Portugal: Selected Issues and Statistical Appendix

This Selected Issues paper and Statistical Appendix examines Portugal’s real convergence experience, and highlights the growth payoff of reform policies in determining the country’s rapid growth during most of the past decade. The paper identifies potential areas for reform by examining developments in the unemployment–vacancy relationship since the early 1980s. The paper highlights that an issue for Portugal as a prospective European Monetary Union member is that of the strength of its financial links with Spain. The paper also examines such linkages as derived from daily data on short-term interest rates.

Abstract

This Selected Issues paper and Statistical Appendix examines Portugal’s real convergence experience, and highlights the growth payoff of reform policies in determining the country’s rapid growth during most of the past decade. The paper identifies potential areas for reform by examining developments in the unemployment–vacancy relationship since the early 1980s. The paper highlights that an issue for Portugal as a prospective European Monetary Union member is that of the strength of its financial links with Spain. The paper also examines such linkages as derived from daily data on short-term interest rates.

II. An Unemployment Vacancy Analysis of the Portuguese Labor Market1

Unemployment in Portugal has long been amongst the lowest in the EU; it has however risen significantly since 1992 (Chart 1, upper panel). Whilst part of this rise is undoubtedly attributable to the cyclical consequences of the 1993 recession, there have also been concerns that the Portuguese labor market may have started to develop structural problems similar to those found in other EU countries. These concerns are heightened by the sharp rise in both youth and long-term unemployment (Chart 1, lower panel), and by signs of an increase in the persistence of unemployment. Work by the staff during last year’s consultation, the OECD in its 1996 Economic Survey of Portugal, and other ongoing research projects have attempted to ascertain the extent to which structural unemployment may have increased in Portugal in recent years—with somewhat contrasting results. This chapter seeks to shed further light on this issue by examining the unemployment vacancy relationship in Portugal.

CHART 1
CHART 1

PORTUGAL Unemployment

(In percent)

Citation: IMF Staff Country Reports 1996, 129; 10.5089/9781451832099.002.A002

Source: Instituto Nacional de Estatistica.1/ Proportion of total unemployed who have been unemployed for a year or more.2/ Proportion of 15-24 year-olds who are unemployed.

The relationship between vacancies and unemployment is often used as a tool to distinguish between structural and cyclical causes of unemployment. Cyclical effects are captured by movements along the unemployment vacancy (UV) curve.2 When an economy is in recession, the rate of job loss increases and firms are unwilling to hire new workers. Accordingly, unemployment is high and the number of vacancies is low. In contrast, when an economy is enjoying an upswing, unemployment declines, and firms have difficulties filling vacancies. Thus, the cyclical relationship between vacancies and unemployment will be a stable, negative one reflected in movements along the UV curve.

Shifts of the UV curve illustrate how efficiently the labor market matches unemployed workers with available vacancies (Pissarides, 1986; Blanchard, 1989). If the labor market works efficiently, unemployed people find jobs quickly, and one would expect a low vacancies rate3 to be correlated with low levels of unemployment. The efficiency of the labor market may deteriorate for a number of reasons: a reduction in the search effectiveness of the unemployed; a greater selectivity on the part of employers; or an increase in labor market mismatch. In the latter case, whilst unemployed people may be effectively searching for work and employers may have jobs to fill, hiring may still fail to occur if unemployed people do not match the job requirements of those firms that have vacancies. Mismatch can occur because workers may not have the appropriate skills (a problem of occupational mismatch) or because unemployed people may not live in those regions where firms are expanding (giving rise to regional mismatch). In all of the above cases, the number of vacancies for any given number of unemployed people will rise, leading to an outward shift of the UV curve.

This chapter highlights two main characteristics of the UV relationship in Portugal since the early 1980s. First, by estimating a simple UV relationship, it is established that this relationship has not been stable over the last 12 or so years. Second, a more general UV relationship is estimated which includes proxy variables for the search intensity of the unemployed, and for structural change in the labor market. The empirical results strongly suggest that the relationship has been subject to two structural shifts. The first inward shift started in the mid-1980s and appears to be closely correlated with the structural impact of Portugal’s entry into the European Community. However this improvement in labor market efficiency gradually subsided and by the early 1990s a second, outward shift of the UV curve occurred. This shift followed changes in the unemployment benefit regime, which loosened eligibility, and increased the level and duration of benefits, with an adverse impact on the search effectiveness of the unemployed.

A. The Unemployment Vacancy Relationship

The change in unemployment (U) between period t and t+1 is equal to inflows into unemployment (It), less outflows from unemployment denoted by (Ot):

Ut+1Ut=ItOt(1)

Both inflows and outflows can be decomposed further. Inflows into unemployment are the sum of people who were previously employed and who have lost their jobs, denoted by Ite, and new entrants to the labor market, such as school leavers in search of their first job. This latter group is denoted by Itn. Thus, we can write:

It=Ite+Itn(2)

Outflows from unemployment can be decomposed into those people who leave unemployment because they have found a new job Ote and those workers who leave unemployment and exit the labor market Otl. So total outflows can be written as:

Ot=Ote+Otl(3)

Applying these two disaggregations to equation (1) gives:

Ut+1Ut=(Ite+Itn)(Ote+Otl)=(IteOte)+(ItnOtl)(4)

An outflow from unemployment into employment (Ote) requires a successful matching of a worker with a vacancy. In turn the number of successful job matchings is assumed to be determined by three factors: the search effectiveness of the unemployed (ctut), the selectivity of employers (dtvt), and the degree of labor market mismatch (Mt).

Ote=f(ctut,dtvt,Mt)f1>0,f2>0,f3<0(5)

where ut = Ut/Lt represents the unemployment rate, vt = Vt/Nt is the vacancy rate, Lt is the total labor force, and Nt is employment. It is assumed that in each period only a proportion ct of the unemployed are actively engaged in job search. Thus, the number of effective job-seekers is given by ctUt with ctut being the “effective” unemployment rate.

Outflows from unemployment and into employment are also determined by the selectivity of employers. For a given level of search effectiveness on the part of the unemployed, the probability of an unemployed person finding a job is greater, the larger the number of vacancies offered by employers. Therefore, the number of outflows from unemployment is an increasing function of vacancies. However, for a given level of vacancies, increased selectivity on the part of employers will reduce outflows. Selectivity is a somewhat nebulous concept to model. However, Pissarides (1986) approximated the selectivity of employers by assuming that only a fraction—dt—of the available job vacancies go into the job-matching process. Therefore, dtvt measures the number of “effective” vacancies.

Both outflows from unemployment which leave the labor market, and inflows to unemployment by those people entering the labor market, are primarily due to demographic and social factors. For example, school leavers looking for their first job, and mothers returning to work after raising their families, would constitute the majority of those inflows to unemployment who have just entered the labor market. People who both leave unemployment and exit the labor market typically do so because they are retiring, or they obtain other non labor market related welfare benefits. Therefore, we can capture these factors in a generic function:

(ItnOtl)=h(Dt)(6)

where Dt is a vector of social and demographic variables.

In steady state equilibrium unemployment is constant. This gives the following implicit relationship:

Itef(ctut,dtvt,Mt)+h(Dt)=0(7)

Inflows to unemployment which have just entered the labor market—Ite—are assumed to be constant.4 Equation (7) can be inverted to give a relationship between unemployment and vacancies, search intensity, employer selectivity, labor market mismatch, and demographic factors:

ut=g(vt,ct,dt,Mt,Dt)(8)

Bowden (1980) showed that any change in either aggregate demand or in real wages will result in movements along the UV curve, whilst changes in any other variable will cause the UV curve to shift.

In order to move to an empirical model, equation (8) is approximated by a log linear functional form. However, there is a further problem. Many of the factors which would shift the UV curve (changes in the search intensity of the unemployed, increased selectivity by employers, and labor market mismatch) are not directly observable. Observable proxy variables must be used which will hopefully capture the impact of these underlying unobservable variables which shift the function. Our estimated equation is:

LN(UL)t=θ0+θ1LN(VN)t+Σi=1NΨiZit+γ1t+γ2t2+εt(9)

where Zit is a vector of proxy explanatory variables which capture search intensity, demographic factors, and labor market mismatch. Details of the variables included in Zit are set out in the appendix to this chapter.

Search intensity is proxied by two variables: the benefits replacement ratio (Z1) and the number of the long-term unemployed as a proportion of total unemployment (Z2). Increases in the ratio of unemployment benefits relative to average wages reduces the opportunity cost of unemployment, increases the reservation wage and thereby reduces effective job search. For a given number of vacancies, the number of successful job matches in any period will be reduced, shifting the UV relationship outward. Therefore, one would expect a positive estimated coefficient for the benefits replacement ratio when it is included in equation (9). The hypothesis that changes in the unemployment benefits regime might have contributed to rising unemployment seems plausible. In the late 1980s Portugal enacted a series of changes in benefit regulations which increased the coverage and generosity of unemployment benefits.5 As a consequence, government expenditure on unemployment benefits as a proportion of GDP rose sharply in the early 1990s—well above the levels recorded during the previous peak in unemployment in the mid-1980s (Chart 2, bottom panel).

CHART 2
CHART 2

PORTUGAL Unemployment

Citation: IMF Staff Country Reports 1996, 129; 10.5089/9781451832099.002.A002

1/ Based on relative unit labor costs.

The extent of mismatch in Portugal is difficult to assess because of data limitations. For example, regional unemployment data is unavailable before 1988 and so it is not possible to construct the type of mismatch indices favored by Layard and others (1991). One potential measure of occupational mismatch might be the variance of unemployment by age (Z3). However, benefit duration in Portugal is determined by the age of the unemployed person, with older workers eligible for longer periods of benefits. If effective benefit duration increases, one would expect workers who are older to remain unemployed longer, and so the variance of unemployment by age would increase. This may create a potential bias because this variable might also pick up duration effects as well as mismatch effects.

Our model of unemployment flows shows that demographic factors will affect inflows/outflows from those workers who have not previously participated in the labor force. As a consequence, the steady state relationship, given by equation (8), will also be affected by demographic factors. Female labor force participation as a proportion of total labor force (Z4) is used to proxy this effect.

The real effective exchange rate (Z5) is included to assess the impact of external competitiveness upon the labor market. Papers by Duffy (1984) and Jones and Manning (1990) argue that a rise in the real exchange rate would shift the UV curve outwards. An appreciation of the real exchange rate, other things being equal, would reduce the demand for labor in those sectors most exposed to international competition. Firms facing such pressures would attempt to improve productivity by cutting their workforce. This would lead to a lower employment retention rate and a reduced hiring rate.

A dummy variable for Portugal’s accession to the European Union is also included in the regression. Finally, linear and quadratic time trends are included in the regression to pick up any additional structural change in the labor market which is not picked up by any of the other variables included in the regression.6 Whilst these variables cannot be used to attribute the source of any structural change, they can determine both the sign and magnitude of any omitted variable, and the timing of its impact upon the UV relationship.

B. The UV Relationship in Portugal

The most significant investigation of the UV relationship in Portugal was by Luz and Pinheiro (1994) who estimated a simple error correction model. They found a stable relationship for the period 1983q4 to 1994q1 by including lagged changes of unemployment and current changes in vacancies. Chart 2 (upper panel), which illustrates the UV relationship for Portugal since 1983, suggests that between 1983 and 1986 there was indeed a relatively stable negative relationship between the two variables, with the data points for these years bunched closely together. After 1986, however the relationship takes a very clear anticlockwise journey. Initially, vacancies rise and unemployment falls. Between 1989 and 1991 unemployment continues to fall, but the trend in vacancies is reversed. After 1992, the vacancy rate falls, and unemployment rises.

In order to assess whether or not the unemployment vacancy relationship is stable in Portugal, the restrictions Ψi = 0 and λi = 0 (which imply no change in the efficiency of the labor matching process) are applied to equation (9) to give the following equation: Parameter estimates and test statistics for equation (10) are given in Table 5. The autocorrelation and heteroscedasticity statistics are highly significant and strongly suggest functional misspecification. A structural stability test, which splits the sample in 1990, rejects the hypothesis that the equation is stable over the entire sample. The plot of the actual versus fitted values from this equation confirms the presence of a structural break during this period (Chart 3). The equation initially tracks the downward shift in unemployment, albeit with a consistent bias. However the regression fails to capture the continued fall in unemployment after 1989. Thus it is clear both from the econometric results and from the cross plot that the UV relationship has shifted significantly.

Table 5.

Portugal: Simple Vacancy Unemployment Relationship

IV Estimation, Sample 1984(2) to 1995 (4) Dependent Variable is Ln(U/L)t

article image
(t statistics in parenthesis)Endogenous Ln(V/N)tInstruments Ln(V/N)t-1, Ln(V/N)t-2, Ln(V/N)t-3, Ln(V/N)t-4
CHART 3
CHART 3

PORTUGAL Unemployment Vacancy Relationship Actual and Fitted Values for Equation 10

Citation: IMF Staff Country Reports 1996, 129; 10.5089/9781451832099.002.A002

Source: Staff estimates.
LN(UL)t=θ0+θ1LN(VN)t+ωt(10)

C. Empirical Results

Table 6 presents a number of alternative estimated specifications for equation (9). In order to avoid potential simultaneity problems, the vacancy variable (Ln(V/N)t) and the long-term unemployment variable (Z2) were instrumented by lagged values of these variables. None of the diagnostic tests suggest that any of the equations suffer from misspecification. All regressions give significant, correctly signed, coefficients for the vacancy rate, the proportion of long-term unemployed, and the benefit replacement ratio (Z1). However, the estimates for the variance of male unemployment rates (Z3) do not seem particularly robust. The variable measuring female labor force participation (Z4) did not seem to have any explanatory power.

Table 6.

Portugal: Estimated Unemployment Vacancy Relationship

IV Estimation, Sample is 1984(2) to 1995 (4) Dependent Variable is Ln(U/L)t

article image
t statistics in parentheses.Endogenous variables: Ln(V/N)t, Z1tInstruments: Ln(V/N)t-1, Ln(V/N)t-2, Ln(V/N)t-3, Ln(V/N)t-4Z1t-1, Z1t-2, Z1t-3, Z1t-4

The most interesting results were the estimates for the real exchange rate (Z5). Contrary to the a priori hypothesis, estimates of this coefficient were consistently negative across all equations. Chart 2 (middle panel) illustrates the point that there does seem to be a negative correlation between the unemployment rate and the real exchange rate, which is inconsistent with the hypothesis that a decline in competitiveness has led to an increase in unemployment.

The linear and quadratic time trends proved to be significant in all those regressions where they were included. The sign on the coefficient for the linear trend is negative, capturing an inward shift in the UV relationship. However, the quadratic time trend is positive but with a much smaller coefficient. These time trends suggest that the UV curve shifted inwards but, as time progressed, the inward shift slowed and gradually stopped.

The dummy variable for EU membership (Z6) gave a positive significant coefficient. This result is contrary to what one would expect given that it is widely agreed that EU membership had a positive effect upon economic performance. However, a closer examination of its interaction with the linear and quadratic time trends explains this result. When the EU is included, as in equations (1a) and (2a) of Table 6, the coefficient on the time trend is -0.03. When the EU coefficient is excluded, as in equations (5a) and (6a), the coefficient on the time trend lies between -0.01 to -0.02. If the impact of both the time trend and the EU dummy are taken together, the negative effect of the time trend effect quickly dominates the positive effect of the EU. The inclusion of the EU dummy merely picks up a few outliers in the early part of the sample, which in turn affects the estimated coefficients on the time trend. The strong suspicion must be that the EU dummy variable, along with the linear and quadratic trends, are picking up the overall structural impact of Portugal’s accession to the EU. A dummy variable for two quarters on 1991 was included in some equations in order to eliminate large outliers found in 1991 which improve the overall fit of the regression.

In equations (4a) and (5a) the benefit replacement ratio is dropped in order to assess how robust are the estimates for the other regressors. The coefficient on the long-term unemployment variable increases but the other coefficients are more or less unchanged. In equations (6a) and (7a) the long-term unemployment variable is excluded. The variance of male unemployment becomes significant, suggesting that this variable is competing with the long-term unemployment variable. The most plausible explanation for this is that both variables are picking up the impact of benefit duration upon the UV relationship. Hence, the variance of male unemployment seems to be a biased, unreliable measure of occupational mismatch.

D. Conclusion

Econometric estimates of a simple UV curve reveal that this relationship has been unstable, suggesting that the Portuguese labor market has experienced appreciable structural change. The residuals from this regression suggest that there was a structural break at some point between 1989 and 1991. Estimation of a more general UV curve which includes proxy variables for search intensity, labor market mismatch, and structural change reveals that the UV has been subjected to two shifts—an inward shift which seems to be closely timed with Portugal’s entry into the EU and an outward shift which closely follows changes in the eligibility rules, level, and effective duration of unemployment benefits.

Part of the inward shift of the UV curve is picked up by the negative relationship between the real exchange rate and unemployment in all estimated regressions.7 The first explanation for this correlation is that these variables were jointly influenced by economic reforms associated with EU membership which have been omitted from the estimated equations. Trade liberalization and economic deregulation following EU accession encouraged a significant expansion of foreign direct investment. Foreign firms relocated their unskilled, labor intensive activities to Portugal where there is a relative abundance of cheap, unskilled labor. This created more unskilled vacancies, and so it became easier for unskilled Portuguese workers to find employment. The matching process became more efficient and the UV curve shifted inwards. As for the real exchange rate, both a capital inflow and the effects of trade liberalization led to a real appreciation. A second interpretation to the observed correlation between unemployment and the real effective exchange rate is that EU transfers contributed to the real exchange appreciation. EU membership gave Portugal access to EU transfers, which were used to develop infrastructure. Previously underdeveloped regions gained access to outside markets, reducing a potential source of regional mismatch. The empirical evidence presented here cannot distinguish between these potential explanations, and probably both made a significant contribution.

The outward shift of the UV curve in recent years is correlated with increases in both the benefit replacement ratio, and long-term unemployment, strongly suggesting that the search effectiveness of the unemployed has fallen. The joint significance of both these variables suggests that search intensity fell because both the level of benefits and effective benefit duration increased. The empirical evidence argues that the full impact of the changes in the 1989 law governing unemployment benefits eligibility were partly obscured by the beneficial effects of EU membership. The detrimental effects of the changes in the unemployment benefit system gradually dominated the structural effect, so that by the early 1990s the unemployment rate began to rise. Changes in benefit duration, which favor older, less skilled workers, inevitably encourage long spells of unemployment amongst older workers. Increased levels of benefits reduced the opportunity cost of unemployment for those workers eligible to claim unemployment benefit, while the eligibility criteria themselves were loosened.

This study suggests that reforming the current benefit system would improve the efficiency of the labor market. Such a reform could entail, first, the harmonization of benefit duration, so that all workers would be entitled to the same period of benefits. Second, the unemployment benefits replacement ratio could be reduced to the lower levels of the 1980s.8 The Portuguese government has introduced a pilot program for a guaranteed minimum income scheme and is planning to extend it nationwide in 1997. This offers the opportunity to reform the current welfare benefit system, including unemployment benefits, with a simpler system that provides an effective and targeted social safety net.

Data Description

Unemployment - OECD standardized unemployment rate.

Vacancies - The stock of job vacancies at the end of the period. Data supplied by I.N.E.

Z1 - The unemployment benefit replacement ratio - The data is supplied by the Ministerio para a Qualificação e Emprego and I.G.F.S.S., who produce an annual series for the unemployment insurance replacement ratio where unemployment benefit is expressed as a percentage of the average remuneration in the economy. This series was interpolated into a quarterly series by assuming a constant replacement ratio throughout all four quarters of the year. Data was only available after 1986. In order to avoid reducing the sample size, it was assumed that between 1983 and 1985 the replacement ratio was the same as in 1986. This assumption is supported by the fact that during 1983-1986 government expenditure on unemployment benefits as a proportion of GDP remained roughly stable.

Z2 - Long-term unemployment - The benefit replacement ratio does not adequately capture the influence of benefits duration upon unemployment rates. In theory, a high benefit replacement ratio could be consistent with low levels of unemployment if benefit duration were also low. In other words, generous but short lived eligibility may encourage short spells of unemployment which in the aggregate lead to a comparatively low level of unemployment.

One must also distinguish between the effective duration of benefits and the maximum duration of benefits. If a social security system requires that all unemployed people must undertake effective job search as a condition for receiving unemployment benefits, then people who do not actively search for jobs could lose their benefits before they reach the maximum duration. Effective duration will in large part depend upon the monitoring of the unemployed by the social security agency. Any reduction in monitoring is likely to result in an increase in effective duration.

In order to capture how effective benefit duration might influence the UV relationship, the number of people unemployed for more than 13 months as a proportion of the total unemployed is included in equation (9). Budd and others (1988) use this variable as a proxy measure for search intensity of the unemployment in their estimation of the UV curve for the United Kingdom. Increasing numbers of the long-term unemployed are interpreted as a decline in search intensity, and so a positive estimated coefficient is anticipated. It is argued here that this variable most probably picks up changes in search intensity due to changes in effective benefit duration. The data on long-term unemployment is suppled by I.N.E.

Z3 - The variance of male unemployment rates - Typically younger workers have undertaken more years of education than older workers. Therefore, one might expect a skill mismatch to be correlated with unemployment by age. In order to construct this variable the male labor force is divided by age into intervals of five years. For example, all males between the age of 21 and 25 participating in the labor force make up the first cohort. The unemployment rate for every cohort is calculated by dividing the number of unemployed in each cohort by the total labor force participation in that cohort. The variance of these age related unemployment rates is then calculated for every time period. Workers under 21 and over 55 are eliminated because unemployment rates in these cohorts are influenced by additional factors such as early retirement, greater training opportunities, etc. The number of unemployed by age and labor participation rates by age are taken from data provided by the Ministerio para a Qualificação e Emprego.

Z4 - Female labor force participation - Female labor force participation may influence the UV relationship because women are more likely to intermittently enter and exit the labor market than men. However, there are a number of other possible mechanisms whereby changes in female participation may influence the UV relationship. An increase in female labor supply, which was matched by an equal increase in labor demand, could still result in an outward shift in the UV curve if these workers did not have the skills necessary to fill these vacancies. Alternatively, if benefit eligibility initially differs across gender, then changes in benefits eligibility may encourage more women to enter the labor force in order to register as unemployed and obtain benefits. In order to test for any female labor supply effects, the ratio of the female labor force over the total labor force is included in the regression. This variable is created by taking the total number of females participating in the labor force and deflating it by the total labor force. The data source is the Ministerio para a Qualificação e Emprego.

Z5 - Real effective exchange rate - Data taken from the IMF International Financial Statistics.

Z6 - European Union Membership - A simple shift dummy dated at Portugal’s accession to the EU is used to see if the timing of EU membership has any explanatory power. However, a simple shift dummy will not pick up any gradual or delayed effects of membership, and so care must be taken when interpreting the empirical results related to this variable.

References

  • Blanchard, O., 1989, “Two Tools for Analyzing Unemployment,” NBER Working Paper No. 3168.

  • Bowden, R.J., 1980, “On the Existence and Secular Stability of U-V Loci,” Economics 47, pp. 33-50.

  • Budd, A., P. Levine, and P. Smith, 1988, “Unemployment, Vacancies and the Long-term Unemployed,” Economic Journal, 98, pp. 1071-91.

  • Duffy, M., 1984, “The Relationship Between Unemployment and Unfilled Vacancies in Great Britain: An Extended Job Search, Labour Turnover View,” Bulletin of Economic Research, 36, Vol. 2, pp. 143-172.

    • Search Google Scholar
    • Export Citation
  • Jackman, R., R. Layard, and C. Pissarides, 1989, “On Vacancies” Oxford Bulletin of Economics and Statistics, 51, Vol. 4, pp. 377-394.

    • Search Google Scholar
    • Export Citation
  • Jones, D. R., and D. N. Manning, 1990, “The Unemployment-Vacancy Relation in the UK,” Scottish Journal of Political Economy, 4, Vol. 37.

    • Search Google Scholar
    • Export Citation
  • Layard, R., S. Nickell, and R. Jackman, 1991, Unemployment: Macroeconomic Performance and the Labour Market, Oxford University Press.

  • Luz S., and M. Pinheiro, 1994, “Wage Rigidity and Job Mismatch in Europe: Some Evidence,” Bank of Portugal Working Paper, 2:94.

  • Pissarides, C., 1986, “Unemployment and Vacancies in Britain,” Economic Policy, 3, pp. 499-559.

1

Prepared by Jimmy McHugh.

2

The unemployment vacancy curve is sometimes referred to as the Beveridge curve.

3

The vacancy rate is normally defined as the number of vacancies available as a proportion of total employment.

4

This is a contentious assumption. If changes in unemployment are primarily due to inflows rather than outflows, then unemployment will move independently of vacancies, changes in the efficiency of the labor matching process, and labor mismatch. Few countries produce reliable, consistent inflow series. However, Layard and others (1991) report that inflow rates tend to be high in North America, and low in the EU, Scandinavia, and Japan. There is no reliable inflow data available in Portugal.

5

Portugal provides two types of benefits for the unemployed: the unemployment subsidy and the unemployment social subsidy. The value of both these benefits is tied directly to the minimum wage. The unemployment subsidy is a contributions-based benefit. The decree-law No. 79-A/89 which was passed in March 1989 eased the eligibility requirements and extended the duration of benefits. Under the previous law an unemployed person could only claim benefits if they had made contributions for at least 36 consecutive months. Under the 1989 law, an unemployed person would be eligible if they had made 18 months contributions over the proceeding 24 months. The 1989 benefits law introduced a sliding scale of benefit duration linked to the age of the worker. The minimum level of duration was set at 10 months for workers aged 25 or under, rising to 30 months for workers who are 55 and over. The previous system set a minimum benefit duration of 6 months plus an additional month for each 12 months of employment with a record of wages earned. The 1989 law also increased generosity of the unemployment social subsidy. Unemployed people with four or more dependents could claim a benefit equal to 100 percent of the minimum wage. Previously, only people with 6 or more dependents could claim this level of benefits. Benefit duration of the social subsidy was also extended and harmonized with the general unemployment subsidy.

6

Jackman, Layard, and Pissarides (1989) and Layard, Nickell, and Jackman (1991) use a similar method for capturing structural change in their estimates of the UV relationship in the United Kingdom.

7

There are two possible explanations for this result.

8

OECD data shows that Portugal has the sixth highest benefit replacement ratio, (defined as benefits before tax as a percent of previous earnings before tax).

Portugal: Selected Issues and Statistical Appendix
Author: International Monetary Fund