Growth and the Welfare State in the EU: A causality analysis*
by José A. Herce Simón Sosvilla-Rivero Juan J. de Lucio
DOCUMENTO DE TRABAJO 98-12
June, 1998
The authors would like to thank Daniele Franco, former advisor at DG II of the European Commission, and Teresa Bento, at EUROSTAT, the supply of complete data from the SEepROS base. Remaining shortcomings are our own responsibility
** FEDEA and Universidad Complutense de Madrid
*** FEDEA and Universidad Complutense de Madrid
**** FEDEA and Universidad de Alcalá de Henares
Abstract
In this paper, we test for causality between GDP and social protection expenditures. To that end, we apply Hsiao's (1981) sequential procedure to data for 12 EU countries during the 1970-1994 period. Our results suggest that, for Belgium, Germany, Ireland, Luxembourg, the Netherlands, Portugal, and Spain, causality runs only from social protection growth to GDP growth, while for Denmark, France, Greece, Italy, and the United Kingdom, no causality is found between social protection growth and GDP growth.
JEL Codes: C32, H53, O52
Key words: Social Protection Benefits and Growth, Causality, European Union
1. INTRODUCTION
Most of the member states of the European Union have a generalised and sophisticated Welfare State (WS, standing also for the plural), unparalleled in the world. Many years of a common economic and institutional framework have made these WS very much alike to each other so that, despite certain substantive differences, a “European model” WS is referred too often. Equally often, analysts blame or praise the WS, in Europe or elsewhere, for having strong implications for general economic performance or growth, although a conclusive answer to the implicit question above is not yet available.
WS are generally the result of complex historical evolutions, along with programmes designed to cope with emerging dependency situations have been staking on top of previous ones. Insurance considerations, however, have always accompanied the birth and development of the core programmes of the WS, like pensions or unemployment benefits. Very often, incomes granted by certain welfare programmes, are higher than those from work, once the tax system is taken into account. On the other hand, the insurance component may be seen as differed wage and economic security in case of predictable or unpredictable risk. As a result, the mixture of incentives and disincentives is difficult to disentangle. Last, but not least, the financing of such huge programmes puts strong pressure on personal incomes, company profits or expenditure what in turn affects behaviour by households and firms.
In a series of recent papers we have explored the relationship between economic growth and the WS in the EU, either simply searching for correlation between growth and the WS (Herce, Sosvilla-Rivero and de Lucio, 1998a) or trying to find evidence of convergence in WS expenditure-GDP ratios across
European countries (Herce, Sosvilla-Rivero, and de Lucio, 1998b). Our evidence points towards a positive correlation between the WS and growth and weak causality or convergence (catching-up) in social protection expenditures amongst European countries. These results however beg for a causality analysis given the intricate nature of this relationship. This is the objective of this paper. The outline of the remainder of the paper is as follows. Section 2 elaborates on the WS-growth relationship while reviewing some recent literature on it. The econometric procedure used to test for causality is briefly described in Section 3, and the empirical results are presented in Section 4. Some concluding remarks are offered in Section 5.
2. DOES THE WELFARE STATE PROMOTE GROWTH?
The Welfare State has, on a priory grounds, many implications for growth. Atkinson (1995, 1996), on which Chart 1 is based, discusses some of them as he addresses the lack of conclusiveness of empirical evidence on this relationship. Indeed, as we discussed in the previous section, the WS is a conglomerate of different targeted programmes, financed through social contributions levied on wages or general taxes on income or expenditure amounting to an important proportion of output. As we summarise in Chart 1, four arguments can be invoked to explain evidence collected from econometric studies for different countries and time periods: the disincentives, dependency, social asset and normal good arguments. Chart 1 is self-explanatory.
| Chart 1The role of the Welfare State (WS) on output growth (gy) | ||||
| Causality | ||||
| The WS causes gy | gy causes the WS | No causality | ||
| Correlation between WS and gy | negative | - High transfers and taxes cause poor growth (the disincentives argument) | - Poor growth implies larger transfers (the dependency argument) | - The mix of disincentives and dependency prevents one-way causation but correlation is clear- The economy catches-up and the WS reaches maturity |
| positive | - High transfers cause high growth (the social asset argument) | - High growth permits higher transfers (the normal good argument) | - The mix of incentives and normality prevents one-way causation but correlation is clear- Industrialisation first and globalisation latter increase growth and, at the same time, require a larger and more sophisticated WS | |
| No correla-tion | - The mix of incentives and disincentives has no definite effect on growth but causation is clear | - The mix of dependency and normality has no definite effect on growth but causation is clear | - Too many factors at play, in fact a complex mix of the various arguments contained in the shadowed cells, prevent a clear-cut pattern of causality and correlation between the WS and growth | |
Sources: Atkinson (1995) and Atkinson (1996) and own elaboration
As mentioned before, Herce, Sosvilla-Rivero and de Lucio (1998a) found a positive correlation between WS and growth although no causality test was performed and thus several of the arguments shown in Chart 1 can be applied simultaneously to justify this results, namely the social asset and the normal good arguments. This result contrasts with the negative role that government expenditure exerts on growth in studies like those by Landau (1985) or Hansson and Henrekson (1994). The first author finds however mixed evidence for transfer expenditures when total government outlays are split into different categories. If one focus in the social asset argument mentioned above, considering for instance the positive role that less inequality has on economic performance as found by Persson and Tabellini (1994) or González-Páramo (1994), it should not be difficult to justify our result provided that social protection programmes reduce inequality. The level of social protection is also important to this respect. McCallun and Blais (1987) find that social expenditure plays a positive role towards economic growth below a certain level and a negative one beyond it. Neither of these studies however, perform causality test. We now turn to the methodology we use in this paper in order to do so.
3. ECONOMETRIC METHODOLOGY
Granger's causality test is widely used to test for the relationship between two variables. However, the causality tests are sensitive to lag length and, therefore, it is important to select the appropriate lengths. Otherwise, the model estimates will be inconsistent and, therefore, it is likely we draw misleading inferences. In this paper, we use Hsiao's (1981) sequential method to test for causality, which combines the Akaike's final predictive error (FPE, from now on) and the definition of Granger causality.
Consider the following models,
\[X _ {t} = \alpha_ {0} + \sum_ {i = 1} ^ {m} \delta_ {i} X _ {t - i} + \varepsilon_ {t}\tag{[1]}\]
\[X _ {t} = \alpha_ {0} + \sum_ {i = 1} ^ {m} \delta_ {i} X _ {t - i} + \sum_ {j = 1} ^ {n} \gamma_ {j} Y _ {t - j} + \varepsilon_ {t}\tag{[2]}\]
where and are stationary variables [i. e., they are I(0) variables]. The following steps are used to apply Hsiao's procedure for testing causality:
(i) Treat as a one-dimensional autoregressive process (1), and compute its FPE with the order of lags varying from 1 to . Choose the order which yields the smallest FPE, say m, and denote the corresponding FPE as .
(ii) Treat X as a controlled variable with m number of lags, and treat Y as a manipulated variable as in (2). Compute again the FPE of (2) by varying the order of lags of Y from 1 to N, and determine the order which gives the smallest FPE, say n, and denote the corresponding FPE as .
(iii) Compare with [i. e., compare the smallest FPE in step (i) with the smallest FPE in step (ii)]. If , then Y is said to cause X. If , then X is an independent process.
(iv) Repeat steps (i) to (iii) for the Y variable, treating X as the manipulated variable.
When X and Y are not stationary variables, but they are first-difference stationary [i. e., they are I(1) variables] and they are cointegrated (see Dolado et al., 1990), it is possible to investigate the causal relationships from to and from to , using the following error correction models:
\[\Delta X _ {t} = \alpha_ {0} + \beta Z _ {t - 1} + \sum_ {i = 1} ^ {m} \delta_ {i} \Delta X _ {t - i} + \varepsilon_ {t}\tag{[3]}\]
\[\Delta X _ {t} = \alpha_ {0} + \beta Z _ {t - 1} + \sum_ {i = 1} ^ {m} \delta_ {i} \Delta X _ {t - i} + \sum_ {j = 1} ^ {n} \gamma_ {j} \Delta Y _ {t - j} + \varepsilon_ {t}\tag{[4]}\]
where is the OLS residual of the cointegrating regression . Note that, if and are I(1) variables, but they are not cointegrated, then in (3) and (4) is assumed to be equal to zero.
In both cases [i. e., and are I(1) variables, and they are or they are not cointegrated], we can use Hsiao's sequential procedure substituting with and with in steps (i) to (iv), as well as substituting expressions (1) and (2) with equations (3) and (4).
4. EMPIRICAL RESULTS
We have applied the methodology described in the previous section to data on the twelve EU countries existing before the enlargement to Austria, Finland and Sweden. We have used data on Gross Domestic Product (GDP) and Social Protection Benefits (SPB), both in per capita terms. Data on GDP and total population comes from Bell (1994) and have been extended to 1994 using OECD (1996) data. The data on SPB comes from EUROSTAT (1988 and 1996). Both GDP and SPB were expressed at constant 1985 prices, and then the local currencies converted to a common standard using the OECD (1996) purchasing power parity estimates. We have taken logarithms of both variables, so that in the tables that follow y and sp denote the logs of real GDP per capita and real SPB per capita, respectively.
As a first step, we tested for the order of integration of the variables y and sp by means of the Dickey-Fuller tests. The results, shown in Table 1, suggest that all the variables could be treated as first-difference stationary.
Second, we have tested for cointegration between the pair of variables y and sp for all the countries in our sample. As can be seen in Table 2, the Dickey-
Fuller test applied to the cointegrating residuals indicates that the null hypothesis of no cointegration cannot be rejected in all cases at the usual levels of significance.
Since y and sp are I(1) variables but they are not cointegrated, we tested for causality in first differences of the variables, with no error-correction term added [i. e., equations (3) and (4), with ]. Notice that, since we are dealing with logs of the variables, we would be testing Granger causality between the rates of growth of SPB and GDP. The resulting FPE statistics are reported in Table 3.
As can be seen, for Belgium, Germany, Ireland, Luxembourg, the Netherlands, Portugal, and Spain, and , suggesting Granger causality running from SPB growth to GDP growth. On the other hand, for Denmark, France, Greece, Italy, and the United Kingdom, and , and no Granger causality would be present between SPB growth and GDP growth.
In order to further check our results, we test whether the differences between RMEs suggested by the FPE statistics in Table 3 are statistically significant or not. To that end, we have considered the Williams-Kloot test for forecasting accuracy described in Williams (1959). Let and denote alternative forecasts of the variable , the Williams-Kloot test statistic is the t-ratio for the hypothesis that the coefficient on is zero in a regression of on . A significantly negative value implies that is statistically superior to that of (and vice versa). Therefore, we generated forecasts for and both considering only past values of the forecasted variable and considering in addition past values of the other variable.
The results are shown in Table 4. As one can see, the Williams-Kloot test suggests that for Belgium, Germany, Ireland, Luxembourg, the Netherlands, Portugal, and Spain, GDP growth can be better predicted by adding the information content of the SPB growth, rather than by past values of GDP growth alone. This is not the case for Denmark, France, Greece. On the other hand, forecasting accuracy for SPB growth rates cannot be gained by considering also the information content of GDP growth. Therefore, these results reinforce our earlier conclusion from Table 3.
5. CONCLUDING REMARKS
This paper has tried to establish whether the welfare state through social protection expenditure causes growth. This also implies a positive correlation between the WS and economic performance, although the opposite is not true. Our results point towards statistically significant Granger causality running from social protection expenditures towards growth, the latter being promoted by the former for seven countries out of twelve in the EU. We cannot associate a definite pattern for those countries where the WS seems to cause growth, for both sets of countries present a variety of characteristics related to size, composition and financing of the WS, as probit modelling of the relationships between these characteristics and the degree of causality leads us to conclude .
Based thus on our characterisation of the possible interactions between the WS and economic performance we would tend to focus on the relative dominance of the social asset argument over the normal good argument and, even more, over the disincentives and dependency arguments as the explanation of our results.
This does not imply that the other three arguments do not apply, but rather that their operation is overtaken by the social asset effect. Neither, given that our analysis is not able to discriminate why certain countries pass the causality test while the rest do not, we can claim that social protection promotes growth up to a size of the WS or not.
Some externalities favouring growth, however, seem to be present when a country has a developed system of social or collective protection. This is relevant, in our view, for the current debate about the privatisation of social security in Western countries. In particular, the important thing is that citizens have a sense of security solidly rooted on the existence of widespread institutions for collective security. Whether these institutions are safer or more efficient under public or private management is another issue. Privatisation of social security, if this is the final outcome of the efficiency comparison, should not diminish the protection of individuals against the standard contingencies covered by the Welfare State.
To save space we do not report the results here, but they are available from the authors upon request.
References:
Atkinson, A. B. (1995): “The Welfare State and Economic Performance”, National Tax Journal 48, 171-198.
Atkinson, A. B. (1996): “Growth and the Welfare State: Is the Welfare State Necessarily Bad for Economic Growth”, New Economy 3, 182-198.
Bell, B. (1994): “The CEP-OECD Data Set”, CEP Discussion Paper 118.
Dolado, J.J., Jenkinson, T. And Sosvilla-Rivero, S. (1990): “Cointegration and Unit Roots”, Journal of Economic Surveys 4, 249-273.
EUROSTAT (1988): Social Protection, Current Expenditure and Receipts 1970-1985, Mimeo.
EUROSTAT (1996): Social Protection Expenditure and Receipts 1980-94, Luxembourg: Office des Publications Officielles des Communautés Européennes.
González-Páramo, J. M. (1994): “Gasto Social y Crecimiento Económico en el Estado del Bienestar”, Hacienda Pública Española Monografía 2, 135-153.
Granger, C. W. J. (1988): “Some Recent Developments in a Concept of Causality”, Journal of Econometrics 39, 199-211.
Hansson, P. and Henrekson, M. (1994): “A New Framework for Testing the Effect of Government Spending on Growth and Productivity”, Public Choice 81, 381-401.
Herce, J. A., Sosvilla-Rivero, S. and de Lucio, J. J. (1998a): “Social Protection Benefits and Growth: Evidence from the European Union”, Working Paper 98-01, FEDEA.
Herce, J. A., Sosvilla-Rivero, S. and de Lucio, J. J. (1998b): “A time-series Examination of Convergence in Social Protection across EU Countries”, Working Paper 98-10, FEDEA
Hsiao, C. (1981): “Autoregressive Modelling and Money-income Causality Detection”, Journal of Monetary Economics 7, 85-106.
Landau, D. L. (1985): “Government Expenditure and Economic Growth in the Developed Countries: 1952-76”. Public Choice 47, 459-477.
McCallum, J. and Blais, A. (1987): “Government, Special Interest Groups and Economic Growth”. Public Choice 54, 3-18.
OECD (1996): National Accounts. Main Aggregates 1960-1994, Paris: OECD.
Persson, T. and Tabellini, G. (1994): “Is Inequality Harmful for Growth?”. American Economic Review 84, 600-621.
Williams, E. J. (1959): Regression analysis, New York, Wiley.
| Table 1: Dickey-Fuller tests for unit roots | ||||||
| $\Delta y$ | $\Delta sp$ | |||||
| $\tau_{\tau}$ | $\tau_{\mu}$ | $\tau$ | $\tau_{\tau}$ | $\tau_{\mu}$ | $\tau$ | |
| Belgium | $-3.7728^b$ | $-3.1836^b$ | $-2.5648^b$ | $-3.6531^b$ | $-3.4789^b$ | $-2.8134^a$ |
| Denmark | $-4.9717^a$ | $-3.5630^b$ | $-2.7671^a$ | $-3.6670^b$ | $-3.7520^b$ | $-2.5528^b$ |
| France | $-3.9715^b$ | $-3.1294^b$ | $-2.2194^b$ | $-4.5424^b$ | $-3.5015^b$ | $-2.3661^b$ |
| Germany | $-4.0479^b$ | $-3.6077^b$ | $-2.7701^a$ | $-3.8899^b$ | $-3.6022^b$ | $-2.6019^b$ |
| Greece | $-3.6784^b$ | $-2.9183^c$ | $-2.0494^b$ | $-4.3229^b$ | $-3.8542^b$ | $-1.9791^b$ |
| Ireland | $-3.7691^b$ | $-2.9920^c$ | $-2.0470^b$ | $-3.7525^b$ | $-3.1193^b$ | $-2.7916^a$ |
| Italy | $-3.7411^b$ | $-3.3298^b$ | $-2.5132^b$ | $-3.7231^b$ | $-3.1541^b$ | $-2.5715^b$ |
| Luxembourg | $-3.7603^b$ | $-3.1384^b$ | $-1.9804^b$ | $-3.8379^b$ | $-3.4246^b$ | $-2.1968^b$ |
| Netherlands | $-3.8820^b$ | $-3.5334^b$ | $-2.2115^b$ | $-3.7897^b$ | $-3.3797^b$ | $-2.2615^b$ |
| Portugal | $-3.7503^b$ | $-3.1898^b$ | $-2.3871^b$ | $-3.9819^b$ | $-3.2753^b$ | $-2.1679^b$ |
| Spain | $-3.7544^b$ | $-3.4397^b$ | $-2.4664^b$ | $-4.2530^b$ | $-3.5601^b$ | $-2.3615^b$ |
| U.K. | $-3.7347^b$ | $-3.1599^b$ | $-2.7736^b$ | $-3.7995^b$ | $-3.8964^a$ | $-3.0439^b$ |
Note: a and b detone significance at the 5% and 1% level, respectively.
| y | sp | |||||
| $\tau_{\tau}$ | $\tau_{\mu}$ | $\tau$ | $\tau_{\tau}$ | $\tau_{\mu}$ | $\tau$ | |
| Belgium | -1.5537 | -2.2454 | 0.7866 | -2.6459 | -0.3882 | -1.3593 |
| Denmark | -1.1831 | 0.4375 | -1.0182 | -1.9535 | 0.2184 | -1.0858 |
| France | -1.9417 | -2.0766 | 0.1076 | -1.6477 | -1.9360 | -1.1879 |
| Germany | -0.5875 | -1.6478 | 0.9018 | -1.5168 | -2.0515 | 1.3514 |
| Greece | -1.5355 | -0.9652 | 1.8162 | 0.0086 | -1.6554 | -0.0581 |
| Ireland | -0.0585 | -1.0582 | -0.2881 | -1.4649 | -0.4656 | -1.1669 |
| Italy | -0.6404 | -1.7429 | 1.2185 | -1.0331 | -2.0823 | 0.4747 |
| Luxembourg | -2.0803 | 0.5902 | -1.2439 | -1.6264 | -0.1557 | -1.0982 |
| Netherlands | -2.8820 | 0.4561 | -0.9716 | 0.1692 | -2.1439 | -1.3141 |
| Portugal | -1.5738 | -1.6989 | 1.0680 | -1.3516 | -1.4823 | 0.0488 |
| Spain | -2.2466 | -1.4794 | 0.2799 | -2.2473 | -1.5012 | -0.1912 |
| U.K. | -1.4018 | -1.9610 | 0.7741 | -1.6099 | -1.0104 | -0.8105 |
| Table 2: Cointegration tests | ||
| CRADF | CRDW | |
| Belgium | -0.9267 | 0.1725 |
| Denmark | -1.5708 | 0.2756 |
| France | -2.1539 | 0.2939 |
| Germany | -1.9660 | 0.2916 |
| Greece | -0.3086 | 0.2441 |
| Ireland | -1.8121 | 0.1058 |
| Italy | -0.2959 | 0.2219 |
| Luxembourg | -2.0548 | 0.3566 |
| Netherlands | -1.2421 | 0.1470 |
| Portugal | 0.2230 | 0.1286 |
| Spain | -2.1554 | 0.4370 |
| U.K. | -1.7594 | 0.1440 |
| Note: CRADF and CRDW are the are the Augmented Dickey-Fuller Statistic and the Durbin-Watson statistic for the cointegrating residuals (see Dolado et al., 1990). Their critical values at the 5% significance level are -3.29 and 0.78, respectively | ||
| Table 3: FPE STATISTICS | ||||||
| $FPE_{\Delta y}$ (m,0) $X10^{-3}$ | $FPE_{\Delta y}$ (m,n) $X10^{-3}$ | Comment | $FPE_{\Delta sp}$ (m,n) $X10^{-3}$ | $FPE_{\Delta sp}$ (m,o) $X10^{-3}$ | Comment | |
| Belgium | 1.3328 | 1.2888 | c. | 1.7787 | 1.8062 | n.c. |
| Denmark | 0.5762 | 0.6263 | n.c. | 0.7789 | 0.8723 | n.c. |
| France | 0.6640 | 0.6988 | n.c. | 1.0951 | 1.1543 | n.c. |
| Germany | 1.1182 | 1.0885 | c. | 3.0218 | 3.5586 | n.c. |
| Greece | 4.9808 | 5.4126 | n.c. | 9.5308 | 10.9301 | n.c. |
| Ireland | 3.2917 | 2.9089 | c. | 6.4222 | 6.6511 | n.c. |
| Italy | 1.6516 | 1.7917 | n.c. | 3.0176 | 3.2017 | n.c. |
| Luxembourg | 1.4039 | 1.0754 | c. | 9.7720 | 10.6420 | n.c. |
| Netherlands | 0.7582 | 0.5515 | c. | 1.3517 | 1.4520 | n.c. |
| Portugal | 3.5000 | 3.0472 | c. | 11.4500 | 15.5260 | n.c. |
| Spain | 1.0500 | 0.8755 | c. | 3.6705 | 3.7867 | n.c. |
| U.K. | 2.4344 | 2.5436 | n.c. | 2.5940 | 2.7238 | n.c. |
| Table 4: Willian-Kloot tests | |
| $\Delta y = \alpha + \sum_{i=1}^{m} \delta_i \Delta y_{t-i}$ vs $\Delta y = \alpha + \sum_{i=1}^{m} \delta_i \Delta y_{t-i} + \sum_{j=1}^{n} \gamma_j sp_{t-j}$ | |
| Belgium | -2.6615 |
| Denmark | 3.9734 |
| France | 2.1447 |
| Germany | -3.4315 |
| Greece | 2.1573 |
| Ireland | -2.8462 |
| Italy | 2.4852 |
| Luxembourg | -2.6054 |
| Netherlands | -3.9351 |
| Portugal | -3.6344 |
| Spain | -2.7915 |
| U.K. | 2.0495 |
| $\Delta y = \alpha + \sum_{i=1}^{m} \delta_i \Delta y_{t-i}$ | vs $\Delta y = \alpha + \sum_{i=1}^{m} \delta_i \Delta y_{t-i} + \sum_{j=1}^{n} \gamma_j sp_{t-j}$ | |
| Belgium | 2.6152 | |
| Denmark | 3.1272 | |
| France | 2.1548 | |
| Germany | 2.3331 | |
| Greece | 2.0511 | |
| Ireland | 2.7393 | |
| Italy | 3.1956 | |
| Luxembourg | 3.9187 | |
| Netherlands | 2.8049 | |
| Portugal | 2.5038 | |
| Spain | 2.6048 | |
| U.K. | 2.4534 | |
COLECCION RESUMENES
98-01: “Negociación colectiva, rentabilidad bursátil y estructura de capital en España”, Aljenadro Inurrieta.
TEXTOS EXPRESS
98-02: “Sector turístico y cremiento del empleo en la Comunidad Autónoma de Canarias: Un ejercicio de prospección al horizonte 2011”, José A. Herce y Simón Sosvilla.
98-01: “El gasto sanitario en España: Evolución reciente y perspective”, Javier Alonso y José A. Herce.
DOCUMENTOS DE TRABAJO
98-12: Growth and the Welfare State in the EU: A casualty analysis" José A. Herce, Simón Sosvilla-Rivero y Juan J. de Lucio.
98-10: A time-series examination of convergence in social protection across EU countries", José A. Herce, Simón Sosvilla-Rivero y Juan J. de Lucio.
98-09: “Estructura Demográfica y Sistemas de Pensiones. Un análisis de equilibrio general aplicado a la economía española”, María Montero Muñoz.
98-08: “Earnings inequality in Portugal and Spain: Contrasts and similarities”, Olga Cantó, Ana R. Cardoso y Juan F. Jimeno.
98-07: “Labour reallocation and labour market institutions: Evidence from Spain”, Carlos García-Serrano y Juan F. Jimeno.
98-06: “Benchmark priors for Bayesian model averaging”, Carmen Fernández, Eduardo Ley, Mark F. J. Steel.
98-05: “The effects of externatilites on value added and productivity growth in Spanish industry”, Juan J. de Lucio, José A. Herce y Ana Goicolea.
98-04: “Employment segmentation, labour mobility and Mismatch: Spain, 1987-1993”, Sonsoles Castillo, Juan F. Jimeno and Omar Licandro.
98-03: “Un análisis global, regional y sectorial de los efectos externos de conocimiento”, Juan José de Lucio.
98-02: “A tale of two neighbour economies: Does wage flexibility make the difference between Portuguese and Spanish unemployment?”, Sonsoles Castillo, Juan J. Dolado y Juan F. Jimeno.
98-01: “Social protection benefits and growth: Evidence from the European Union”, José A. Herce, Simón Sosvilla-Rivero y Juan J. de Lucio.