative factor costs and a series of technology parameters proper to he CES specification that will be explained below. Equally, cost minimizing capital demand per unit of output will also be a function of relative factor prices and its own technology parameters.

The CES production function, following Bradley et al. (1993), can be formulated as:

\[O = A \left\{\delta \left(\exp \left(\lambda_ {L} t\right) L\right) ^ {- \rho} + (1 - \delta) \left(\exp \left(\lambda_ {K} t\right) K\right) ^ {- \rho} \right\} ^ {- (1 / \rho)}\tag{2.1}\]

where O, L and K are, respectively, added-value, employment and the capital stock. A is a scale parameter, is the elasticity of substitution, is a factor intensity parameter and and are the rates of technical progress embodied, respectively, in labour and capital. Whenever is positive the technological progress will be factor saving.

Cost minimization for a given level of output restricted to fulfill equation (2.1) above will give factor demands as:

\[\mathrm{K} = \mathrm{G} (\mathrm{O}, \mathrm{w/c})\tag{2.2}\]

and

\[\mathrm{L} = \mathrm{H} (\mathrm{O}, \mathrm{w/c})\tag{2.3}\]

where, c is the user cost of capital and w the wage rate.

Equivalently, by dividing by O, factor demands per unit of output can be expressed as:

\[\mathrm{K} / \mathrm{O} = \mathrm{g} (\mathrm{w} / \mathrm{c})\tag{3.2 \( ^{+} \)}\]

and

\[\mathrm{L} / \mathrm{O} = \mathrm{h} (\mathrm{w} / \mathrm{c})\tag{2.3'}\]

Furthermore, in order to apply the estimation techniques to the existing data on employment and capital formation we take gross investment instead of the capital stock. The huge simplification this procedure entails can be justified assuming that the flow of gross investment is a proxy for the latest vintage of capital in a "putty-clay" world. Thus our final joint demand system will be:

\[\mathrm{I} / \mathrm{O} = \mathrm{g} (\mathrm{w} / \mathrm{c}); \text { with } \mathrm{g} ^ {\prime} > 0\tag{2.2"}\]

and

\[\mathrm{L} / \mathrm{O} = \mathrm{h} (\mathrm{w} / \mathrm{c}); \text { with } \mathrm{h} ^ {\prime} < 0\tag{2.3 \( ^{++} \)}\]

where I stands for gross investment.

Once I is computed, one can build a proxy for the capital stock following the perpetual inventory rule:

\[\mathrm{K} _ {\mathrm{t}} = \mathrm{I} _ {\mathrm{t}} + (1 - \alpha) \mathrm{K} _ {(\mathrm{t} - 1)}\tag{2.4}\]

As indicated above, given the cost minimization criterion adopted, we need an alternative determination of the level of output in every period in order to derive the levels of employment and investment.

The equation for output in the tradable sector is:

OT = f(FDDWOT, OW, CCOMPT)

where OT is the tradable sector added-value at factor cost, FDDWOT is a measure of final domestic demand where each of its components is weighted by their private domestic T sector added-value content (see Appendix 3), OW is a measure of world trade, and CCOMPT is a measure of relative (to the rest of the world) cost competitiveness in the tradable sector. CCOMPT is defined as follows:

\[\mathrm{CCOMPT} = \mathrm{ULCT/ULCEC11}\tag{2.5}\]

Where ULCT are nominal labour costs in the tradable sector and ULCEC11 are nominal (in pesetas) unit labour costs in the European Community countries except Spain, weighted by their share of total GDP.

Output in the non-tradable sector (ON) is assumed to be demand driven, the driving variable being final demand where its components are weighted by there private domestic N sector added-value content (FDWON see Appendix 3):

\[\mathrm{ON} = \mathrm{f} (\text { FDWON })\tag{2.6}\]

Output and employment in the agricultural and public sectors will be estimated differently (see section 3.1).

2.3.2. Wage determination and wage bargaining

Wages result out of a process of wage bargaining between employers and trade unions. Of the very many elements that are relevant in this process (Layard, Nickel and Jackman, 1991; De Lamo y Dolado 1993) we will only consider consumption prices, productivity of labour, the unemployment rate and the tax wedge.

In the bargaining process, unions will, above all, try to translate price and productivity increases into nominal wages and, as long as they realize them fully, do the same with all kind of taxes that diminish their disposable income and/or its purchasing power. In doing so they will be restrained by the state of the labour market reflected in the current unemployment rate or its rate of change. Clearly, their power in the negotiation will be endangered by an excessive unemployment level. This is the so-called "Phillips curve" effect.

The proposed equation for the determination of wages, consistent with the above description of the bargaining process will thus be:

\[\mathrm{W} = \mathrm{f} (\mathrm{PC}, \mathrm{PROD}, \mathrm{UR}, \mathrm{WEDGE})\tag{2.7}\]

where PC is the private consumption deflator, PROD is the productivity of labour obtained as real output over employment in the corresponding sector, UR is the percentage unemployment ratio and WEDGE is the ratio of over (inclusive of social contributions).

Again, only in the private non-agricultural (i.e., T and N) sector wages will be modelled in this way. In the government and in the agricultural sectors, wages will simply be assumed as following wage setting patterns in the T sector.

2.3.3. Price determination

All the prices (deflators) for which equations are estimated in HERMIN-S4 are formed upon the evolution of either the deflator (PGDPFC) or the consumption deflator (PC). According, however, to the deflator considered, other variables have been taken into account. In general, prices will evolve driven by a combination of two factors: costs, notably labour costs and so mark-up pricing, and international prices. Exposure to competition at home and abroad will determine the weights in the above mentioned combination. Costs at home will be proxied by the deflator of and international prices will be represented by the imports deflator (PMP).

HERMIN-S4 contains equations for the following deflators:

\[\mathrm{PC} = \mathrm{f} (\text { PGDPFC }, \text { PMP }, \text { indirect taxes }); \text { consumption deflator } \tag {2.8}\]

\[\mathrm{PG} = \mathrm{f} (\text { PGDPFC }); \text { public consumption deflator }\tag{2.9}\]

\[\mathrm{PIH} = \mathrm{f} (\text { PGDPFC }, \text { PMP }); \text { residential investment deflator } \tag {2.10}\]

\[\mathrm{PIG} = \mathrm{f} (\text {PGDPFC}, \text {PMP}); \text {public investment deflator} \tag {2.11}\]

; private investment deflator (2.12)

; T sector exports deflator (2.13)

PXNTUR = f(PC); Non-turistic services exports deflator (2.14)

PXTUR = f(PC); Turistic exports deflator (2.15)

PGSUB = f(POP); defaltor of Government subsidies (2.16)

PGTE = f(PC); deflator of Government indirect tax revenue

(2.17)

2.3.4. Labour force participation

The extent to which the working age population (N1564, or total population of age between 15 and 64 years old) opts for participating into the labour force (LF) depends, leaving apart demographic factors captured here by a time trend, on the situation of the labour market given by the unemployment rate. In a rather simplistic way, thus, we will represent the labour supply by the following set of equations:

\[\mathrm{LFPRF} = \mathrm{f} (\mathrm{UR}, \text { time })\tag{2.18}\]

\[\mathrm{LFPRM} = \mathrm{f} (\mathrm{UR}, \text { time })\tag{2.19}\]

\[\mathrm{LFPR} = \mathrm{LFPRM} + \mathrm{LFPRF}\tag{2.20}\]

\[\mathrm{LF} = \mathrm{LFPR} * \mathrm{N1564}\tag{2.21}\]

\[\mathrm{UR} = 1 0 0 ^ {*} ((\mathrm{LF-L}) / \mathrm{LF})\tag{2.22}\]

where LFPRF, LFPRM and LFPR are female, male and total labour force participation rates respectively, UR the unemployment ratio, time a time trend, and L employment. The effect of UR on the participation rates, the encouragement effect, should have a negative sign (as established by several authors -see e.g., De Hevia and Novales, 1992), that is, the greater UR, the lower LFPRF or LFPRM. No sectoral breakdown is adopted for this part of the model.

2.3.5. Private consumption

Consumption by households is the largest single item in aggregated demand. A liquidity constrained consumption function is used, linking private consumption to real personal disposable income and real financial wealth.

The consumption function is thus:

\[\mathrm{CONS} = \mathrm{f} (\text { Y R P E R D }, \text { R G N D D })\tag{2.23}\]

where YRPERD is real personal disposable income (see Appendix 1 for a precise definition of how it is derived) and RGNDD is the outstanding debt of the government held by domestic residents, in real terms (as a proxy for financial wealth).

2.3.6. Total investment

The way factors demand where estimated saves us the separate specification of a private non residential investment equation. Total (gross) fixed investment, I, results out of the following identity:

\[\begin{array}{l} \mathrm{I=IP+IH+IG} \\ \mathrm{IP=IT+IN+IA} \end{array}\tag{2.24}\]

(2.25)

where IP is gross fixed private non residential investment; IT, IN and IA is the split of IP between the tradable, non-tradable and the agricultural sectors; IH is the gross fixed private residential investment; and IG is gross fixed public investment.

Residential investment, IH, will be a function of personal disposable income, YRPERD, both expressed in per capita terms:

\[\mathrm{IH} = \mathrm{f} (\text { Y R P E R D }, \mathrm{N1564})\tag{2.26}\]

Public investment, IG, is an exogenous variable.

2.3.7. Imports, exports and the balance of trade

In order to assure the closing of the model it was decided that imports should not be modelled behaviourally but rather determined residually once output and exports had been estimated (see Appendix 1 for details). Exports in the private sector, XP, are:

\[\mathrm{XP} = \mathrm{XT} + \mathrm{XN} + \mathrm{XA}\]

that is, exports in the T sector, XT, exports in the N sector, XN, and agricultural exports, XA.

Agricultural exports (XA) are estimated as a ratio to agricultural output against a constant and a time trend.

It is well known (see, e.g., Fernández and Sebastián, 1991) that industrial exports in Spain increase when domestic producers find difficulties to sell their products at home.

Manufacturing sector exports (XT) are thus estimated as follows:

XT = f(OW, GNPDOT, CCOMPT, DUMCEE)

that is, XT is a function of world output, the rate of growth of gross national product (proxing the economic cycle), cost competitiveness, and a dummy variable to take account of the joining of Spain to the European Community in 1986 since this fact changed the trade regime of the Spanish economy.

Exports of the N sector have been split into tourism exports (XTUR) and non-tourism ones (XNTUR). The former are assumed dependent of world output and (cost) competitiveness, and the latter only of world output. Indeed, touristic exports are known to be highly dependent on the western economies cycle.

The high income elasticity of Spanish imports is also well documented (see, e.g., Sebastián, 1991). This fact is properly captured by the way imports are residually determined in HERMIN-S4.

2.4. The HERMIN-S4 data

The HERMIN-S4 model has 225 variables of which 46 are exogenous, 39 behavioral and the other 140 derived through identities. This means that data series for a large number of variables, ranging between 1964 and 1990 had to be constructed. Of them, a large number were already available in the public domain MOISSES data base "DEMO" (M/DEMO here after) [Molinas, Sebastián and Zabalza (eds.) 1991], although the range was 1964-1988 which required extension of the series using the standard sources. Although a complete listing of the variables names and sources is offered in Appendix 2, we refer here to those for which ad hoc procedures had to be used.

Linkage of series expressed in different base years.

The Spanish National Accounts (CNE) offer data series starting in 1964. These series have a break in 1985-1988 due to the change of the base year from 1980 to 1986. We have linked the different subperiods by multiplying the observations from 1988 on by , where is the 1986 observation for the variable X computed in yy base year. This simple procedure harmonizes the level of both subseries while keeping the profile unchanged. All real variables are expressed at 1980 prices. While nominal variables can be either obtained directly from their respective sources or through their deflators.

As it is well known, however, the bias introduced by this procedure increases with time. The procedure has, nevertheless, been used to push the series from 1988 up to 1990 and the bias is small. As the Statistical Office issues more recent data, a linkage procedure will be devised in order to link backwards the 1986-base series down from 1985.

The above mentioned linkage procedure has been applied to all series listed in Appendix 2 whose sources are in M/DEMO and/or CNE (see Appendix 2 for a precise definition of the Sources).

The obtention of sectoral data has been proved particularly difficult and we had to adopt several ad hoc procedures. Given that we wanted to build-up also a four sectors model, it was decided to gather data on capital formation, employment, compensation of employees, added value and exports at the R6 level (i.e., six branches) of the National Accounts system. The six branches are: agriculture, energy, manufacturing, building and construction, marketed services and non-marketed services. The last branch is identified with the Government sector.

From this sectoral breakdown we collapsed the data to the four sectors level described in Section 2.1 above. The ad hoc procedures used are described in what follows.

Added value

The Spanish Statistical Office has recently published data for added value at branch level in the base year 1986 for the period 1964-1991 (see INE (1992)). From this publication we obtained the sectoral shares in gross added-value and applied then to the linked series in base 1980 obtained from M/DEMO and CNE sources:

\[A V \text { Sector } _ {B 1 0} ^ {\text { Pre60 }} \cdot \left(\frac {A V \text { Sector }}{G A V}\right) _ {B 1 6} ^ {\text { Pre66 }} \cdot G A V _ {B 1 0} ^ {\text { Pre60 }}\]

\[A V \text { Sector } _ {B 1 0} ^ {\text { Plan }} = \left(\frac {A V \text { Sector }}{G A V}\right) _ {B 1 6} ^ {\text { Plan }} \cdot G A V _ {B 1 0} ^ {\text { Plan }}\]

for constant (i.e., 1980) and current (cc) prices magnitudes.

Employment and wage-earners

Employment in the Government sector (LG) has been taken from the Labour Force Survey (EPA) for the years 1977-1992. The years before where obtained projecting backwards the 1977 figure with the corresponding growth rates for this variable obtained from the M/DEMO data base.

Once LG was obtained we subtracted this from the total employment series taken from García Perea (1992) and EPA to derive private total employment (LP).

From LP we obtained the sectoral series multiplying it by the sectoral employment shares (in private employment) computed from M/DEMO and EPA.

These procedures where needed due to the frequency with which the employment series change their methodology in Spain. The only harmonized series for total employment and wage-earners are those offered by García Perea (1992).

The procedures followed in order to construct the sectoral wage-earners series where exactly the same than in the previous case.

Compensation of employees

These data are only available since 1970, when a first figure was published, for the different branches, by the Statistical Office. No data is available between 1970 and 1980, when yearly data publication was resumed. We proceeded to the linkage of the different base years as explained above, but the data for 1971 though

1979 was intrapolated, for each sector, after computation of the average annual growth rate. We further ensured that the sectoral figures added-up to the aggregate figure published by the Statistical Office.

Capital formation

The aggregate series for capital formation by the private sector and by the government were obtained, after linking the different base years, from M/DEMO and CNE. Then, to the private capital formation, we applied the sectoral shares computed from the corresponding sectoral series (volume and current prices) of the MIDE model data base. These series are only available since 1970.

Capital stock

We computed the capital stock for the A, T, N and G sectors using the perpetual inventory formula:

\[\mathrm{K} = (1 - \delta) ^ {*} \mathrm{K} (- 1) + \mathrm{I}\]

where we arbitrarily chose a depreciation rate of 5% (i.e., ) for the A and the G sectors, and a 10% depreciation rate (i.e., ) for the T an N sectors.

The capital stock in 1969 was fixed as follows:

K (69) in T = 2,800 bn. Pta. (1980 prices)

K (69) in N = 4,200 bn. Pta. (1980 prices)

K (69) in A = 700 bn. Pta. (1980 prices)

K (63) in G = 1,227 bn. Pta. (1980 prices)

The figures for G and P sectors were obtained from M/DEMO data base. The T, N and A sectors totals were obtained using capital formation data in the sample period to infer the shares of the three sectors.

User cost of capital

The conventional formula to compute the cost of capital services:

\[\mathrm{PK} = \mathrm{PI} ^ {*} (\mathrm{i} - \pi + \delta)\]

(where, i is an interest rate, is an inflation rate and is the depreciation rate), produced very volatile results due to the excessive variation of real interest rates. We averaged through the sample period so that, in fact, PK for each sector resulted in a constant proportion of the price of investment in the corresponding sector.

Final demand measures

These are FDDWOT and FDWON and result of a weighted sum of the Final Domestic Demand and Final Demand components, where the weights reflect respectively, the T and N sectors output contents of those components. See Appendix 3 for a detailed discussion of their construction.

Variables set to zero

The variables IGVOTH, IHG an IHGV have been set to zero due to insufficient desaggregation of the National Accounts data. Their true values are however recovered in other variables, so that the aggregate out of them are correct.

III. ESTIMATING HERMIN-S4

In this section we present the results obtained when the behavioural equations in HERMIN-S4 were estimated using annual time series. Each set of estimation results is preceded by a short description of the equation, with reference to the theoretical model described in Section II. A mathematical formulation of the equation is then presented, followed by estimation results from the econometric package TSP 4.2, which are largely self-explanatory, as well as by some brief remarks on the implications of the results.

In spite of the simultaneity of the system, the equations in HERMIN-S4 are estimated by ordinary least squares (OLS). In some cases, correction for first order serial-correlation is added. Instead of using ordinary Cochrane-Orchutt corrections, we use the Beach and McKinnon's (1978) maximum-likelihood estimation that assumes that the disturbance term follows a stationary first-order autoregressive process. The small sample properties of the estimators obtained from this method (that we describe as AR1 estimation for brevity) are preferable to those of the conventional procedures.

After the four-sector prototypical models have been completed for all countries involved in the present research project (Greece, Ireland, Portugal and Spain), it is intended to address the issues of the possible non-stationarity of the time series and the dynamic specification of the equations making use of the so-called "cointegration analysis" (see Dolado, et. al. 1990, for a survey) and its associated concept of error correction model, combining the latter flexibility in the dynamic specification with desiderable long-run properties (see Hendry and Richard, 1983).

3.1. The Supply Side of HERMIN-S4

3.1.1. Tradable private sector behavioural equations

As mentioned in Section 2.2.1, T sector output is determined in an equation involving final demand weighted by T sector output content (FDDWOT), world output (OW) and cost competitiveness (CCOMPT). The equation was estimated by OLS using data for the period 1970-1990; where a time trend was included to account for variations in the weights used in composing FDDWOT from the 1987 input-output tables:

From this estimated equation, the long-run elasticity of T sector output with respect to final domestic demand weighted by T sector output content, world output and expected cost competitiveness are 0.57, 0.64 and -0.50, respectively.

Concerning the estimation of the parameters of the CES production function in the T sector, although the static CES function is probably too simple to capture the structure and dynamics of factor demands, estimation yielded plausible estimates. Our justification for this simplified CES approach is that we are mainly interested in the medium to long-run properties of the model, rather than in the short-run dynamics and tracking performance.

The estimation method is however rather complex (see Bradley and Fanning, 1984, Appendix 3.1). Recall that the marginal production function can be approximate a by:

\[Q = A \left\{\delta \left(\exp \left(\lambda_ {L} t\right) L T\right) ^ {- \rho} + (1 - \delta) \left(\exp \left(\lambda_ {K} t\right) I T\right) ^ {- \rho} \right\} ^ {- (1 / \rho)}\]

It can hence be deduced that cost-minimisation implies the factor proportions equation

\[\log \left(\frac {I T}{L T}\right) = \log \left(\frac {1 - \delta}{\delta}\right) + \sigma \log (R F P T) + (1 - \sigma) (\lambda_ {L} - \lambda_ {K}) t\]

where RFPT is the relative factor price in the T sector. Using AR1 over 1970-1990 yields the estimated cost-minimising factor proportions equation:

Two of the production function parameters can be recovered from this equation: and , together with the bias of technical change. We solve the factor proportions equation and production function for the long-run labour demand, yielding the equation:

\[\begin{array}{r l} & \log (L T) - \log (A) + \frac {\sigma}{1 - \sigma} \log (\delta) + \log (O T) + \\ & \frac {\sigma}{1 - \sigma} \log \left[ \left(\frac {\delta}{1 - \delta}\right) ^ {\delta} R F P T ^ {- (1 - \sigma)} e ^ {- (\sigma - 1) (\lambda_ {L} - \lambda_ {D})} + 1 \right] - \lambda_ {L} t \end{array}\]

Taking everything except the first and last right-hand-side terms over to the left-hand-side gives us a regression in a constant and a time trend from which we can trivially estimate all the remaining parameters of the production function. The estimated parameters are as follows:

A (the scale parameter) = 0.6882

σ (the elasticity of substitution) = 0.7750

(the factor intensity parameter)

(the rate of technical progress embodied in labour) = 0.0371

(the rate of technical progress embodied in capital) = -0.0687

Notice that we therefore estimate technical progress to be capital using and labour saving at a rate of 6.9 and 3.7 percent per annum, respectively.

In HERMIN-S4, the price of the T sector output (POT) is determined in an equation involving the world price (PWORLD) and unit labour costs (ULCT). Price homogeneity in PWORLD, not rejected by the data, is imposed for simplicity. Estimation by AR1 over 1969-1990 yields the following:

As mentioned in Section 2.3.2, nominal wages in the T sector (WT) are modelled as the outcome of bargaining between employers and trade unions, the relevant explanatory variables being private consumption deflator (PC), labour productivity in the T sector (PRODT), the rate of unemployment (UR), and the "wedge" driven by taxes between the wage denominated in the employer's (output) price and the take-home consumption wage enjoyed by workers (WEDGE). Estimation by AR1 using data for the period 1970-1990 yields:

Notice that, in contrast to the Irish case, we did not find a significant tax-wedge effect. Regarding this, it has been argued that the wedge variable could have very strong short-run effects, but its long-run effects could be not very well determined. Layard, Nickell and Jackman (1991) state that "the effect of the wedge on wage pressure probably does not last for ever" (p. 33). Anderson (1993) finds that the tax wedge effects in Spain are only present in the short-run dynamics. Note that a dummy variable DUMMON taking the value 1 from 1978 and zero before was introduced to represent the move to a more centralized, and moderate, wage outcome with the the Moncloa Agreements signed by trade unions and employers. The estimated equation shows a change in the elasticity of wages in the T sector with respect to consumption prices after the Moncloa Agreements: before 1978 there was a strict over-indexation of wages to prices, and after that year an under-indexation. Notice also that productivity gains are not fully translated into wages, since the coefficient on log (PRODT) is 0.70. The Phillips curve effect (-0.005) is roughly half the size to that reported for Spain in Drèze and Bean (1990, p. 23), and also much smaller than those obtained for Ireland (-0.026) (Bradley and Wright, 1994) and Portugal (-0.046) (Modesto and Neves, 1993). Finally, we find a significant and negative effect for a dummy variable proxing Spanish integration into the EC (DUMCEE). This may well have been driving by the removal of trade barriers after entry into the Community.

3.1.2. Non-tradable sector behavioural equations

Output in the non-tradable sector (ON) is determined by final demand weighted by N sector output content (FDWON), and a time trend, to take account of variations in the weights from their 1987 input-output values. OLS estimation using the 1971-1990 period yields the following:

Concerning the estimation of the parameters of the CES production function in the N sector, using data for the 1965-1990 period, AR1 estimation of the cost minimizing marginal factor proportions equations yields:

The values obtain for the elasticity of substitution ( ) is 0.51, suggesting that sustitution possibilities at smaller in this sector than in the tradables sector. Technical progress in this sector is estimated to be both capital and labour using ( and ). The scale parametre (A) is estimated to be 1.17, and the factor intensity parameter ( ) is found to be 0.99.

Prices in the N sector are determined by a mark up on labour cost (ULCN). Using AR1 with data for the 1970-1990 period yields:

Wage determination in the non-tradable sector is identical to wage formation in the tradable sector. Nominal wages (WN) are determined by private consumption deflator (PC), labour productivity (PRODN), and the rate of unemployment (UR). The estimation results using OLS over the 1964-1990 period are as follows:

Note again, that no significant tax-wedge effect was found, and that a dummy variable DUMMON was introduced to take into account the deceleration in wage inflation following the Moncloa Agreements in 1978. The estimated equation shows a change in the elasticity of wages in the N sector with respect to consumption prices after 1978: before that year there was over-indexation of wages to prices and, after that year, under-indexation. Notice also that productivity gains are more than fully translated into wages, since the coefficient on log(PRODN) is 1.69. Finally, the

Phillips curve effect (-0.007) is greater than that estimated for the tradable-sector wages (-0.005).

3.1.3. Agricultural sector behavioural equations

HERMIN-S4 does not attempt to model the agricultural sector behaviourally. Nevertheless, in order to isolate agricultural activities from the rest of the private sector, we model agricultural output (OA), employment (LA) and the capital/output ratio (KA/OA) by simple time trends. The results obtained by OLS for the period 1970-1990 are as follows, where no attempt is made to correct by serial correlation except in the case of LA.

3.1.4. Public sector behavioural equations

Real output in the public sector (OG) is driven in our model by public sector employment (LG). Estimation by AR1 using data for the period 1971-1990 yields:

In HERMIN-S4 the wage rate in the public sector (WG) is assumed to follow the wage rate in the tradable sector (WT). Estimation by AR1 for the period 1971-1990 yields:

Finally, since public sector output is largely measured by labour cost inputs, the deflator (POG) is modelled as a function of WG. The results obtained by AR1 using data for the period 1965-1990 are as follows:

3.1.5. Labour supply equations

As mentioned in Section 2.3.4, the female and male labour force participation rates (LFPRF and LFPRM) are a function of the rate of unemployment (UR) and demographic factors captured by a time trend. OLS and AR1 estimation for the period 1970-1990 yields the following for LFPRF and LFPRM, respectively:

Notice that the time trend does not appear in the male participation ratio equation due to colinearity with UR given also its negative sign. Indeed, female participation is steadily growing while the downwards trend in male participation has been overtaken by rapidly rising unemployment.

3.2. The Absorption Side of HERMIN-S4

3.2.1. Private consumption

As mentioned in Section 2.3.5, households consumption in constant prices (CONS) is a function of real personal disposable income (YRPERD) and real financial wealth, proxied by government debt held by residents deflated by PC (GNDD/PC). The long-run interest rate proved non-significant. Estimation by OLS for the period 1970-1990 yields:

We obtain a marginal propensity to consume of 0.94, which implies a marginal propensity to save of 0.06. DUM7680 is a dummy variable taking the value 1 in 1976 and 1980 0 in the rest of the period. The wealth effect is of the expected sign and size.

3.2.2. Residential investment

In HERMIN-S4 residential investment per capita (IHP/N1564) is assumed to be dependent on real personal disposable income per capita (YRPERD/N1564). AR1 estimation for the period 1970-1990 yields the following:

3.2.3. Agricultural exports

The ratio of agricultural exports (XA) to agricultural output (OA) is modelled as a simple time trend. OLS estimation for the period 1965-1990 yields:

3.2.4. Exports of the tradable sector

Export of the tradable sector (XT) are assumed to be driven by world output (OW), relative cost competitiveness (CCOMPT), and the Spanish economic cycle represented by GNPDOT. The results obtained by OLS using data for the period 1970-1990 are as follows, where DUMCEE is a dummy variable taking the value 1 from 1986 on and 0 before that year:

3.2.5. Exports in the non tradable sector

Exports in this sector are divided into turistic services exports (XTUR) and non turistic exports (XNTUR). Both are dependent on world output (OW), and XTUR also on cost competitiveness (CCOMPN).

Estimation by OLS over 1970-1990 yields for XTUR:

3.2.6. Imports

In HERMIN-S4, for simulation purposes, imports are obtained residually from the product expenditure equation, so they are not estimated. Given the effect of

GNPDOT on exports this procedure ensures that imports will increase with aggregated demand.

3.3. Income Distribution in HERMIN-S4

3.3.1. Absorption price determination equations

The majority of the absorption deflators (PABS) are modelled in the following way:

\[\log (\mathrm{PABS}) = \mathrm{a} _ {0} + \mathrm{a} _ {1} \log (\mathrm{PGDPFC}) + (1 - \mathrm{a} _ {1}) \log (\mathrm{PMP})\]

where PGDPFC is the deflator of total GDP at factor cost and PMP is the deflator of total imports. The only exceptions are the consumption deflator (PC), where we add a term to account for the influence of net indirect taxes (TINC) on consumer prices, the price of public consumption, the price of exports of turistic and non turistic services and the price of residential investment, where no significant effect was found for PMP. The results obtained by AR1 for the period 1970-1990 are as follows:

Price of private consumption:

From this estimated equation, a one percentage point rise in the net indirect tax rate (TINC) will be translated in an increase in the consumption deflator rate by 0.80 percentage points, ceteris paribus. This effect is smaller than those computed for Ireland (1.14) (Bradley and Wright, 1994) and Portugal (0.996) (Modesto and Neves, 1993).

ii) Price of public consumption:

iii) Price of investment in housing:

where the dummy variable DUM7579 eliminates the anomalous data for the years 1975 and 1979.

v) Price of government investment:

v) Price of private investment:

vi) Price of the tradable-sector exports:

vii) Price of the non-turistic services exports:

Note that here we take the price of private consumption to be the relevant national deflator of the exports of non-turistic services.

viii) Price of turistic exports

ix) Deflator of Government subsidies:

x) Deflator of Government indirect tax revenue:

Both the deflator for government subsidies and that for taxes on production and imports are estimated in order to have proper real adjustments for the corresponding nominal magnitudes, being thus able to compute real GDD at market prices out of real GDD at factor costs.

3.3.2. Income Determination Behavioural Equations

There are only two estimated equations in the income determination part of the model: total depreciation and undistributed profits. The rest of the magnitudes are derived through identities (see Appendix 1).

i) Total depreciation:

Total depreciation (DEP) is determined in a "technical" relation as a function of the value of total private capital stock (PIP*KP). OLS estimation over the period 1965-1990 yields the following:

ii) Undistributed profits:

Undistributed (i.e. retained) profits (YCU) are cyclically influenced by the growth rate of GNP (GNPDOT) and by total profits (YC). Estimation by OLS for the period 1971-1990 yields the following:

Notice that these results imply that the undistributed profits increase both as company income rises and during recoveries.

IV. POLICY ANALYSIS WITH HERMIN-S4

Empirical econometric models can be tested in two complementary ways. First, its ability to track the evolution of the relevant aggregate and sectoral variables over the sample period can be studied, by carrying out static and dynamic system simulations and then examining the differences between the simulated and historical values of the endogenous variables. Second, the partial derivatives or multiplier properties of the model can be examined by constructing a benchmark or baseline simulation and then re-run the model over the same period after shocking one or more exogenous variables. The perturbed values of the new simulation can then be compared to the baseline values providing an assessment of the consequences of the shock over time.

The first form of testing is less relevant here as the model was not constructed to be a short-term forecasting model. Nevertheless, the within-sample simulation errors were reasonably small. The second type of analysis requires a full understanding of the system-wide properties of the model, not all of which will be obvious from an isolated analysis of the component equations.

In this chapter we study the effects of variations in the settings of some policy instruments or other exogenous variables included in the specification of HERMIN-S4 by comparing their effects on a baseline projection for the years 1990-2010. Section 4.1 describes the assumptions concerning the baseline simulation. Section 4.2 investigates how the model behaves under changes in fiscal policy and other government exogenous variables. Section 4.3 illustrates how the model describes the effects on the Spanish economy of an increase in world economy activity. Finally, section 4.4 explores the effects of an increase in working age population.

4.1. A Baseline for 1990-2010

The baseline against which we carry out our experiments was obtained as follows. For the period 1978-1990, the core model was forced to track the historical data exactly by means of additive constant adjustments in the behavioural equations. For the period 1991-2010, we project the exogenous variables assuming a 'no-change' or 'neutral' policy stance, a non-inflationary international environment, and a growth in world output. This is not intended to be a realistic economic projection, the construction of which would require much more work.

Before explaining the consequences of the various shocks, it will be useful to look at the way the model translates the shocks to the different endogenous variables. This will be done with the help of the flow diagram shown in Figure 4.0

Two basic blocks can be distinguished: a wage-price block (nominal) and a product/employment block (real) the link between the two are wages (here represented by WP-i.e., wages in the private sector).

As can be seen in the lower-right corner of the diagram, an increase in demand (e.g. through IG) will be translated to output and employment in the private sectors (OP, LP). This will increase wages through the Phillips curve effect (a decrease in UR) which will, in turn, fuel the wage-price round-about and increase the relative factor price thus leading to a lower employment/output ratio. This would not necessarily prevent prices and nominal wages from raising moderately as it turns to be the case within HERMIN-S4. The product/income/demand feed-back is also represented in the lower-right corner of the diagram.

\[\mathrm{ULCP} = \frac {\mathrm{WP}}{\mathrm{PRODP}}\]

Exogenous shocks through: POA, OW, G, IG, TINC, ULCEC11, WEDGE

4.2. Domestic Policy Shocks

To study the effects in the model of changes in fiscal policy, four simulations were carried out concerning public sector employment, public investment and tax rates.

i) A shock to public employment

Table 4.1 and Figure 4.1 present the model's response to a 5% sustained rise in the number of public sector employees above its baseline level starting in 1991. In this simulation, tax rates are exogenous, extra public expenditure being financed by public borrowing at fixed debt interest rates.

As can be seen, the impact on consumption is positive and thus expands GDP. The initial impact on the labour market is to increase total employment and decrease the unemployment rate. The reduced unemployment rate acts through the Phillips curve to drive up wages and prices. The rise in wages leads to a strong downward wards correction in employment. Since the subsequent rise in wages is lower than the rise in the user cost of capital, there is also a similar decrease in investment.

The strong effect on consumption also leads to increased tax revenue. Since the increase in public consumption is greater than the increase in tax revenue, public borrowing increases. On the other hand, there is a rise imports following the increase in GDP which also induces a fall in exports. This however keeps relatively stable the current account.

ii) A shock to public investment

Table 4.2 and Figure 4.2 show the model's response to a sustained increase in public sector investment of 1% of GDP above its baseline level. At this stage no specific supply side responses (such as externalities) are assumed.

The impact on domestic demand is positive, increasing GDP. There are an increase in private employment and a decrease in the unemployment rate, which leads to an increase in wages and prices that is translated into a fall in competitiveness. Exports go up and imports decrease, causing an increase in the trade balance/GNP ratio.

The increase in the level of activity affects positively private investment. The lower unemployment and higher wages increase personal disposable income, increasing consumption and residential investment, and this brings a further increase in domestic demand.

The increase in public borrowing as percentage of GNP following the increase in public sector investment is less than 1%, since higher activity leads to higher Government revenues and lower Government transfers.

Notice the differentiated response of output employment and wages in the T and N sectors due to the fall in competitiveness after an expansionary demand shock.

iii) A shock to indirect taxation

Table 4.3 and Figure 4.3 present the model's response to a sustained rise in the implicit indirect tax rate of ten percent points above its baseline level.

An increase in the indirect tax rate forces the consumption price to go up, reducing real private consumption, forcing a decrease in both GDP and investment. The recession in turn leads to a decrease in imports and to a fall in employment and an increase in the unemployment rate. As the latter increases, the Phillips curve effect begins to push wages downwards, but this effect is not sufficient to boost employment. Government revenues decrease due to the diminishing activity. Finally, there is a rise in exports since the increase in GDP an imports in greater than the increase in final demand.

iv) A shock to direct taxation

Table 4.4 and Figure 4.4 show the model's response to a sustained rise in the implicit personal income tax rate of a ten percent points above its baseline level.

The decline in real disposable income negatively effects. Both consumption and investment are negatively affected by this shock. The contraction in aggregate demand leads to a decrease in GDP and a strong reduction in the rate of inflation the format reducing further investment. Despite the latter, the decrease in activity implies a fall in employment and a rise of the unemployment rate. Since imports decrease due to the reduction in aggregate demand and exports increase subsequently, the balance of payments also improves. Finally, the increase on the rate of personal income tax rate increases Government revenues, despite the reduction in the level of activity and in consumption which in turn means a reduction of public deficit.

4.3. World Demand Shock

Table 4.5 and Figure 4.5 present the model's response to a sustained increase in the level of world economic activity above its baseline level.

As can be seen, GDP responds initially by about 0.12% declining latter to about 0.04 in the year 2010. Both T production and exports go up, having a positive effect on employment. The increase in employment and wage leads to an increase in personal disposable income, which in turn means an increase in consumption. Both the increased activity and higher wages induce a rise in investment. The rises in consumption and investment lead to an increase in imports. Finally, there is an increased tax revenue an, therefore, a reduction in public borrowing.

4.4. A Shock in Working Age Population

Table 4.6 and Figure 4.6 show the once and for all expansion of the labour force by 1% that leads to a rise in the unemployment rate, which, through the Phillips curve effect, reduces wages. Employment increases, but this increase is not enough to offset the rise in the labour force.

In the first years, the increased number of unemployed reduced consumption and GDP, leading to a reduction in the rate of inflation. As employment raises, so does consumption and GDP.

Greater transfers to the unemployed leads to an increase in public borrowing and lower domestic prices induce a fall in imports. There is a reduction in exports since the fall in GDP and imports is greater than the fall in final demand.

TABLE 4.1. Effects of a sustained 5% increase in Public Employment.

* Percent deviation from baseline ** Deviation from baseline.

TABLE 4.2. Effects of a sustained 1% GDP equivalent decrease in Public Investment.

* Percent deviation from baseline ** Deviation from baseline.

TABLE 4.3. Effects of a sustained 10% increase in Indirect Tax Rate

* Percent deviation from baseline ** Deviation from baseline.

TABLE 4.4. Effects of a sustained 10% increase in Direct Tax Rate.

* Percent deviation from baseline ** Deviation from baseline.

TABLE 4.5. Effects of a sustained 1% increase in World Output.

* Percent deviation from baseline ** Deviation from baseline.

TABLE 4.6. Effects of a sustained 1% increase in Working Age Population.

* Percent deviation from baseline ** Deviation from baseline.

FIGURE 4.1. Effects of a sustained 5% increase in Public Employment

Percent deviation of GDP, CONS and I from baseline

Deviation of Prices from baseline

Deviation of Wages from baseline

Percent deviation of L, LN and LT from baseline

Percent deviation of EXPORTS and IMPORTS from baseline

Deviation of GBORR and BPTR from baseline

Deviation of Unemployment Rate from baseline

FIGURE 4.2. Effects of a sustained 1% increase in Public Investment.

Percent deviation of GDP, CONS and I from baseline

Deviation of Prices from baseline

Deviation of Wages from baseline

Percent deviation of L, LT and LN from baseline

Percent deviation of XP and MP from baseline

Deviation of GBORR and BPTR from baseline

Deviation of Unemployment Rate from baseline

FIGURE 4.3. Effects of a sustained 10% increase in Indirect Tax Rate.

Percent deviation of GDP, CONS and I from baseline

Deviation of Prices from baseline

Deviation of Wages from baseline

Percent deviation of L, LT and LN from baseline

Percent deviation of EXPORTS nd IMPORTS from baseline

Deviation of GBORR and BPTR from baseline

Deviation of Unemployment Rate from baseline

FIGURE 4.4. Effects of a sustained 10% increase in Direct Tax Rate.

Percent deviation of GDP, CONS and I from baseline

Deviation of Prices from baseline

Deviation of Wages from baseline

Percent deviation of L, LT and LN from baseline

Percent deviation of EXPORTS and IMPORTS from baseline

Deviatin of GBORR and BPTR from baseline

Deviation of Unemployment Rate from baseline

FIGURE 4.5. Effects of a sustained 1% increase in World Output. Percent deviation of GDP and OW from baseline

Percent deviation of Prices from baseline

Percent deviation of Wages from baseline

Percent deviation of L, LN and LT from baseline

Percent deviation of EXPORTS and IMPORTS from baseline

Percent deviation of GBORR and BPTR from baseline

Deviation of Unemployment Rate from baseline

FIGURE 4.6. Effects of a sustained 1% increase in Working Age Population.

Percent deviation of GDP, CONS and I from baseline

Deviation of Prices from baseline

Deviation of Wages from baseline

Percent deviation of L, LN and LT from baseline

Percent deviation of EXPORTS and IMPORTS from baseline

Deviation of GBORR and BPTR from baseline

Deviation of Unemployment Rate from baseline

V. CONCLUDING REMARKS

In its present format, HERMIN-S4 captures in a very simplified way the structure of the Spanish economy and allows the analysis of conventional policy shocks. The estimation of the behavioral equations of the model, given the standard specifications imposed, is quite acceptable. We always found the expected signs of the coefficients and multipliers, whose size is in line with those obtained for Portugal and Ireland, countries for which we have similar models. As the improvement of sectoral data continues, we hope to obtain better fits for the standard behavioral relationships of the supply side of the model.

The results of the simulations show that HERMIN-S4 has the desirable properties concerning the stability of the multipliers, their size and signs. This multipliers compare well with those derived from alternative models of the Spanish economy.

We believe that, based on the work done so far, HERMIN-S4 has a suitable formulation to attempt more ambitious tasks in the immediate future. First is to carefully develop the theoretical foundations of the presumed supply-side effects of CSF funds. This involves an effort in order to import into the simple framework of HERMIN type models the relevant parts of the recent literature on growth. Next is to add to HERMIN-S4 a north/south "module" to give the model the ability to supply differentiated results in regions eligible for CSF and regions not eligible. Finally, to further develop the very simple econometrics of the model, proceeding to block simultaneous estimation and differentiated short-run/long-rung specifications for the relevant behavioral relationships of the model in order to apply it to medium term macroeconomic analysis of the Spanish economy.

REFERENCES:

1. Anderson, B. (1993): "Spain: Evaluating the Effects of Macro Policy Using an Econometric Model", National Institute Economic Review, Num. 146, pp. 76-89.
2. Beach, C. and McKinnon, J. (1978): "A Maximum Procedure for Regression with Autocorrelated Errors", Econometrica, Vol. 46, pp. 51-58.

Bradley, J. and Fanning, C. (1984): Aggregate Supply, Aggregate Demand and Income Distribution in Ireland: A Macrosectoral Analysis, Dublin: The Economic and Social Research Institute.

Bradley, J. and Wright, J. (1994): "HERMIN-I4: A Four-Sector Structural Model of the Republic of Ireland", Technical Paper submitted to the European Commission.

Bradley, J., Whelan, K. and Wright, J. (1993): Stabilization and Growth in the EC Periphery: A study of the Irish Economy, Aldrshot: Avebury.

Christ, C. F. (1966): Econometric Models and Methods, New York: J. Wiley and Sons.

De Hevia, J. and A. Novales (1992): "¿Es la participación activa procíclica en España?, Working Document 92-05. FEDEA. Madrid.

De Lamo, A. R. and Dolado, J. J. (1993): "Un Modelo del Mercado de Trabajo y la Restricción de Oferta en la Economía Española", Investigaciones Económicas, Vol. 17, pp. 87-118.

Dolado, J., Jenkinson, T. and Sosvilla-Rivero, S. (1990): "Cointegration and Unit Roots", Journal of Economic Surveys, Vol.4, pp. 249-273.

1. Draper, M. and Herce, J. A. (1993): "Infraestructuras", Documento de Trabajo 93-07, FEDEA.

Drèze, J. H. and Bean, C. R. (1990): "Europe's Unemployment Problem: Introduction and Synthesis", in Bean, C. R. and Drèze, J. H. (eds.) Europe's Unemployment Problem, Cambridge, Mass.: The MIT Press, pp. 1-65.

1. EE (1993): Statistical Annex, European Economy, Num. 54, 1993.

Fair, R. C. (1993): "Testing Macroeconometric Models", American Economic Review, Vol. 83, pp. 287-293.

1. Fernández, I. and Sebastian, M. (1991): "El Sector Exterior y la Incorporación de España a la CEE. Análisis a partir de Funciones de Exportaciones e Importaciones", in Molinas, C., Sebastián, M. and Zabalza, A. (eds.): La Economía Española: Una Perspectiva Macroeconómica, Barcelona: Antoni Bosch, pp. 209-303.
2. García Delgado, J.L. et al. (1993): España, Economía. Madrid: Espasa Calpe.
3. García Perea, P. (1991): "Obtencion de series históricas de algunas variables clave del mercado de trabajo a partir de la Contabilidad Nacional", in Bentolila, S. and Toharia, L. (eds.): Estudios de Economía del Trabajo en España III. El Problema del Paro, Madrid: Ministerio de Trabajo y Seguridad Social, pp. 1263-1287.
4. Helliwell, J.F. (1993): "Macroeconometrics in a Global Economy", American Economic Review, Vol. 83, pp. 294-299.
5. Hendry, D. and Richard, J. F. (1983): "The econometric analysis of time series", International Statistic Review, Vol. 51, pp. 111-163.
6. Herce, J.A. and Sosvilla-Rivero, S. (1994): "HERMINS3: A Trhree-Sector Structural Model of the Spanish Economy for the Analysis of Community Support Frameworks", Documento de Trabajo 94-01, FEDEA, Madrid.
7. INE (1992): Contabilidad Nacional. Serie enlazada 1964-1991 - Base 1986, Madrid: Instituto Nacional de Estadística.
8. Layard, R., Nickell, S. and Jackman, R. (1991): Unemployment, Macroeconomic Performance and The Labour Market, Oxford. Oxford University Press.
9. Modesto, L. and Neves, P.D. (1993): "HERMINP4: A Four-Sector Structural Model of the Portuguese Economy", Technical Paper submit to the European Commission.
10. Molinas, C., Sebastián, M. and Zabalza, A. (1991): La Economía Española: Una Perspectiva Macroeconómica, Barcelona: Antoni Bosch.
11. Sebastián, M. (1991): "Un Anaálisis Estructural de las Exportaciones e Importaciones Españolas: Evaluación del Período 1989-1991 y Perspectivas a Medio Plazo", Información Comercial Española, Num. 699, pp. 9-23.
12. Taylor, J.B. (1993): "The Use of New Macroeconometrics for Policy Formulation", American Economic Review, Vol. 83, pp.300-305.
13. Wallis, K. F. (1991): "On Macroeconomic Policy and Macroeconometric Models", The Economic Record, Vol. 69, pp. 113-130.

APPENDIX 1

THE SIMPC VERSION OF THE HERMIN-S4 FOUR-SECTOR MODEL OF SPAIN

(T) : Tradable sector - Industry (excluding energy) (N) : Non-tradable sector - Private services, building and construction and energy (A) : Agriculture sector - Agriculture, forestry & fishing (G) : Public sector - Public administration, health and education

FEDEA: May 11, 1994

Exogenous variables in the model are as follows:

APENDIX 2: Variables Description

The table below lists all the variables defined in HERMINS3. "Type" reffers to identities (I), behavioural (B) or exogenous (E) variables. Those variables for which no source is indicated are obtained through an identity in the TSP data generation file irrespectively of the role (I, B or E) they play in the model. For example PC (the private consumption deflator) is a behavioural variable in the model, however, the data generation file defines it as CONSV/CONS. These two variables have proper sources in the National Accounts. Nevertheless, in the model, CONSV is obtained through an identity.

APPENDIX 3: Computation of FDDWOT and FDWON

Final domestic demand (FDD) and final demand (FD) are in HERMIN-S4 the major determinants of, respectively, output in the N and T sectors. In order to derive measures of FDD and FD that adequately reflect the sectoral output content of each of their components, we have computed two variables, FDDWOT and FDWON, defined as follows:

\[\mathrm{FDDWOT} = \mathrm{a} _ {1} ^ {*} \text {CONS} + \mathrm{a} _ {2} ^ {*} \mathrm{G} + \mathrm{a} _ {3} ^ {*} \mathrm{IBC} + \mathrm{a} _ {4} ^ {*} \mathrm{IME}\]

\[\text { FDWMON } = b _ {1} ^ {*} \text { CONS } + b _ {2} ^ {*} \text { G } + b _ {3} ^ {*} \text { IBC } + b _ {4} ^ {*} \text { IME } + b _ {5} ^ {*} \text { XP }\]

where the a and b coefficients have been derived from the Input/Output Tables of the Spanish economy for 1987 (TIO, 1991) according to the following expression:

\[\mathrm{PI} ^ {*} (\mathrm{I-A}) ^ {- 1} * \mathrm{FD}\tag{1}\]

in which:

PI is the matrix of primary inputs coefficients for the A, T, N and G sectors. Primary inputs have been simplified to three: Value-added (VA), Imports (MP) and Taxes on production and imports net of subsidies (Net Tax). Only the first, in the T and N sector, are however relevant for computation of the a and b coefficients.

(I-A) is the Leontieff inverse matrix.

FD is the final demand matrix where rows represent sectors and columns FD components. Each element of the matrix has been divided by its corresponding column total.

The result of computing expression (1) above is shown in the table below. Each figure amounts to the value-added, imports or taxes (net of subsidies) generated, in the corresponding sector, by a unit increase in each of the final demand components.

DOCUMENTOS DE TRABAJO

References

1. 93-01: "¿Son las Cajas y los Bancos estratégicamente equivalentes?, Juan Coello.

References

1. 93-03: "Indiciación salarial y empleo: un análisis desagregado para el caso español", María Draper.

References

1. 93-04: "The productivity effects of fixed term employment contracts: are temporary workers less productive than permanent workers?", Juan F. Jimeno and Luis Toharia

References

1. 93-05: "The determinants of labour mobility in Spain: who are the migrants?", Luis Albériko Gil and Juan F. Jimeno

References

1. 93-06: "A survey of recent applied macroeconomic and modelling research on the Spanish economy", José A. Herce y Simón Sosvilla-Rivero.

References

1. 93-07: "Infraestructuras", Maria Draper y José A. Herce.

References

1. 93-08: "Los Servicios de transporte aéreo, marítimo y terrestre: estructura económica y regulación", Ginés de Rus.

References

1. 93-09: "Situación actual, resultados y perspectivas del sector de las telecomunicaciones en España", Francisco Caballero.

References

1. 93-10: "Estructura y Regulación del sistema sanitario Español", Guillem López i Casasnovas.

References

1. 94-01: "HERMINS3, A three-sector structural model of the Spanish Economy for the analysis of Community Support Frameworks", José A. Herce y Simón Sosvilla-Rivero.

References

1. 94-02: "El mercado de depósitos a la vista en España: Banco Vs Cajas de Ahorro", Juan Coello.

References

1. 94-03: "An Econometric Analysis of Foreign Direct investment in Spain, 1964-89", Oscar Bajo-Rubio y Simón Sosvilla-Rivero.

References

1. 94-04: "The management of redundancies in Spain: Econometric report", Juan F. Jimeno y Luis Toharia.

References

1. 94-05: "Modelling international capital movements in the Spanish economy: A portfolio-balance approach". Oscar bajo y Simón Sosvilla-Rivero.

References

1. 94-06: "Demanda de tráfico telefónico nacional en España 1985-1989: Un estudio econométrico con datos de panel provinciales", Teresa Garín.

References

1. 94-07: "Demanda de tráfico telefónico internacional en España 1985-1989: Un estudio econométrico con datos de panel provinciales", Teresa Garín.

References

1. 94-08: "HERMIN-S4, a four-sector structural model of the Spanish economy for the analysis of COMMUNITY SUPPORT FRAMEWORKS", Simón Sosvilla-Rivero y José A. Herce.