Age- and gender-specific unemployment in Scandinavian countries: an analysis based on Okun's law.

Hutengs, Oliver ; Stadtmann, Georg

INTRODUCTION

Youth unemployment in European Union (EU) member states has been on the rise since the beginning of the world financial crisis. This negative development was further accelerated by the crises in Europe. Southern European countries have experienced record high levels of youth unemployment (Hutengs and Stadtmann, 2013). Similarly, youth unemployment problems have been identified in Central and Eastern European economies, such as Poland or Hungary (Hutengs and Stadtmann, 2014). Scandinavian countries have strong economic links to other European economies; therefore, the recent low-growth period in Europe also affected the Scandinavian economies as well as their labor markets.

This paper investigates Scandinavian countries and their male and female unemployment rates. An EU15 aggregate of countries is used for comparison. (1) Okun's (1962) law is used to estimate age cohort and gender-specific Okun coefficients to provide evidence on the business-cycle dependence of young people across Scandinavian countries. We show that young people are predominantly and significantly more exposed to business-cycle fluctuations than are older ones. This especially holds true for the male population.

The remainder of the paper is structured as follows: The next section briefly reviews the literature. The section after that provides a general description of the dataset and the descriptive statistics regarding youth and total unemployment rates. We also highlight differences in male and female unemployment. The penultimate section discusses the regression approach and results. The final section concludes.

LITERATURE REVIEW

In general, Okun's law describes the negative relation between the real GDP growth rate and a change in the unemployment rate. Positive economic growth will, ceteris paribus, lead to a decrease in the unemployment rate. Research on Okun's law is extensive, with many studies confirming the general negative relationship between economic growth and change in the unemployment rate (see, for example, Knoester, 1986; Freeman, 2001; Sogner and Stiassny, 2002; Cevik et al., 2013). The focus of many Okun-related studies lies with the Okun coefficient, that is, the parameter showing the strength of the relation between GDP growth and the change in the unemployment rate. Many studies concentrate on comparisons of the Okun coefficient across countries.

Hutengs and Stadtmann (2013, 2014) have applied Okun's law to different age cohorts for Eurozone and Eastern European countries, respectively, thus focusing on the effect of business-cycle movements on the unemployment rate of different subgroups on the labor market. These studies provide evidence that Okun's law is also valid for young people and not only for the total population. Results of these studies indicate that the absolute value of the Okun coefficient decreases with growing age, that is, GDP growth has, ceteris paribus, a larger impact on the unemployment rates of younger cohorts. In fact, the impact on the youngest cohort (15-24 years old) is much stronger than for subsequent cohorts.

DATASET AND DESCRIPTIVE STATISTICS

The data used for the estimates of Okun's law consist of real annual GDP growth rates and the unemployment rate for various age cohorts as provided by the Organisation for Economic Co-operation and Development (OECD, 2013). (2) The data start in 1984 for all countries except Iceland (1992) and end in 2011. (3)

The trends in youth unemployment, that is, the people between the age of 15 and 24 years, provide some insights into the differences and similarities among Scandinavian economies. Figure 1 shows that all Scandinavian countries, with the exception of Norway, show rising unemployment rates in recent years. This can be attributed to the ongoing weak GDP growth following the financial crisis. Nonetheless, data for most countries also indicate the possibility of decreasing rates as of late. There is no such indication for the Danish economy. Norway's development does not resemble the performance of the other countries. In fact, its unemployment rate has been relatively flat in recent years. The reason probably lies in its role as a relatively large oil and natural gas producer, thus providing its economy with additional stimulus over the past several decades, when oil prices increased.

When looking at developments over the last decade, Scandinavian countries can be classified by their unemployment trend. Norway, Denmark and Iceland show a relatively flat development since the start of the century and an increase in recent years, with the exception of Norway. Sweden shows a general upward trend that even strengthened during the recent crisis. In contrast, Finland, the only Scandinavian eurozone member, shows a downward trend that was only slightly interrupted by the recent crisis. Some countries, like Sweden, Iceland and Denmark, exceeded their previously highest levels of youth unemployment. Rates in Iceland rose toward 20% and in the case of Sweden surpassed this level.

Table 1 compares the differences between youth unemployment and total unemployment across countries and over time. We would expected youth unemployment rates to be higher than the total unemployment rates because young people are endowed with fewer skills and are less experienced than their older peers, which makes their employment less likely (OECD, 2010], Furthermore, labor market institutions can directly influence the level and difference between the two rates. The European Economic Advisory Group (2013, p. 86) argues, for example, that a minimum wage adversely affects young people by artificially lowering pay differences between younger and older workers, thus directly decreasing young people's employment chances during a recession. Nonetheless, youth unemployment rates are not independent of the total unemployment rate. Both usually move into the same direction (Brenke, 2012).

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

Table 1 reveals that Scandinavian youth and total unemployment is, on average, below the respective levels of the EU15 countries. Only Finland comes close to the European levels, all other countries show much lower rates. Nonetheless, this does not necessarily lead to a better youth (15-24 years) to adult (25-64 years) unemployment ratio compared with the EU15 average. Thus, young people in some Scandinavian countries may be even worse off than their peers in the EU15 when comparing their average unemployment rates with those of adults.

Figure 2 shows the development of the youth (15-24 years) to adult (25-64 years) unemployment ratio among Scandinavian countries. The ratio is not constant over time in any country. Despite some large variations, the data show an upward trend of the ratio in Denmark, Finland and Sweden since the mid-1990s. This change is strongest in Sweden where the ratio jumped from 2 to over 4, indicating that the youth unemployment problem has become more severe in Sweden than in other Scandinavian countries. Interestingly, the increasing ratio in Sweden cannot be solely contributed to the recent financial and economic crisis as it started long before its emergence. Moreover, there is no sign of a trend reversal yet.

[FIGURE 3 OMITTED]

Figure 3 shows a comparison between the respective male and female unemployment rate. During the observation period there is no homogeneous development of both rates among Scandinavian countries. Especially the Danish and Finnish economies showed higher female rates over a long time whereas Sweden and Iceland showed a reverse pattern. More interestingly, the recent economic crisis led to a strong difference between male and female youth unemployment, caused by a much sharper increase in the male unemployment rate. This effect might be because of the fact that the crisis had a strong negative impact on the construction and manufacturing sectors of the economies. These sectors predominantly employ males, thus strongly increasing their unemployment rates in a downturn (see European Commission, 2013).

REGRESSION ANALYSIS

We use Okun's (1962) law in order to focus on the effect of business-cycle fluctuations on the change in the unemployment rate. Following Knotek (2007), the differenced version of Okun's law can be written as a linear regression model by

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

with [DELTA][u.sub.t] as change in unemployment rate, [??] as discrete real GDP growth rate and [[epsilon].sub.t] an assumed white noise error term. The parameter [beta], the Okun coefficient, is expected to be negative, that is, an increase in real GDP growth rate will lower the unemployment rate.

The estimation of [beta] coefficients for each age cohort and each country independently from another as suggested by equation 1 precludes the possibility of efficiently testing for equality of coefficients with respect to a country's different age cohorts. Thus, we make use of a balanced panel for each country, following the same approach as Hutengs and Stadtmann (2014). Each country panel consists of the yearly changes in the unemployment rate for five different age cohorts and the real GDP growth rate. Thus, we assume there exists unobserved heterogeneity between the age cohorts in each country, which we now control for. In detail, we estimate a fixed-effects model for each country through the following panel least squares dummy variable model:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)

where [DELTA][u.sub.i,t] represents the change in unemployment rate for cohort i at time t, D, is a dummy variable accounting for the different age cohorts and [[epsilon].sub.i, t] is an assumed white noise error term. Thus, [[beta].sub.i] captures the different cohort-specific Okun coefficients.

Common to panel analysis are the presence of heteroscedasticity, serial correlation and cross-sectional dependencies. These effects may lead to inefficient estimates with biased standard errors and thus misleading results. We test all country panels for these properties. Thus, all our reported results show panel-corrected standard errors allowing inference on statistical significance. (4)

Table 2 summarizes our regression results of Equation 2. Similar to other cross-country studies (see, for example, Hutengs and Stadtmann, 2013; Sogner and Stiassny, 2002; Lee, 2000; Moosa, 1997), the Okun coefficients are negative across all countries and age cohorts, confirming Okun's law. Norway stands out again by having significantly lower absolute Okun coefficients than all other countries. This suggests that Norwegians are less affected by GDP changes than are their Scandinavian peers.

Norwegian unemployment is less sensitive to cyclical changes compared with the other Nordic countries for several reasons. First, according to Table 3, the Norwegian economy is characterized by the lowest industrial share of all the Nordic economies. The industrial sector with its manufacturing and construction branches has higher fluctuations of employment over the business-cycle than does the service sector. Second, the Norwegian mining sector, including its crude oil and natural gas production, has shown increases in employment during the crisis (Eurostat, 2014), helping to offset negative effects from employment reductions in industry. Third, the Norwegian employment in the health sector, which roughly counts for 20% of overall employment, increased by over 14% during the crisis, helping to offset negative developments in other branches (see Table 4).

In addition, the results provide more important insights:

Figure 4 shows one common pattern where all Scandinavian countries have their largest absolute Okun coefficient in their youngest cohort (15-24 years). These people are most affected by business-cycle fluctuations. Furthermore, the Okun coefficient becomes smaller in absolute terms with increasing age. The decrease in the coefficient is most distinct from the 15-24 years to the 25-34 years cohort. Afterwards it levels off rather rapidly. In fact, significance tests show that the high absolute coefficients of the youngest cohort are significantly different from the lower coefficients of all subsequent cohorts. (5) This is true for all countries with the exception of Denmark and Norway, where the differences are not statistically significant even between the first two cohorts. Results further show that the three oldest cohorts tend to have coefficients that are close to zero. Furthermore, the coefficients are not significantly different from each other as well. Thus, in Scandinavian economies, people tend to be less exposed to business-cycle fluctuations after passing the age of 35.

The first two insights confirm results of two other studies. Hutengs and Stadtmann (2013, 2014) studied the size of the Okun coefficients for various age cohorts and show that the observed pattern also holds with respect to Eurozone and Eastern European countries, respectively. Thus, the results for the Scandinavian countries are in line with previous literature. Furthermore, the absolute values of the Okun coefficients decline with age as the unemployment rate is much more volatile for younger age cohorts, that is, the standard deviation of the change in the unemployment rate also decreases with progressing age. The data do not take into account different kinds of unemployment, that is, long-term versus short-term unemployment. Nonetheless, Hutengs and Stadtmann (2013) show for a group of Eastern European countries that young people tend to be less affected by long-term unemployment than older cohorts. Furthermore, O'Higgins (2003) and Ryan (2001) provide a discussion of the youth unemployment rate, finding that it is more strongly influenced by a search for better jobs and a generally higher fluctuation of the younger cohorts into and out of employment. These facts support the obtained results showing that a higher share of long-term unemployed, as in the older cohorts, should dampen volatility with respect to changes in the unemployment rate, whereas a higher fluctuation regarding the employment status, as in the youth cohort, should increase volatility. Thus, young people are more strongly affected by business-cycle changes than are older ones. Moreover, the reported unemployment rates do not distinguish between people seeking part-time and full-time work because they are compiled according to ILO standards, so people are identified as unemployed when they are not currently employed, actively looking for work and are able to start working within 2 weeks.

[FIGURE 4 OMITTED]

Splitting the overall analysis between men and women, Tables 5 and 6 provide a similar conclusion; all coefficients are negative and decrease in absolute terms over time, independent of gender. Two big differences exist. First, the model fit is much better for the male regression than for the female regression as shown by the higher [R.sup.2] suggesting that unemployment changes of women are not as well explained by GDP growth. Second, as Figure 5 shows, the male coefficients are significantly higher than their female counterparts. This holds for all countries except Sweden and the EU15 aggregate. This result is in line with the already presented argument that males are predominantly employed by more cyclical sectors than are females. Thus, their unemployment rate reacts more strongly to any change in GDP.

[FIGURE 5 OMITTED]

Generally, employment is concentrated in five sectors, with manufacturing, construction, wholesale/retail, education and health together accounting for 50% -60% of overall employment in the sample countries (Eurostat, 2014). Table 4 shows that changes in employment between 2008 and 2010 are strongest in the industrial sectors. Moreover, the employment share of men is considerably higher in those industrial sectors, whereas female employment is concentrated in the service sector of the economy. In fact, it is the manufacturing and construction sector that shows an all-out decrease in employment numbers, while some service sectors even report an increase in employment during the crisis, thus mainly benefiting women. This sectoral distribution of employment, that is, 23%-37% of all men but only 7%-12% of women work in the industrial sector of each country, in combination with the strong fluctuation in the industrial sector, make male employees more vulnerable to business-cycle changes. (6)

CONCLUSION

This paper has analyzed unemployment in Scandinavian economies by evaluating recent unemployment trends within age cohorts and the differences in male- and female-specific Okun coefficients. The main results can be summarized as follows:

Youth unemployment became more severe in recent years compared with adult unemployment as shown by a rising ratio of young to total unemployment in Scandinavian countries. Sweden exhibits the worst development in this regard. Young people (age 15-24) are most exposed to business-cycle fluctuations showing the highest absolute Okun coefficients among all cohorts. Men show significantly higher Okun coefficients in absolute terms than do their female counterparts because of a higher representation in cyclical occupations.

Scandinavian youth unemployment has not reached Southern European levels of 50% (Hutengs and Stadtmann, 2013) nor is it conceivable that this will happen. Nevertheless, lower absolute youth unemployment figures compared with the south should not hide the fact that young people in Scandinavia are disproportionately worse off than their older peers and require additional attention. The rising unemployment ratio in countries like Sweden and Finland indicate a problem for policymakers. Unemployment at an early stage of the career has negative effects on lifetime income and career possibilities (Olofsson and Wadensjo, 2012; Bell and Blanchflower, 2011). Moreover, high youth unemployment threatens the social cohesion of the societies itself as the riots by young people in Sweden have demonstrated. In addition, a rising unemployment ratio might also indicate that the youth unemployment development is not only business-cycle driven but rather a sign of severe structural problems in the respective labor markets. The Swedish training system, for example, provides little working experience for its apprentices (Korpi and Mertens, 2003). In contrast, countries such as Germany and Austria, with a focus on employer provided apprenticeships and thus more direct working experience, exhibit lower youth unemployment rates (Hutengs and Stadtmann, 2013).

REFERENCES

Bell, DNF and Blanchflower, DG. 2011: Youth unemployment in Europe and the United States. Nordic Economic Policy Review 1:11-37.

Brenke, K. 2012: Arbeitslosigkeit in Europa: Jugendliche sind viel starker betroffen als Erwachsene. Wochenbericht des DIW Berlin 30: 3-12.

Cevik, El, Dibooglu, S and Barisik, S. 2013: Asymmetry in the unemployment-output relationship over the business cycle: Evidence from transition economies. Comparative Economic Studies 55(4): 557-581.

European Commission. 2013: The impact of the economic crisis on the situation of women and men and on gender equality policies. Publications Office of the European Union: Luxembourg.

European Economic Advisory Group. 2013: The EEAG report on the European economy. CESifo: Munich.

Eurostat. 2014: The Labour Force Survey (LFS)--Detailed Annual Survey Results. Eurostat Employment and Unemployment (LFS) Statistical Database, accessed 12 May 2014.

Freeman, DG. 2001: Panel tests of Okun's law for ten industrial countries. Economic Inquiry 39(4): 511-523.

Hutengs, O and Stadtmann, G. 2013: Age effects in Okun's law within the Eurozone. Applied Economics Letters 20(9): 821-825.

Hutengs, O and Stadtmann, G. 2014: Don't trust anybody over 30: Youth unemployment and Okun's law in CEE countries. Bank i Kredyt 45(1): 1-16.

Knoester, A. 1986: Okun's law revisited. Weltwirtschaftliches Archiv 122(4): 657-666.

Knotek, ES. 2007: How useful is Okun's law? Economic Review, Federal Reserve Bank of Kansas City Q IV: 73-103.

Korpi, T and Mertens, A. 2003: Training systems and labor mobility: A comparison between Germany and Sweden. The Scandinavian Journal of Economics 105(4): 597-617.

Lee, J. 2000: The robustness of Okun's law: Evidence from OECD countries. Journal of Macroeconomics 22(2): 331-356.

Moosa, IA. 1997: A cross-country comparison of Okun's coefficient. Journal of Comparative Economics 24(3): 335-356.

OECD. 2010: Off toagood start? Jobs for youth. OECD Publishing: Paris.

OECD. 2013: Labour Force Statistics: Population and labour force. OECD Employment and Labour Market Statistics (database), accessed 14 January 2013, doi: 10.1787/lfs-lfs-data-en.

O'Higgins, N. 2003: Trends in the youth labour market in developing and transition countries. World Bank Social Protection Discussion Paper Series 0321, http://go.worldbank.org/XTNM664DJ2.

Okun, AM. 1962: Potential GNP: Its measurement and significance. In: Baily, MN and Okun, AM (eds). The Battle Against Unemployment and Inflation (1982). Norton: New York, 132-145.

Olofsson, J and Wadensjo, E. 2012: Youth, education and labour market in the Nordic countries. Friedrich-Ebert-Stiftung, International Dialogue, http://library.fes.de/pdf-files/id/09468.pdf.

Ryan, P. 2001: The school-to-work transition: A cross-national perspective. Journal of Economic Literature 39(1): 34-92.

Signorelli, M, Choudhry, M and Marelli, E. 2012: The impact of financial crises on female labour. European Journal of Development Research 24(3): 413-433.

Sogner, L and Stiassny, A. 2002: An analysis on the structural stability of Okun's law-A cross-country study. Applied Economics 34(14): 1775-1787.

(1) EU15 includes all countries that were members of the EU before the eastern enlargement in May 2004. A more detailed and country-specific analysis of major EU15 countries is found in Hutengs and Stadtmann (2013).

(2) Unemployment rates are based on International Labour Organisation (ILO) standards, ensuring comparability among different countries.

(3) The end date (2011) is determined by data availability at the time of the study. Countries included in the study with number of observations per cohort: Denmark, Finland, Norway, Sweden, EU15 (28 years) and Iceland (22 years).

(4) Estimation results are obtained through linear regression with Prais-Winsten standard errors.

(5) Significance test results are available from the authors on request.

(6) Own calculations based on Eurostat data: Employment by sex, age and economic activity (from 2008 onwards, NACE Rev. 2)-1000 [lfsa_egan2]. See Signorelli et al. (2012) for a discussion about the impact of crises on females.

OLIVER HUTENGS [1] & GEORG STADTMANN [1, 2]

[1] European University Viadrina Grosse Scharrnstrasse 59, Frankfurt (Oder), 15320, Germany.

Email: hutengs@europa-uni.de

[2] Department of Business and Economics, University of Southern Denmark, Campusvej 55, Odense M, 5230, Denmark.

E-mail: stadtmann@europa-uni.de Table 1: Summary statistics (1) Average youth (2) Average total (3) [DELTA] columns Country unemployment unemployment (1) and (2) Denmark 10.00 6.40 3.60 Finland 18.26 8.75 9.51 Iceland 8.79 3.89 4.90 Norway 10.02 3.87 6.15 Sweden 14.99 6.02 8.97 EU15 18.31 9.12 9.12 (4) Maximum (5) Minimum Country [DELTA] youth and total [DELTA] youth and total Denmark 6.66 2.05 Finland 14.74 3.86 Iceland 8.71 0.00 Norway 7.98 3.09 Sweden 16.84 2.30 EU15 12.30 6.70 Notes: (1) Average unemployment rate of the youngest age cohort in percentage; (2) Average unemployment rate of the total population in percentage; (3) Difference between Column (1) and (2) in percentage points; (4) Highest value of the difference between the young and the total unemployment rate within one country over time in percentage points; (5) Lowest value of the difference between the young and the total unemployment rate within one country over time in percentage points. Table 2: Regression results: Okun coefficients and standard errors Country 15-24 25-34 35-44 45-54 Denmark -0.755 *** -0.543 *** -0.297 *** -0.312 *** (0.128) (0.093) (0.067) (0.066) Finland -0.667 *** -0.357 *** -0.249 *** -0.259 *** (0.096) (0.064) (0.060) (0.050) Iceland -0.465 *** -0.349 *** -0.155 *** -0.151 *** (0.094) (0.079) (0.046) (0.041) Norway -0.382 ** -0.258 ** -0.121 * -0.119 ** (0.138) (0.085) (0.061) (0.039) Sweden -0.795 *** -0.390 *** -0.212 *** -0.193 *** (0.160) (0.074) (0.047) (0.036) EU15 -0.708 *** -0.392 *** -0.244 *** -0.207 *** (0.077) (0.056) (0.047) (0.039) Country 55-64 [R.sup.2] N Denmark -0.256 ** 0.49 140 (0.085) Finland -0.261 * 0.47 140 (0.110) Iceland -0.151 *** 0.50 100 (0.040) Norway -0.081 * 0.21 140 (0.033) Sweden -0.128 * 0.46 140 (0.053) EU15 -0.185 ** 0.63 140 (0.060) Note: (N) number of observations, standard errors in parentheses, significance at *** 1% level, ** 5% level, * 10% level. Table 3: Sectoral composition of employment as % of total employment Industry (a) Country With mining (%) Without mining (%) Services (b) Denmark 23.2 23.0 74.5 Finland 25.2 25.0 70.6 Iceland 21.9 21.9 73.8 Norway 20.7 19.0 76.9 Sweden 21.4 21.2 76.7 EU15 25.9 25.6 71.3 (a) industry includes NACE Rev. 2 sectors B-F. (b) Services includes all sectors except industry and agriculture. Source: Own computations based on data from Eurostat, Employment LFS series Table 4: Employment changes and men's share in selected sectors between 2008 and 2010 Manufacturing (a) Construction (b) Wholesale (c) 2008- share 2008- share 2008- share Country 2010 (%) (%) 2010 (%) (%) 2010 (%) (%) Denmark -19.2 67.0 -19.5 91.1 -9.4 55.7 Finland -13.4 74.4 -7.8 92.9 -4.0 51.6 Iceland -4.5 71.9 -42.1 93.6 -13.3 56.9 Norway -9.6 76.3 -3.0 93.4 -3.2 52.4 Sweden -12.7 76.6 -1.2 92.2 -0.9 56.1 EU15 -9.8 72.8 -11.5 90.6 -3.7 51.8 Education (d) Human health (e) 2008- share 2008- share Country 2010 (%) (%) 2010 (%) (%) Denmark 11.3 42.4 3.0 17.6 Finland 5.6 32.8 -0.8 10.5 Iceland -2.5 23.0 7.0 17.8 Norway -4.6 34.4 14.6 17.3 Sweden 0.1 24.9 -2.4 16.6 EU15 3.6 30.3 4.9 21.8 (a) C Manufacturing. (b) F Construction. (c) G Wholesale and retail trade; repair of motor vehicles and motorcycles. (d) P Education. (e) Q Human health and social work activities. Source: Own computations based on data from Eurostat, Employment LFS series. Selected sectors official NACE Rev. 2 codes Table 5: Regression results: Okun coefficients and standard errors--men only Country 15-24 25-34 35-44 45-54 Denmark -0.876 *** -0.639 *** -0.420 *** -0.423 **** (0.132) (0.109) (0.077) (0.074) Finland -0.900 *** -0.469 *** -0.364 *** -0.353 *** (0.094) (0.074) (0.061) (0.059) Iceland -0.526 *** -0.381 *** -0.177 *** -0.171 ** (0.138) (0.106) (0.051) (0.064) Norway -0.462 ** -0.343 *** -0.130 -0.153 ** (0.164) (0.104) (0.071) (0.056) Sweden -0.964 *** -0.504 *** -0.282 *** -0.238 *** (0.178) (0.074) (0.055) (0.040) EU15 -0.866 *** -0.503 *** -0.314 *** -0.258 *** (0.087) (0.057) (0.043) (0.038) Country 55-64 [R.sup.2] N Denmark -0.309 ** 0.53 140 (0.099) Finland -0.292 ** 0.61 140 (0.106) Iceland -0.197 *** 0.40 100 (0.056) Norway -0.083 * 0.22 140 (0.041) Sweden -0.168 ** 0.52 140 (0.059) EU15 -0.232 *** 0.71 140 (0.064) Note: (N) number of observations, standard errors in parentheses, significance at *** 1% level, ** 5% level, * 10% level. Table 6: Regression results: Okun coefficients and standard errors--women only Country 15-24 25-34 35-44 45-54 Denmark -0.622 *** -0.438 *** -0.162 -0.194 * (0.158) (0.118) (0.096) (0.091) Finland -0.554 *** -0.311 *** -0.198 ** -0.236 *** (0.116) (0.072) (0.066) (0.055) Iceland -0.403 *** -0.312 * -0.131 * -0.129 *** (0.114) (0.122) (0.065) (0.029) Norway -0.281 * -0.155 * -0.099 -0.080 * (0.112) (0.068) (0.056) (0.038) Sweden -0.629 *** -0.267 ** -0.143 ** -0.148 *** (0.148) (0.085) (0.045) (0.037) EU15 -0.512 *** -0.259 *** -0.162 ** -0.146 ** (0.082) (0.059) (0.057) (0.048) Country 55-64 [R.sup.2] N Denmark -0.189 0.26 140 (0.116) Finland -0.293 * 0.34 140 (0.120) Iceland -0.100 0.38 100 (0.060) Norway -0.075 0.16 140 (0.047) Sweden -0.097 0.34 140 (0.063) EU15 -0.118 0.42 140 (0.063) Note: (N) number of observations, standard errors in parentheses, significance at *** 1% level, ** 5% level, * 10% level.