- Email: [email protected]
Wage mobility: do institutions make a difference?
Labour Economics 13 (2006) 387 – 404 www.elsevier.com/locate/econbase
Wage mobility: do institutions make a difference? Ana Rute Cardoso* IZA Bonn, Schaumburg-Lippe-Str 7/9, 53113 Bonn, Germany University of Minho and CEPR Received 1 November 2003; received in revised form 1 May 2004; accepted 1 October 2004 Available online 27 April 2005
Abstract This paper checks empirically the assertion that strict labour market regulation leads to low wage mobility, addressing an apparent puzzle. Indeed, most economic reasoning links a combination of regulations in the labour market to low wage flexibility and mobility, but the scarce empirical evidence available challenges that view. I focus on Portugal, one of the most regulated labour markets in Europe. The evidence gathered indicates that an aggregate view—of minimum wage enforcement, unionisation rates and extension of collective bargaining contracts—provides a misleading idea of the actual constraints imposed by the institutional framework on wage setting. Instead, micro conditions at the firm level play a major role shaping wage mobility in Portugal. Some comparisons with the UK, traditionally pointed out as a flexible labour market, are provided. Remarkable similarities in mobility level and trend are detected between the two countries, further suggesting that a regulated institutional framework does not necessarily reduce individual mobility in the wage distribution. D 2004 Elsevier B.V. All rights reserved. JEL classification: J31; J60 Keywords: Wage dispersion; Collective bargaining
1. Introduction Most economic reasoning links a combination of regulations in the labour market— strict employment protection legislation, national minimum wages, and widespread * Tel.: +49 2283894508; fax: +49 2283894510. E-mail address: [email protected]. 0927-5371/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.labeco.2004.10.004
388
A.R. Cardoso / Labour Economics 13 (2006) 387–404
influence of collective bargaining on wage setting—to low wage flexibility and mobility. However, empirical evidence, though scarce, challenges that view. Burkhauser et al (1997) point to the remarkable similarity between the USA and Germany in terms of mobility levels and trend, discarding the hypothesis that countries with higher Government intervention in the labour market and higher union influence will have lower wage mobility. Similarly, Aaberge et al. (2002) detect lower levels of earnings mobility in the USA than in the Scandinavian countries. Also, the OECD (1997) found that less regulation in labour and product markets, in countries such as the UK and the USA, does not result in higher levels of wage mobility. This paper addresses this apparent puzzle, by concentrating on wage mobility in Portugal. It checks whether empirical support is found for the assertion that strict labour market regulations lead to low wage mobility. In Portugal collective bargaining is extensively applied, both to unionised and non-unionised workers, minimum wages have long been enforced, and employment protection legislation is among the most strict in the OECD. A comparison with the UK is provided, relying on the work by Dickens (2000). As opposed to Portugal, the British labour market is recognized as one of the most flexible in Europe. Specifically, the following questions are addressed: What is the extent of wage mobility in Portugal? Is a strongly regulated labour market likely to protect low-wage workers, sustaining their wages, while nevertheless reducing their chances of upward mobility? Which forces drive wage changes in Portugal: mainly collective bargaining, or firmspecific conditions? How does mobility in Portugal compare with the UK, two countries with very different institutional frameworks? A remarkable longitudinal matched employer-employee data set, covering all the wageearners in manufacturing and services private sector in Portugal, is used. Problems commonly faced by panel data sets, such as under-sampling or over-sampling of certain groups and panel attrition, are thus reduced. Also, the legal requirement for the data to be posted in the company contributes to its reliability, reducing measurement errors. Section 2 outlines the major contrasts between the institutional framework in the British and Portuguese labour markets. Section 3 discusses theoretical and empirical evidence on the link between labour market regulations and wage mobility. Section 4 describes the data set used. Section 5 overviews trends in wage dispersion, while Sections 6 and 7 concentrate on wage mobility. Section 8 discusses the results, identifying the role of institutional forces versus firm level conditions shaping wage variability across workers. Concluding comments are presented in Section 9.
2. Labour market institutions in Portugal and the UK The UK and Portugal shared, like most industrialized economies, a remarkable drop in unionisation rates over recent decades. In Portugal the unionisation rate reached 61 percent in 1970 and 1980, but by 1990 it had dropped to 32 percent; a similar trend was followed in the UK, where it dropped from 50 percent in 1980 to 39 percent in 1990 (OECD, 1994: 184). Nevertheless, collective bargaining has a sharply different role in these two economies. While it covers virtually all the workforce in Portugal, just approximately one fourth of the
A.R. Cardoso / Labour Economics 13 (2006) 387–404
389
workers in the UK are covered (European Commission, 2002: 47). Such contrast results from different mechanisms of extension of contracts. Voluntary extensions are common in Portugal, when a workers’ or employers’ representative subscribes to an agreement it had initially not signed. Also, agreements subscribed by an employer most often apply to all of the company’s workforce, irrespective of the worker’s actual union membership status. This is the default mechanism in Portugal that results from a legal provision, whereas in the UK it applies more rarely and on a voluntary basis (European Commission, 2002: 39– 47). Moreover, in Portugal a mandatory bargaining regime can be applied, when workers are not covered by unions, when one of the parties involved refuses to negotiate or bargaining is obstructed in any other way, in which case a contract is imposed by the Government (Cardoso and Portugal, 2004). The degree of centralization of collective bargaining is also different in these two countries. Agreements covering a whole industry predominate among the Portuguese workforce. Centralized negotiations take place yearly at the national level between trade union confederations, employersT federations and the Government, setting indicative guidelines for wage increase. It should also be noted that bargaining is highly synchronized in the Portuguese economy, with wage clauses in collective contracts updated yearly, most often to take effect January each year. Job protection is another aspect of the institutional framework clearly setting the British and the Portuguese systems apart. Portugal is invariably pointed out as one of the economies in the OECD with highest employment protection, whereas the UK shows up at the other extreme, with one of the lowest levels of employment protection (see for example OECD, 1999). During the period under analysis, the minimum wage legislation was also different across the two countries. A national minimum wage was not enforced in the UK, while in Portugal it was quite binding, reaching in 1999 44 percent of the mean wage in the economy.
3. Why would different labour market institutions lead to different degrees of wage mobility? There is theoretical and empirical support to the idea that employment protection combined with other regulations in the labour market, such as strong union influence and minimum wages, may lead to lower wage mobility. Employment protection tends to reduce worker turnover between firms. In the model by Lindbeck and Snower (1989), higher firing costs raise the power of incumbents, increasing their wages and reducing layoffs. Lazear (1990) argues that more restrictive employment protection legislation leads to less dismissals and less hirings. Relying on macro crosscountry data, he finds that countries with high firing costs have lower employment levels and higher unemployment rates. The theoretical and empirical analyses by Bertola (1990) also find that employment protection lowers flows in the labour market, but highlight that the average employment in the economy is not necessarily lower—it is just more stable, as firms opt for a less fluctuating workforce. Abraham and Houseman (1994) compared Germany, Belgium, France and the USA, having found that the adjustment of employment
390
A.R. Cardoso / Labour Economics 13 (2006) 387–404
to changes in output is faster in countries with less employment protection.1 Bertola and Rogerson (1997), on the other hand, claim that yearly job creation and destruction rates are similar across countries with very different degrees of employment protection. Blanchard and Portugal (2001) reconcile this conflicting evidence by analysing quarterly data on the USA and Portugal, which enabled the identification of patterns that are not apparent from yearly data. They found that behind similar unemployment rates hide two labour markets operating in very different ways: high employment protection leads in Portugal to much lower flows of workers into unemployment—but a much higher duration of unemployment—, which results from lower job creation and destruction rates, together with lower flows of workers given the rate of job creation and destruction. There thus seems to be support to the assertion that employment protection reduces in Portugal worker turnover among firms. One major source of wage mobility—the transition between firms—is thus reduced.2 Tougher job protection legislation could therefore reduce the scope for wage mobility. Different authors who have addressed explicitly the link between job protection and wage flexibility have nevertheless stressed that job protection by itself does not reduce wage flexibility. In particular, Bertola (1990) found that, in countries with high employment protection, wages react more to the unemployment rate. He therefore concludes that the impact of employment protection reducing wage flexibility may be undone by other labour market institutions, and that it would take a combination of institutions—employment protection legislation, widespread trade union influence and national minimum wages—to reduce wage flexibility. Such combination of institutions is certainly present in the Portuguese labour market, as described in the previous section. That would reinforce the expectation of low wage mobility. Widespread collective bargaining could lead to lower wage mobility, since it defines a strict wage hierarchy, based on the worker category, and clear career progression rules, which in Portugal apply to both unionized and non-unionized workers. Firms do have discretionary power over promotions, but because careers and associated wages are strongly regulated, possibilities of relative wage mobility in the economy would be weakened. Also, high union influence in the economy could give insiders power to protect wages from market forces.
4. Data source and concepts used A large longitudinal data set matching firms and workers in the Portuguese economy is used. The data are gathered annually by the Ministry of Employment, based on an inquiry that every establishment with wage-earners is legally obliged to fill in. Reported data cover all the personnel working for the establishment in a reference week (in March until 1993, and in October from 1994 onwards). Public administration and domestic service are not covered, and the coverage of agriculture is low, due to the low share of wage earners. For the remaining sectors, the mandatory nature of the survey leads to an extremely high 1 2
Whereas the adjustment of hours worked is similar. On the impact of job changes on wage changes, see for example Gottschalk and Moffitt (1994).
A.R. Cardoso / Labour Economics 13 (2006) 387–404
391
response rate and in fact the population of firms with wage-earners in the manufacturing and services private sectors is covered. Reported data on the worker include gender, age, schooling, monthly earnings (split into several components), duration of work, the collective bargaining contract and the detailed worker category within the contract. A worker identification code, based on a transformation of the social security number, enables tracking him/her over time. The panel data set covers the period 1986 to 19993. The final panel includes over 22 million worker-year observations and almost 5 million workers. Extensive checks have been performed to guarantee the accuracy of the data, using gender, date of birth, and highest schooling level achieved (for details on the procedure followed, see Cardoso, 2004). The analysis is carried out separately for males and females and it focuses on workers aged 22 to 59. Workers with missing wage (because they are not wage-earners, but instead owners of the company, unpaid family members or members of cooperatives, according to the Portuguese classification) were kept in the analysis of transition matrices. Once a worker drops out of the database, one of several situations may have occurred: (s)he may have become unemployed, inactive, or an independent worker, may have joined an uncovered sector (Public Administration) or the Army. Even though unemployment is the most likely situation, when analyzing transition matrices, the cell out of the data base should be interpreted with caution. When choosing the time spans under analysis, two constraints should be kept in mind: in 1994, the reference week for the inquiry changed from March to October, and therefore 1993–94 captures changes over a longer period (19 months), rendering it infeasible for comparison with the remaining periods; in 1990, no worker data were reported. Gross hourly earnings were computed as hw ¼
bw þ sen þ reg þ overtw ; normh þ overth
ð1Þ
where bw stands for base-wage, sen are seniority-indexed components of pay, reg are other regularly paid components, and overtw is payment for overtime; normh refers to the normal duration of work and overth are overtime hours. Wages were deflated using the Consumer Price Index.
5. Wage dispersion: trends and patterns Wage dispersion increased at a fast pace in Portugal until mid-90s (Fig. 1), slowly declining afterwards. Fig. 2 considers separately dispersion at the bottom and at the top halves of the distribution, reporting the ratios between the percentiles 50/10 and 90/50. It shows that the reported overall trend was shared by the top and the bottom of the distribution, since in either case dispersion increased until mid-90s and declined afterwards.
3
Note, however, that no worker data were reported for the year 1990.
392 A.R. Cardoso / Labour Economics 13 (2006) 387–404
Fig. 1. Trends in wage dispersion (Gini index), Portugal and the UK. Source: For Portugal- Computations based on Portugal, MTSS (1986–1999); for the UK- Machin (1998). Note: The values for Portugal in 1990 result from linear extrapolation of the trend, since no worker data are available.
A.R. Cardoso / Labour Economics 13 (2006) 387–404 Fig. 2. Trends in wage dispersion at the bottom and top of the distributions, Portugal and the UK. Source: for Portugal- computations based on Portugal, MTSS (1986– 1999); for the UK- Machin (1998). Note: The values for Portugal in 1990 result from linear extrapolation of the trend, since no worker data are available. 393
394
A.R. Cardoso / Labour Economics 13 (2006) 387–404
It is worth highlighting the comparison with the UK. The similarity between the level of wage dispersion in both countries in 1986, the first year with overlapping data, is outstanding.4 Moreover, in both countries inequality increased very sharply until mid-90s. In Portugal it increased at a faster pace up to 1994 but declined afterwards, whereas in the UK it kept increasing. In synthesis, both countries became more unequal labour markets, having departed from similar levels of wage inequality in mid-80s. However, the pattern of wage dispersion contrasts sharply. The bottom part of the distribution is much more compressed in Portugal and, conversely, the top is much more stretched. Similar levels of overall wage dispersion therefore result from a very different pattern. The compression at the bottom in Portugal reflects the impact of the national minimum wage, which pulls low wages closer to the middle of the distribution.5
6. Wage mobility: contrasts and similarities The mobility matrix in Table 4 in Appendix A reports for males a high degree of earnings persistence, with the proportion of workers who do not change decile ranging from 31 percent to 66 percent, depending on the decile of origin. Most of those who travel in the wage distribution, do so for a short distance. These patterns hold for females as well. To check changes in wage mobility over time and make comparisons of both levels and trends across countries, a synthesis measure M will be used (Dickens, 2000): 2 M¼
N X
jrank it rank is j
i¼1
N
ð2Þ
where rank stands for percentile rank, the subscripts t and s refer to the year, and N refers to the total number of workers. M is twice the average absolute jump in rank (percentile) and it ranges between 0 (no mobility) and 1 (maximum mobility). It is computed over workers who are wage-earners in both periods under analysis. Since wages tend to rise over the life cycle, the analysis is conducted removing the age effect from wages. Actual mobility is evaluated using the residuals of a wage regression on age dummies.6 Mid-90s were a turning point in the evolution of the wage distribution in Portugal, in particular for males. Until 1994, wage inequality was rising, and afterwards it has been declining; until 1993, mobility was high, and after 1994 it has been lower (Fig. 3). Rising inequality was thus associated with relatively high mobility, whereas declining inequality coexisted with lower mobility. 4
For both countries, employer reported worker level data are used. Using data from the Family Expenditure Survey, instead of the New Earnings Survey, both reported in Machin (1998), indicates a similar level of wage dispersion in Portugal and the UK in 1994, with lower departing values in Portugal in 1986 and a faster increase in inequality during that period. 5 In the UK, the national minimum wage was introduced in 1999, and its impact is therefore beyond the time frame of the data presented. 6 The patterns and trends are not affected by this change, but the mobility levels are lower.
A.R. Cardoso / Labour Economics 13 (2006) 387–404
Fig. 3. Trends in wage mobility (index M), Portugal and the UK. Source: For Portugal- Computations based on Portugal, MTSS (1986–1999); for the UK- Dickens (2000). Note: For Portugal, information for 1989–90 and 1990–91 are not available, and for 1993–94 it has not been used in the plots. Thus the values for those periods result from linear extrapolation of the trend.
395
396
A.R. Cardoso / Labour Economics 13 (2006) 387–404
Table 1 One-year mobility index M, by decile of origin, Portugal MALES
period rising inequality
period declining inequality
decile origin
88–89
91–92
92–93
94–95
95–96
96–97
97–98
change in av. mobility between the two periods
All 1 2 3 4 5 6 7 8 9 10
0.150 0.128 0.157 0.178 0.186 0.185 0.175 0.157 0.134 0.110 0.088
0.146 0.129 0.151 0.165 0.171 0.172 0.165 0.151 0.135 0.109 0.114
0.146 0.131 0.153 0.164 0.168 0.169 0.164 0.148 0.135 0.109 0.122
0.129 0.114 0.132 0.144 0.147 0.144 0.137 0.130 0.116 0.094 0.136
0.125 0.120 0.138 0.147 0.149 0.146 0.135 0.124 0.110 0.087 0.094
0.129 0.111 0.138 0.151 0.154 0.150 0.142 0.132 0.115 0.091 0.101
0.129 0.131 0.151 0.158 0.157 0.151 0.140 0.128 0.108 0.086 0.079
1.151 1.087 1.098 1.126 1.150 1.185 1.212 1.184 1.197 1.222 1.054
FEMALES
period rising inequality
period declining inequality
decile origin
88–89
91–92
92–93
94–95
95–96
96–97
97–98
between the two periods
All 1 2 3 4 5 6 7 8 9 10
0.138 0.157 0.163 0.161 0.171 0.167 0.152 0.135 0.116 0.087 0.069
0.151 0.169 0.175 0.186 0.193 0.183 0.166 0.138 0.117 0.083 0.098
0.157 0.175 0.184 0.198 0.198 0.189 0.166 0.143 0.119 0.087 0.113
0.136 0.152 0.149 0.155 0.162 0.152 0.138 0.121 0.107 0.087 0.136
0.132 0.152 0.153 0.157 0.164 0.156 0.142 0.127 0.107 0.076 0.085
0.139 0.149 0.163 0.170 0.177 0.166 0.148 0.131 0.113 0.079 0.096
0.143 0.179 0.178 0.183 0.182 0.168 0.153 0.130 0.109 0.074 0.075
1.081 1.058 1.083 1.092 1.094 1.119 1.112 1.091 1.077 1.084 0.950
change in av. mobility
Source: Computations based on Portugal, MTSS (1988–1998). Notes: For 1989–90 and 1990–91, mobility measures cannot be computed because no worker data are available for 1990; similarly, for 1993–94, mobility measures are not presented, because they refer to a wider time horizon, and are thus not comparable to the rest of the series. Different from Dickens (2000), no standard-errors were computed, because the data set includes the population of wage-earners in the private sector.
A crucial question that follows is: Which workers were subject to higher mobility in Portugal until 1993? The low-paid workers? The top wage-earners? Table 1 explores which groups of workers experienced gains in mobility when wage inequality was increasing. It reports the yearly mobility index M by decile of origin, distinguishing between the period when earnings inequality was rising and the period when it was declining.7 The last column reports the change between the two periods in the average yearly mobility index. It is interesting to note a clear pattern regarding identification of the groups of workers who experienced much higher mobility until mid-90s than afterwards. For the top five deciles of male wages in the Portuguese economy, mobility was until mid-90s 7
In 1987, inequality was not yet increasing for females, and in 1999 it was no longer decreasing for males, so those periods were not included.
A.R. Cardoso / Labour Economics 13 (2006) 387–404
397
considerably higher than afterwards (around 20 percent higher). Their earnings instability was higher during that period, with a higher probability of swaps in relative positions. Males on lowest wages were subject until 1993 to somewhat higher mobility than afterwards (9 percent higher, for the first decile). For women, the trend in wage mobility is more homogenous across wage deciles. Comparison between the UK and Portugal for the overlapping period (1987 to 1994) highlights, once again, a notable similarity—mobility levels were very similar at the start, either for males or females (see Fig. 3). Also, in both countries mobility has declined, for males and for females. The pace of decline was faster in the UK, such that by 1994 the degree of mobility was considerably lower there. It can however be noted that male wage mobility kept declining in Portugal, to reach in 1996 values almost as low as the UK two years earlier. In synthesis, both countries became less mobile labour markets, having departed from similar levels of wage mobility in mid-80s. However, mobility declined the most in Portugal for high-wage workers. This pattern contrasts with the results reported by Dickens (2000) for the UK, where mobility declined specially for low-wage workers. The same trend has been reported for the USA by Buchinsky and Hunt (1999).
7. Trends in longer-term mobility Trends in longer-term mobility, over three and five years, can be identified using the mobility measure M, complemented by simple inspection of an informative graphical device (Dickens 2000). It plots on the horizontal axis the rank of the worker in the wage distribution in the initial year, and on the vertical axis his/her rank in the final year. If the worker remained in the same relative position (rank), then (s)he would be represented by a dot on the diagonal of the graph. A higher concentration of points along or close to the diagonal therefore provides indication of lower mobility. Fig. 4 compares mobility in the three-year periods 1988–91 and 1995–98. For longer-term mobility, over five years, the periods 1988–93 and 1993–98 are compared in Fig. 5.8 It can be noted that the concentration of dots along the diagonal increased from 1988– 91 to 1995–98, both for males and females. Mobility over a three–year horizon thus declined. The same trend holds for longer-term mobility, over five years. Thus, between mid-80s and late-90s, wage mobility declined in Portugal irrespective of the time horizon considered (one-, three-or five-year mobility). This result is confirmed by the mobility measures reported in Table 2.
8. Where does the Portuguese wage variability come from? The analysis now turns to explore which type of forces shape wage mobility in the Portuguese economy. The aim is to check the relative importance of institutional forces 8
Note however that the period 1993 to 1998 covers more than five years, because the inquiry took place in March 93 and October 98, therefore overstating the mobility that would have taken place over a five-year horizon.
398 A.R. Cardoso / Labour Economics 13 (2006) 387–404 Fig. 4. Three-year mobility, in the 80s and in the 90s, males and females, Portugal. Source: Computations based on Portugal, MTSS (1986–1999). Note: The plots refer to 25 percent samples of the data set.
A.R. Cardoso / Labour Economics 13 (2006) 387–404 399
Fig. 5. Five-year mobility, in the 80s and in the 90s, males and females, Portugal. Source: Computations based on Portugal, MTSS MTSS (1986–1999). Note: The plots refer to 25 percent samples of the data sets.
400
A.R. Cardoso / Labour Economics 13 (2006) 387–404
Table 2 Mobility (index M) over three year and five-year periods, Portugal Males Females
1988–91
1995–98
1988–93
1993–98
0.230 0.217
0.179 0.190
0.260 0.256
0.231 0.244
Source: Computations based on Portugal, MTSS (1987–1998).
versus firm-level conditions in shaping differences in wage growth. The following specific question is addressed: How much of the variability in annual wage growth across workers is associated with the employer and how much is associated with the collective bargaining unit? The underlying idea is that mobility in the wage distribution will result from different workers having different rates of wage growth. Therefore, the impact of the institutional bargaining unit versus the firm on the variance of wage growth is checked. To capture the institutional framework, two variables available in the data set are used: the collective bargaining contract and the worker category within the contract. In fact, each collective agreement sets wages for detailed categories of workers, which are enforced as minimum contractual wage levels. Firms are free to pay higher wages, and they often deviate from that benchmark, adjusting wages to firm specific conditions. Analysis of variance is performed to check the relative importance of the firm versus collective bargaining in defining differences in wage growth across workers in the Portuguese economy. The employer and the detailed unit in collective bargaining account for about half the dispersion in annual wage growth among males, and a slightly higher share among females (Table 3). In either case, most of the explained variability (around 80 percent) of wage growth is accounted for by the firm, whereas the bargaining unit accounts for just 8 percent. Differences across workers in annual wage growth are therefore mostly shaped by market forces operating at the firm level. Unions, on the other hand, bargain for a much more homogenous wage growth. The remarkable role of firm-level wage adjustments, as opposed to the relatively homogenous wage growth in collective bargaining, suggests that, despite the regulated wage bargaining system, wages are in Portugal quite flexible. This evidence reinforces results by the OECD (1992), when finding that at the macro level wages in Portugal are among the most responsive to changes in the unemployment rate. Studies of wage adjustment at the micro level have also pointed in the same direction. Cardoso and Portugal (2004) found that wage adjustments by firms, beyond the levels defined by collective bargaining, provide ample scope for the firm to adapt the wage levels paid to firm-specific and worker-specific conditions. Similarly, Cardoso (2000) had found that firms diverge widely in the way they reward different worker attributes. The pace of change in the Portuguese economy during the second half of the 1980s is likely to provide a major clue explaining the changes in its wage distribution. During the second half of the 1980s, growth and modernization took place at a fast pace in the Portuguese economy and unemployment decreased from 9 percent in 1985 to 4 percent in 1991, substantially below the European Union average. This framework of growth, in an economy with a shortage of skilled labour,
A.R. Cardoso / Labour Economics 13 (2006) 387–404
401
Table 3 Decomposition of variance of annual wage growth source of variation
degrees of freedom
marginal contribution to sum squares
percent of total sum of squares
percent model sum of squares
marginal F-statistic
Males year barg. unit*year firm*year Model Residual Total
9 169,759 558,046 727,814 907,664
1588 10,407 49,819 61,815 56,141 117,957
1.3 8.8 42.2 52.4 47.6 100
2.6 16.8 80.6 100
2853 0.99 1.44 1.37
Females year barg. unit*year firm*year Model Residual Total
9 92,382 456,290 548,681 450,502
788 5801 36,436 43,026 27,148 70,148
1.1 8.3 51.9 61.3 38.7 100
1.8 13.5 84.7 100
1453 1.04 1.32 1.30
Source: Computations based on Portugal, MTSS (1986–1999).
led to a distribution of rents that favoured the most skilled, entitled to rising wage premia, through rising returns on education (see for example Hartog et al. (2001)).9 The interplay between the pace of economic change and the supply of a highly schooled labour force may have played a role leading to wage adjustments promoted mainly at the firm level.
9. Conclusion This study investigated wage mobility in a regulated institutional framework— Portugal—, where traditional indicators of labour market rigidity would lead one to expect low wage mobility. The evidence gathered indicates instead that an aggregate view—of minimum wage enforcement, unionisation rates and mechanisms of extension of contracts—provides a misleading idea of the constraints actually imposed on wage setting by the institutional framework. Indeed, annual wage adjustments promoted at collective bargaining are rather homogenous across the labour force, but afterwards they differ sharply between firms. Micro conditions prevailing at the firm level thus play a major role shaping wage mobility in Portugal. Despite apparent rigidity, 9 The role of inflation facilitating wage mobility has been checked, by regressing the dispersion of wage growth between firms (or between bargaining units) on the inflation rate (for a fuller description of the empirical model, see Groshen and Schweitzer, 1997). Inflation had a significant impact on the dispersion of wage adjustments across firms and across bargaining units. However, given that the relationship presents an inverted U-shape, the high values of inflation during the second half of the 80s would be associated, ceteris paribus, with lower wage mobility than in the 90s, and inflation by itself does not account for the trend in wage mobility detected in Portugal.
402
A.R. Cardoso / Labour Economics 13 (2006) 387–404
institutions in the Portuguese labour market leave ample scope for firm action when setting wages. Major similarities and some contrasts have been detected between the British and the Portuguese wage distributions, which are hard to reconcile with the idea that a strongly regulated labour market should exhibit lower wage mobility. In fact, mobility levels and trends are very similar in the two countries. Strict employment protection legislation and widespread collective bargaining do not seem to have generated lower wage mobility in Portugal. The evidenced gathered is compatible with the idea that changes in mobility levels and patterns resulted mostly from the interplay between modernization and technological progress, on one hand, and the supply of labour holding high levels of schooling, on the other hand, which has been reflected in firmlevel wage adjustments.
Acknowledgements I am grateful to Jan van Ours for clear guidance on the revision of this paper, and to Pedro Portugal, Alexandru Voicu, the participants in seminars held at the Universita` Cattolica del Sacro Cuore, Milan, and IZA Bonn, the participants in the conferences Portuguese Economic Development in the European Framework 2004 and SPIE 2004, and two anonymous referees for their comments. Funding from the Fundac¸a˜o para a Cieˆncia e a Tecnologia (contract POCTI/ECO/37668/2001) and the Bank of Portugal are acknowledged. I thank the Ministry of Employment, Statistics Department, for access to the data set. The usual disclaimer applies.
Appendix A. Additional tables
Table 4 Transition matrix (wage deciles), Portugal, 1991–92 State in 92 Males
out db*
miss w*
1
2
3
4
5
6
7
8
9
10
State in 91 out db* miss w* 1 2 3 4 5 6 7 8 9 10
32.4 32.8 29.2 27.4 25.6 22.9 20.8 17.7 15.4 14.0 15.1
16.1 47.6 4.4 3.9 3.7 3.5 3.6 3.5 3.5 3.3 2.7 3.5
13.9 2.3 41.4 10.4 2.1 1.1 0.7 0.5 0.4 0.2 0.2 0.5
11.7 2.0 12.0 33.7 12.4 3.0 1.4 0.8 0.5 0.3 0.2 0.4
10.7 2.0 3.7 12.8 31.1 13.4 3.6 1.4 0.7 0.5 0.2 0.4
9.7 2.0 2.0 4.6 13.8 31.2 13.5 2.9 1.1 0.7 0.4 0.5
8.5 2.1 1.2 2.3 4.8 14.2 33.7 13.1 2.6 0.9 0.5 0.6
7.7 1.9 0.8 1.3 2.1 4.4 14.1 37.5 12.7 2.4 0.9 0.6
6.2 1.8 0.5 0.8 1.1 1.7 3.6 14.0 43.0 13.4 2.0 0.9
5.4 1.8 0.3 0.4 0.7 0.9 1.4 3.6 13.9 47.9 12.7 1.8
4.8 2.2 0.3 0.3 0.4 0.5 0.7 1.2 3.0 12.7 55.3 10.4
5.3 2.1 0.8 0.5 0.5 0.6 0.6 0.7 1.1 2.3 10.9 65.3
A.R. Cardoso / Labour Economics 13 (2006) 387–404
403
Appendix A (continued) State in 92 Females
out db*
miss w*
1
2
3
4
5
6
7
8
9
10
State in 91 out db* miss w* 1 2 3 4 5 6 7 8 9 10
34.0 33.9 29.7 26.2 24.9 24.1 24.1 22.9 21.4 18.1 15.9
13.2 37.1 6.5 6.0 6.1 5.9 5.6 5.3 5.0 4.9 4.1 3.5
13.5 3.2 33.2 10.5 4.4 1.5 1.3 0.9 0.6 0.4 0.3 0.5
10.3 3.0 11.7 30.4 11.4 5.7 2.4 1.2 0.5 0.3 0.1 0.3
9.5 3.2 5.8 11.6 27.5 13.3 5.8 1.7 0.8 0.4 0.2 0.4
8.8 3.0 2.8 5.4 13.8 28.8 13.0 4.0 1.0 0.5 0.2 0.4
8.7 3.0 2.3 2.9 5.7 12.2 29.7 13.2 2.6 0.8 0.4 0.5
8.7 2.9 1.4 1.5 2.6 4.5 12.0 32.5 13.2 2.0 0.6 0.5
7.9 3.1 1.0 0.8 1.0 1.6 3.6 12.6 38.6 12.1 1.3 0.6
7.6 2.7 0.6 0.5 0.5 0.8 1.3 3.0 12.2 45.3 9.1 1.2
6.2 2.4 0.3 0.2 0.3 0.4 0.6 0.8 2.0 10.3 56.4 9.2
5.7 2.5 0.6 0.6 0.5 0.5 0.6 0.7 0.8 1.6 9.1 67.0
Source: Computations based on Portugal, MTSS (1991–1992). Notes: (*) out db- out of the database; miss wmissing wage (workers who are not wage-earners, but instead owners of the company, unpaid family members or members of cooperatives).
References Aaberge, R., Bjorklund, A., Jantti, M., Palme, M., Pedersen, P.J., Smith, N., Wennemo, T., 2002. Income inequality and income mobility in the Scandinavian countries compared to the United States. Review of Income and Wealth 48, 443 – 469. Abraham, K.G., Houseman, S.N., 1994. Does employment protection inhibit labor market flexibility? Lessons from Germany, France, and Belgium. In: Blank, R.M. (Ed.), Social Protection Versus Economic Flexibility. University of Chicago Press, Chicago, pp. 59 – 93. Bertola, G., 1990. Job security, employment and wages. European Economic Review 34, 851 – 886. Bertola, G., Rogerson, R., 1997. Institutions and labor reallocation. European Economic Review 41, 1147 – 1171. Blanchard, O., Portugal, P., 2001. What hides behind an unemployment rate: comparing Portuguese and U.S. labor markets. American Economic Review 91, 187 – 207. Buchinsky, M., Hunt, J., 1999. Wage mobility in the United States. Review of Economics and Statistics 81, 351 – 368. Burkhauser, R.V., Holtz-Eakin, D., Rhody, S.E., 1997. Labor earnings mobility and inequality in the United States and Germany during the growth of the 1980s. International Economic Review 38, 775 – 794. Cardoso, A.R., 2000. Wage differentials across firms: an application of multilevel modelling. Journal of Applied Econometrics 15, 343 – 354. Cardoso, A.R., 2004. Wage mobility: do institutions make a difference? A replication study comparing Portugal and the UK. Institute for the Study of Labor discussion paper 1086, Bonn. Cardoso, A.R., Portugal, P., 2004. Bargained wages, wage drift and the design of the wage setting system. Centre for Economic Policy Research discussion paper 4405, London. Dickens, R., 2000. Caught in a trap? Wage mobility in Great Britain: 1975–1994. Economica 67, 477 – 497. European Commission, 2002. Industrial Relations in Europe. Office for Official Publications of the European Communities, Luxembourg. Gottschalk, P., Moffitt, R., 1994. The growth of earnings instability in the U.S. labor market. Brookings Papers on Economic Activity 2, 217 – 272. Groshen, E.L., Schweitzer, M.E., 1997. The effects of inflation on wage adjustments in firm-level data: grease or sand? Mimeo, Federal Reserve Bank of New York. Hartog, J., Pereira, P.T., Vieira, J.A.C., 2001. Changing returns to education in Portugal during the 1980s and early 90s: OLS and quantile regression estimators. Applied Economics 33, 1021 – 1037.
404
A.R. Cardoso / Labour Economics 13 (2006) 387–404
Lazear, E.P., 1990. Job security provisions and employment. Quarterly Journal of Economics 105, 699 – 726. Lindbeck, A., Snower, D.J., 1989. Macroeconomic policy and insider power. American Economic Review 79, 370 – 376. Machin, S., 1998. Recent shifts in wage inequality and the wage returns to education in Britain. National Institute Economic Review 166, 87 – 96. OECD [Organisation for Economic Co-operation and Development], 1992. OECD Economic Surveys, Portugal 1991/92. OECD, Paris. OECD [Organisation for Economic Co-operation and Development], 1994. OECD Employment Outloook 1994. OECD, Paris. OECD [Organisation for Economic Co-operation and Development], 1997. Earnings mobility: taking a longerrun view. OECD Employment Outlook 1997. OECD, Paris, pp. 27 – 61. OECD [Organisation for Economic Co-operation and Development], 1999. OECD Employment Outloook 1999. OECD, Paris. Portugal, Ministe´rio do Trabalho e da Seguranc¸a Social, 1986 to 1999. Quadros de Pessoal. Data in magnetic media.
Wage mobility: do institutions make a difference? Ana Rute Cardoso* IZA Bonn, Schaumburg-Lippe-Str 7/9, 53113 Bonn, Germany University of Minho and CEPR Received 1 November 2003; received in revised form 1 May 2004; accepted 1 October 2004 Available online 27 April 2005
Abstract This paper checks empirically the assertion that strict labour market regulation leads to low wage mobility, addressing an apparent puzzle. Indeed, most economic reasoning links a combination of regulations in the labour market to low wage flexibility and mobility, but the scarce empirical evidence available challenges that view. I focus on Portugal, one of the most regulated labour markets in Europe. The evidence gathered indicates that an aggregate view—of minimum wage enforcement, unionisation rates and extension of collective bargaining contracts—provides a misleading idea of the actual constraints imposed by the institutional framework on wage setting. Instead, micro conditions at the firm level play a major role shaping wage mobility in Portugal. Some comparisons with the UK, traditionally pointed out as a flexible labour market, are provided. Remarkable similarities in mobility level and trend are detected between the two countries, further suggesting that a regulated institutional framework does not necessarily reduce individual mobility in the wage distribution. D 2004 Elsevier B.V. All rights reserved. JEL classification: J31; J60 Keywords: Wage dispersion; Collective bargaining
1. Introduction Most economic reasoning links a combination of regulations in the labour market— strict employment protection legislation, national minimum wages, and widespread * Tel.: +49 2283894508; fax: +49 2283894510. E-mail address: [email protected]. 0927-5371/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.labeco.2004.10.004
388
A.R. Cardoso / Labour Economics 13 (2006) 387–404
influence of collective bargaining on wage setting—to low wage flexibility and mobility. However, empirical evidence, though scarce, challenges that view. Burkhauser et al (1997) point to the remarkable similarity between the USA and Germany in terms of mobility levels and trend, discarding the hypothesis that countries with higher Government intervention in the labour market and higher union influence will have lower wage mobility. Similarly, Aaberge et al. (2002) detect lower levels of earnings mobility in the USA than in the Scandinavian countries. Also, the OECD (1997) found that less regulation in labour and product markets, in countries such as the UK and the USA, does not result in higher levels of wage mobility. This paper addresses this apparent puzzle, by concentrating on wage mobility in Portugal. It checks whether empirical support is found for the assertion that strict labour market regulations lead to low wage mobility. In Portugal collective bargaining is extensively applied, both to unionised and non-unionised workers, minimum wages have long been enforced, and employment protection legislation is among the most strict in the OECD. A comparison with the UK is provided, relying on the work by Dickens (2000). As opposed to Portugal, the British labour market is recognized as one of the most flexible in Europe. Specifically, the following questions are addressed: What is the extent of wage mobility in Portugal? Is a strongly regulated labour market likely to protect low-wage workers, sustaining their wages, while nevertheless reducing their chances of upward mobility? Which forces drive wage changes in Portugal: mainly collective bargaining, or firmspecific conditions? How does mobility in Portugal compare with the UK, two countries with very different institutional frameworks? A remarkable longitudinal matched employer-employee data set, covering all the wageearners in manufacturing and services private sector in Portugal, is used. Problems commonly faced by panel data sets, such as under-sampling or over-sampling of certain groups and panel attrition, are thus reduced. Also, the legal requirement for the data to be posted in the company contributes to its reliability, reducing measurement errors. Section 2 outlines the major contrasts between the institutional framework in the British and Portuguese labour markets. Section 3 discusses theoretical and empirical evidence on the link between labour market regulations and wage mobility. Section 4 describes the data set used. Section 5 overviews trends in wage dispersion, while Sections 6 and 7 concentrate on wage mobility. Section 8 discusses the results, identifying the role of institutional forces versus firm level conditions shaping wage variability across workers. Concluding comments are presented in Section 9.
2. Labour market institutions in Portugal and the UK The UK and Portugal shared, like most industrialized economies, a remarkable drop in unionisation rates over recent decades. In Portugal the unionisation rate reached 61 percent in 1970 and 1980, but by 1990 it had dropped to 32 percent; a similar trend was followed in the UK, where it dropped from 50 percent in 1980 to 39 percent in 1990 (OECD, 1994: 184). Nevertheless, collective bargaining has a sharply different role in these two economies. While it covers virtually all the workforce in Portugal, just approximately one fourth of the
A.R. Cardoso / Labour Economics 13 (2006) 387–404
389
workers in the UK are covered (European Commission, 2002: 47). Such contrast results from different mechanisms of extension of contracts. Voluntary extensions are common in Portugal, when a workers’ or employers’ representative subscribes to an agreement it had initially not signed. Also, agreements subscribed by an employer most often apply to all of the company’s workforce, irrespective of the worker’s actual union membership status. This is the default mechanism in Portugal that results from a legal provision, whereas in the UK it applies more rarely and on a voluntary basis (European Commission, 2002: 39– 47). Moreover, in Portugal a mandatory bargaining regime can be applied, when workers are not covered by unions, when one of the parties involved refuses to negotiate or bargaining is obstructed in any other way, in which case a contract is imposed by the Government (Cardoso and Portugal, 2004). The degree of centralization of collective bargaining is also different in these two countries. Agreements covering a whole industry predominate among the Portuguese workforce. Centralized negotiations take place yearly at the national level between trade union confederations, employersT federations and the Government, setting indicative guidelines for wage increase. It should also be noted that bargaining is highly synchronized in the Portuguese economy, with wage clauses in collective contracts updated yearly, most often to take effect January each year. Job protection is another aspect of the institutional framework clearly setting the British and the Portuguese systems apart. Portugal is invariably pointed out as one of the economies in the OECD with highest employment protection, whereas the UK shows up at the other extreme, with one of the lowest levels of employment protection (see for example OECD, 1999). During the period under analysis, the minimum wage legislation was also different across the two countries. A national minimum wage was not enforced in the UK, while in Portugal it was quite binding, reaching in 1999 44 percent of the mean wage in the economy.
3. Why would different labour market institutions lead to different degrees of wage mobility? There is theoretical and empirical support to the idea that employment protection combined with other regulations in the labour market, such as strong union influence and minimum wages, may lead to lower wage mobility. Employment protection tends to reduce worker turnover between firms. In the model by Lindbeck and Snower (1989), higher firing costs raise the power of incumbents, increasing their wages and reducing layoffs. Lazear (1990) argues that more restrictive employment protection legislation leads to less dismissals and less hirings. Relying on macro crosscountry data, he finds that countries with high firing costs have lower employment levels and higher unemployment rates. The theoretical and empirical analyses by Bertola (1990) also find that employment protection lowers flows in the labour market, but highlight that the average employment in the economy is not necessarily lower—it is just more stable, as firms opt for a less fluctuating workforce. Abraham and Houseman (1994) compared Germany, Belgium, France and the USA, having found that the adjustment of employment
390
A.R. Cardoso / Labour Economics 13 (2006) 387–404
to changes in output is faster in countries with less employment protection.1 Bertola and Rogerson (1997), on the other hand, claim that yearly job creation and destruction rates are similar across countries with very different degrees of employment protection. Blanchard and Portugal (2001) reconcile this conflicting evidence by analysing quarterly data on the USA and Portugal, which enabled the identification of patterns that are not apparent from yearly data. They found that behind similar unemployment rates hide two labour markets operating in very different ways: high employment protection leads in Portugal to much lower flows of workers into unemployment—but a much higher duration of unemployment—, which results from lower job creation and destruction rates, together with lower flows of workers given the rate of job creation and destruction. There thus seems to be support to the assertion that employment protection reduces in Portugal worker turnover among firms. One major source of wage mobility—the transition between firms—is thus reduced.2 Tougher job protection legislation could therefore reduce the scope for wage mobility. Different authors who have addressed explicitly the link between job protection and wage flexibility have nevertheless stressed that job protection by itself does not reduce wage flexibility. In particular, Bertola (1990) found that, in countries with high employment protection, wages react more to the unemployment rate. He therefore concludes that the impact of employment protection reducing wage flexibility may be undone by other labour market institutions, and that it would take a combination of institutions—employment protection legislation, widespread trade union influence and national minimum wages—to reduce wage flexibility. Such combination of institutions is certainly present in the Portuguese labour market, as described in the previous section. That would reinforce the expectation of low wage mobility. Widespread collective bargaining could lead to lower wage mobility, since it defines a strict wage hierarchy, based on the worker category, and clear career progression rules, which in Portugal apply to both unionized and non-unionized workers. Firms do have discretionary power over promotions, but because careers and associated wages are strongly regulated, possibilities of relative wage mobility in the economy would be weakened. Also, high union influence in the economy could give insiders power to protect wages from market forces.
4. Data source and concepts used A large longitudinal data set matching firms and workers in the Portuguese economy is used. The data are gathered annually by the Ministry of Employment, based on an inquiry that every establishment with wage-earners is legally obliged to fill in. Reported data cover all the personnel working for the establishment in a reference week (in March until 1993, and in October from 1994 onwards). Public administration and domestic service are not covered, and the coverage of agriculture is low, due to the low share of wage earners. For the remaining sectors, the mandatory nature of the survey leads to an extremely high 1 2
Whereas the adjustment of hours worked is similar. On the impact of job changes on wage changes, see for example Gottschalk and Moffitt (1994).
A.R. Cardoso / Labour Economics 13 (2006) 387–404
391
response rate and in fact the population of firms with wage-earners in the manufacturing and services private sectors is covered. Reported data on the worker include gender, age, schooling, monthly earnings (split into several components), duration of work, the collective bargaining contract and the detailed worker category within the contract. A worker identification code, based on a transformation of the social security number, enables tracking him/her over time. The panel data set covers the period 1986 to 19993. The final panel includes over 22 million worker-year observations and almost 5 million workers. Extensive checks have been performed to guarantee the accuracy of the data, using gender, date of birth, and highest schooling level achieved (for details on the procedure followed, see Cardoso, 2004). The analysis is carried out separately for males and females and it focuses on workers aged 22 to 59. Workers with missing wage (because they are not wage-earners, but instead owners of the company, unpaid family members or members of cooperatives, according to the Portuguese classification) were kept in the analysis of transition matrices. Once a worker drops out of the database, one of several situations may have occurred: (s)he may have become unemployed, inactive, or an independent worker, may have joined an uncovered sector (Public Administration) or the Army. Even though unemployment is the most likely situation, when analyzing transition matrices, the cell out of the data base should be interpreted with caution. When choosing the time spans under analysis, two constraints should be kept in mind: in 1994, the reference week for the inquiry changed from March to October, and therefore 1993–94 captures changes over a longer period (19 months), rendering it infeasible for comparison with the remaining periods; in 1990, no worker data were reported. Gross hourly earnings were computed as hw ¼
bw þ sen þ reg þ overtw ; normh þ overth
ð1Þ
where bw stands for base-wage, sen are seniority-indexed components of pay, reg are other regularly paid components, and overtw is payment for overtime; normh refers to the normal duration of work and overth are overtime hours. Wages were deflated using the Consumer Price Index.
5. Wage dispersion: trends and patterns Wage dispersion increased at a fast pace in Portugal until mid-90s (Fig. 1), slowly declining afterwards. Fig. 2 considers separately dispersion at the bottom and at the top halves of the distribution, reporting the ratios between the percentiles 50/10 and 90/50. It shows that the reported overall trend was shared by the top and the bottom of the distribution, since in either case dispersion increased until mid-90s and declined afterwards.
3
Note, however, that no worker data were reported for the year 1990.
392 A.R. Cardoso / Labour Economics 13 (2006) 387–404
Fig. 1. Trends in wage dispersion (Gini index), Portugal and the UK. Source: For Portugal- Computations based on Portugal, MTSS (1986–1999); for the UK- Machin (1998). Note: The values for Portugal in 1990 result from linear extrapolation of the trend, since no worker data are available.
A.R. Cardoso / Labour Economics 13 (2006) 387–404 Fig. 2. Trends in wage dispersion at the bottom and top of the distributions, Portugal and the UK. Source: for Portugal- computations based on Portugal, MTSS (1986– 1999); for the UK- Machin (1998). Note: The values for Portugal in 1990 result from linear extrapolation of the trend, since no worker data are available. 393
394
A.R. Cardoso / Labour Economics 13 (2006) 387–404
It is worth highlighting the comparison with the UK. The similarity between the level of wage dispersion in both countries in 1986, the first year with overlapping data, is outstanding.4 Moreover, in both countries inequality increased very sharply until mid-90s. In Portugal it increased at a faster pace up to 1994 but declined afterwards, whereas in the UK it kept increasing. In synthesis, both countries became more unequal labour markets, having departed from similar levels of wage inequality in mid-80s. However, the pattern of wage dispersion contrasts sharply. The bottom part of the distribution is much more compressed in Portugal and, conversely, the top is much more stretched. Similar levels of overall wage dispersion therefore result from a very different pattern. The compression at the bottom in Portugal reflects the impact of the national minimum wage, which pulls low wages closer to the middle of the distribution.5
6. Wage mobility: contrasts and similarities The mobility matrix in Table 4 in Appendix A reports for males a high degree of earnings persistence, with the proportion of workers who do not change decile ranging from 31 percent to 66 percent, depending on the decile of origin. Most of those who travel in the wage distribution, do so for a short distance. These patterns hold for females as well. To check changes in wage mobility over time and make comparisons of both levels and trends across countries, a synthesis measure M will be used (Dickens, 2000): 2 M¼
N X
jrank it rank is j
i¼1
N
ð2Þ
where rank stands for percentile rank, the subscripts t and s refer to the year, and N refers to the total number of workers. M is twice the average absolute jump in rank (percentile) and it ranges between 0 (no mobility) and 1 (maximum mobility). It is computed over workers who are wage-earners in both periods under analysis. Since wages tend to rise over the life cycle, the analysis is conducted removing the age effect from wages. Actual mobility is evaluated using the residuals of a wage regression on age dummies.6 Mid-90s were a turning point in the evolution of the wage distribution in Portugal, in particular for males. Until 1994, wage inequality was rising, and afterwards it has been declining; until 1993, mobility was high, and after 1994 it has been lower (Fig. 3). Rising inequality was thus associated with relatively high mobility, whereas declining inequality coexisted with lower mobility. 4
For both countries, employer reported worker level data are used. Using data from the Family Expenditure Survey, instead of the New Earnings Survey, both reported in Machin (1998), indicates a similar level of wage dispersion in Portugal and the UK in 1994, with lower departing values in Portugal in 1986 and a faster increase in inequality during that period. 5 In the UK, the national minimum wage was introduced in 1999, and its impact is therefore beyond the time frame of the data presented. 6 The patterns and trends are not affected by this change, but the mobility levels are lower.
A.R. Cardoso / Labour Economics 13 (2006) 387–404
Fig. 3. Trends in wage mobility (index M), Portugal and the UK. Source: For Portugal- Computations based on Portugal, MTSS (1986–1999); for the UK- Dickens (2000). Note: For Portugal, information for 1989–90 and 1990–91 are not available, and for 1993–94 it has not been used in the plots. Thus the values for those periods result from linear extrapolation of the trend.
395
396
A.R. Cardoso / Labour Economics 13 (2006) 387–404
Table 1 One-year mobility index M, by decile of origin, Portugal MALES
period rising inequality
period declining inequality
decile origin
88–89
91–92
92–93
94–95
95–96
96–97
97–98
change in av. mobility between the two periods
All 1 2 3 4 5 6 7 8 9 10
0.150 0.128 0.157 0.178 0.186 0.185 0.175 0.157 0.134 0.110 0.088
0.146 0.129 0.151 0.165 0.171 0.172 0.165 0.151 0.135 0.109 0.114
0.146 0.131 0.153 0.164 0.168 0.169 0.164 0.148 0.135 0.109 0.122
0.129 0.114 0.132 0.144 0.147 0.144 0.137 0.130 0.116 0.094 0.136
0.125 0.120 0.138 0.147 0.149 0.146 0.135 0.124 0.110 0.087 0.094
0.129 0.111 0.138 0.151 0.154 0.150 0.142 0.132 0.115 0.091 0.101
0.129 0.131 0.151 0.158 0.157 0.151 0.140 0.128 0.108 0.086 0.079
1.151 1.087 1.098 1.126 1.150 1.185 1.212 1.184 1.197 1.222 1.054
FEMALES
period rising inequality
period declining inequality
decile origin
88–89
91–92
92–93
94–95
95–96
96–97
97–98
between the two periods
All 1 2 3 4 5 6 7 8 9 10
0.138 0.157 0.163 0.161 0.171 0.167 0.152 0.135 0.116 0.087 0.069
0.151 0.169 0.175 0.186 0.193 0.183 0.166 0.138 0.117 0.083 0.098
0.157 0.175 0.184 0.198 0.198 0.189 0.166 0.143 0.119 0.087 0.113
0.136 0.152 0.149 0.155 0.162 0.152 0.138 0.121 0.107 0.087 0.136
0.132 0.152 0.153 0.157 0.164 0.156 0.142 0.127 0.107 0.076 0.085
0.139 0.149 0.163 0.170 0.177 0.166 0.148 0.131 0.113 0.079 0.096
0.143 0.179 0.178 0.183 0.182 0.168 0.153 0.130 0.109 0.074 0.075
1.081 1.058 1.083 1.092 1.094 1.119 1.112 1.091 1.077 1.084 0.950
change in av. mobility
Source: Computations based on Portugal, MTSS (1988–1998). Notes: For 1989–90 and 1990–91, mobility measures cannot be computed because no worker data are available for 1990; similarly, for 1993–94, mobility measures are not presented, because they refer to a wider time horizon, and are thus not comparable to the rest of the series. Different from Dickens (2000), no standard-errors were computed, because the data set includes the population of wage-earners in the private sector.
A crucial question that follows is: Which workers were subject to higher mobility in Portugal until 1993? The low-paid workers? The top wage-earners? Table 1 explores which groups of workers experienced gains in mobility when wage inequality was increasing. It reports the yearly mobility index M by decile of origin, distinguishing between the period when earnings inequality was rising and the period when it was declining.7 The last column reports the change between the two periods in the average yearly mobility index. It is interesting to note a clear pattern regarding identification of the groups of workers who experienced much higher mobility until mid-90s than afterwards. For the top five deciles of male wages in the Portuguese economy, mobility was until mid-90s 7
In 1987, inequality was not yet increasing for females, and in 1999 it was no longer decreasing for males, so those periods were not included.
A.R. Cardoso / Labour Economics 13 (2006) 387–404
397
considerably higher than afterwards (around 20 percent higher). Their earnings instability was higher during that period, with a higher probability of swaps in relative positions. Males on lowest wages were subject until 1993 to somewhat higher mobility than afterwards (9 percent higher, for the first decile). For women, the trend in wage mobility is more homogenous across wage deciles. Comparison between the UK and Portugal for the overlapping period (1987 to 1994) highlights, once again, a notable similarity—mobility levels were very similar at the start, either for males or females (see Fig. 3). Also, in both countries mobility has declined, for males and for females. The pace of decline was faster in the UK, such that by 1994 the degree of mobility was considerably lower there. It can however be noted that male wage mobility kept declining in Portugal, to reach in 1996 values almost as low as the UK two years earlier. In synthesis, both countries became less mobile labour markets, having departed from similar levels of wage mobility in mid-80s. However, mobility declined the most in Portugal for high-wage workers. This pattern contrasts with the results reported by Dickens (2000) for the UK, where mobility declined specially for low-wage workers. The same trend has been reported for the USA by Buchinsky and Hunt (1999).
7. Trends in longer-term mobility Trends in longer-term mobility, over three and five years, can be identified using the mobility measure M, complemented by simple inspection of an informative graphical device (Dickens 2000). It plots on the horizontal axis the rank of the worker in the wage distribution in the initial year, and on the vertical axis his/her rank in the final year. If the worker remained in the same relative position (rank), then (s)he would be represented by a dot on the diagonal of the graph. A higher concentration of points along or close to the diagonal therefore provides indication of lower mobility. Fig. 4 compares mobility in the three-year periods 1988–91 and 1995–98. For longer-term mobility, over five years, the periods 1988–93 and 1993–98 are compared in Fig. 5.8 It can be noted that the concentration of dots along the diagonal increased from 1988– 91 to 1995–98, both for males and females. Mobility over a three–year horizon thus declined. The same trend holds for longer-term mobility, over five years. Thus, between mid-80s and late-90s, wage mobility declined in Portugal irrespective of the time horizon considered (one-, three-or five-year mobility). This result is confirmed by the mobility measures reported in Table 2.
8. Where does the Portuguese wage variability come from? The analysis now turns to explore which type of forces shape wage mobility in the Portuguese economy. The aim is to check the relative importance of institutional forces 8
Note however that the period 1993 to 1998 covers more than five years, because the inquiry took place in March 93 and October 98, therefore overstating the mobility that would have taken place over a five-year horizon.
398 A.R. Cardoso / Labour Economics 13 (2006) 387–404 Fig. 4. Three-year mobility, in the 80s and in the 90s, males and females, Portugal. Source: Computations based on Portugal, MTSS (1986–1999). Note: The plots refer to 25 percent samples of the data set.
A.R. Cardoso / Labour Economics 13 (2006) 387–404 399
Fig. 5. Five-year mobility, in the 80s and in the 90s, males and females, Portugal. Source: Computations based on Portugal, MTSS MTSS (1986–1999). Note: The plots refer to 25 percent samples of the data sets.
400
A.R. Cardoso / Labour Economics 13 (2006) 387–404
Table 2 Mobility (index M) over three year and five-year periods, Portugal Males Females
1988–91
1995–98
1988–93
1993–98
0.230 0.217
0.179 0.190
0.260 0.256
0.231 0.244
Source: Computations based on Portugal, MTSS (1987–1998).
versus firm-level conditions in shaping differences in wage growth. The following specific question is addressed: How much of the variability in annual wage growth across workers is associated with the employer and how much is associated with the collective bargaining unit? The underlying idea is that mobility in the wage distribution will result from different workers having different rates of wage growth. Therefore, the impact of the institutional bargaining unit versus the firm on the variance of wage growth is checked. To capture the institutional framework, two variables available in the data set are used: the collective bargaining contract and the worker category within the contract. In fact, each collective agreement sets wages for detailed categories of workers, which are enforced as minimum contractual wage levels. Firms are free to pay higher wages, and they often deviate from that benchmark, adjusting wages to firm specific conditions. Analysis of variance is performed to check the relative importance of the firm versus collective bargaining in defining differences in wage growth across workers in the Portuguese economy. The employer and the detailed unit in collective bargaining account for about half the dispersion in annual wage growth among males, and a slightly higher share among females (Table 3). In either case, most of the explained variability (around 80 percent) of wage growth is accounted for by the firm, whereas the bargaining unit accounts for just 8 percent. Differences across workers in annual wage growth are therefore mostly shaped by market forces operating at the firm level. Unions, on the other hand, bargain for a much more homogenous wage growth. The remarkable role of firm-level wage adjustments, as opposed to the relatively homogenous wage growth in collective bargaining, suggests that, despite the regulated wage bargaining system, wages are in Portugal quite flexible. This evidence reinforces results by the OECD (1992), when finding that at the macro level wages in Portugal are among the most responsive to changes in the unemployment rate. Studies of wage adjustment at the micro level have also pointed in the same direction. Cardoso and Portugal (2004) found that wage adjustments by firms, beyond the levels defined by collective bargaining, provide ample scope for the firm to adapt the wage levels paid to firm-specific and worker-specific conditions. Similarly, Cardoso (2000) had found that firms diverge widely in the way they reward different worker attributes. The pace of change in the Portuguese economy during the second half of the 1980s is likely to provide a major clue explaining the changes in its wage distribution. During the second half of the 1980s, growth and modernization took place at a fast pace in the Portuguese economy and unemployment decreased from 9 percent in 1985 to 4 percent in 1991, substantially below the European Union average. This framework of growth, in an economy with a shortage of skilled labour,
A.R. Cardoso / Labour Economics 13 (2006) 387–404
401
Table 3 Decomposition of variance of annual wage growth source of variation
degrees of freedom
marginal contribution to sum squares
percent of total sum of squares
percent model sum of squares
marginal F-statistic
Males year barg. unit*year firm*year Model Residual Total
9 169,759 558,046 727,814 907,664
1588 10,407 49,819 61,815 56,141 117,957
1.3 8.8 42.2 52.4 47.6 100
2.6 16.8 80.6 100
2853 0.99 1.44 1.37
Females year barg. unit*year firm*year Model Residual Total
9 92,382 456,290 548,681 450,502
788 5801 36,436 43,026 27,148 70,148
1.1 8.3 51.9 61.3 38.7 100
1.8 13.5 84.7 100
1453 1.04 1.32 1.30
Source: Computations based on Portugal, MTSS (1986–1999).
led to a distribution of rents that favoured the most skilled, entitled to rising wage premia, through rising returns on education (see for example Hartog et al. (2001)).9 The interplay between the pace of economic change and the supply of a highly schooled labour force may have played a role leading to wage adjustments promoted mainly at the firm level.
9. Conclusion This study investigated wage mobility in a regulated institutional framework— Portugal—, where traditional indicators of labour market rigidity would lead one to expect low wage mobility. The evidence gathered indicates instead that an aggregate view—of minimum wage enforcement, unionisation rates and mechanisms of extension of contracts—provides a misleading idea of the constraints actually imposed on wage setting by the institutional framework. Indeed, annual wage adjustments promoted at collective bargaining are rather homogenous across the labour force, but afterwards they differ sharply between firms. Micro conditions prevailing at the firm level thus play a major role shaping wage mobility in Portugal. Despite apparent rigidity, 9 The role of inflation facilitating wage mobility has been checked, by regressing the dispersion of wage growth between firms (or between bargaining units) on the inflation rate (for a fuller description of the empirical model, see Groshen and Schweitzer, 1997). Inflation had a significant impact on the dispersion of wage adjustments across firms and across bargaining units. However, given that the relationship presents an inverted U-shape, the high values of inflation during the second half of the 80s would be associated, ceteris paribus, with lower wage mobility than in the 90s, and inflation by itself does not account for the trend in wage mobility detected in Portugal.
402
A.R. Cardoso / Labour Economics 13 (2006) 387–404
institutions in the Portuguese labour market leave ample scope for firm action when setting wages. Major similarities and some contrasts have been detected between the British and the Portuguese wage distributions, which are hard to reconcile with the idea that a strongly regulated labour market should exhibit lower wage mobility. In fact, mobility levels and trends are very similar in the two countries. Strict employment protection legislation and widespread collective bargaining do not seem to have generated lower wage mobility in Portugal. The evidenced gathered is compatible with the idea that changes in mobility levels and patterns resulted mostly from the interplay between modernization and technological progress, on one hand, and the supply of labour holding high levels of schooling, on the other hand, which has been reflected in firmlevel wage adjustments.
Acknowledgements I am grateful to Jan van Ours for clear guidance on the revision of this paper, and to Pedro Portugal, Alexandru Voicu, the participants in seminars held at the Universita` Cattolica del Sacro Cuore, Milan, and IZA Bonn, the participants in the conferences Portuguese Economic Development in the European Framework 2004 and SPIE 2004, and two anonymous referees for their comments. Funding from the Fundac¸a˜o para a Cieˆncia e a Tecnologia (contract POCTI/ECO/37668/2001) and the Bank of Portugal are acknowledged. I thank the Ministry of Employment, Statistics Department, for access to the data set. The usual disclaimer applies.
Appendix A. Additional tables
Table 4 Transition matrix (wage deciles), Portugal, 1991–92 State in 92 Males
out db*
miss w*
1
2
3
4
5
6
7
8
9
10
State in 91 out db* miss w* 1 2 3 4 5 6 7 8 9 10
32.4 32.8 29.2 27.4 25.6 22.9 20.8 17.7 15.4 14.0 15.1
16.1 47.6 4.4 3.9 3.7 3.5 3.6 3.5 3.5 3.3 2.7 3.5
13.9 2.3 41.4 10.4 2.1 1.1 0.7 0.5 0.4 0.2 0.2 0.5
11.7 2.0 12.0 33.7 12.4 3.0 1.4 0.8 0.5 0.3 0.2 0.4
10.7 2.0 3.7 12.8 31.1 13.4 3.6 1.4 0.7 0.5 0.2 0.4
9.7 2.0 2.0 4.6 13.8 31.2 13.5 2.9 1.1 0.7 0.4 0.5
8.5 2.1 1.2 2.3 4.8 14.2 33.7 13.1 2.6 0.9 0.5 0.6
7.7 1.9 0.8 1.3 2.1 4.4 14.1 37.5 12.7 2.4 0.9 0.6
6.2 1.8 0.5 0.8 1.1 1.7 3.6 14.0 43.0 13.4 2.0 0.9
5.4 1.8 0.3 0.4 0.7 0.9 1.4 3.6 13.9 47.9 12.7 1.8
4.8 2.2 0.3 0.3 0.4 0.5 0.7 1.2 3.0 12.7 55.3 10.4
5.3 2.1 0.8 0.5 0.5 0.6 0.6 0.7 1.1 2.3 10.9 65.3
A.R. Cardoso / Labour Economics 13 (2006) 387–404
403
Appendix A (continued) State in 92 Females
out db*
miss w*
1
2
3
4
5
6
7
8
9
10
State in 91 out db* miss w* 1 2 3 4 5 6 7 8 9 10
34.0 33.9 29.7 26.2 24.9 24.1 24.1 22.9 21.4 18.1 15.9
13.2 37.1 6.5 6.0 6.1 5.9 5.6 5.3 5.0 4.9 4.1 3.5
13.5 3.2 33.2 10.5 4.4 1.5 1.3 0.9 0.6 0.4 0.3 0.5
10.3 3.0 11.7 30.4 11.4 5.7 2.4 1.2 0.5 0.3 0.1 0.3
9.5 3.2 5.8 11.6 27.5 13.3 5.8 1.7 0.8 0.4 0.2 0.4
8.8 3.0 2.8 5.4 13.8 28.8 13.0 4.0 1.0 0.5 0.2 0.4
8.7 3.0 2.3 2.9 5.7 12.2 29.7 13.2 2.6 0.8 0.4 0.5
8.7 2.9 1.4 1.5 2.6 4.5 12.0 32.5 13.2 2.0 0.6 0.5
7.9 3.1 1.0 0.8 1.0 1.6 3.6 12.6 38.6 12.1 1.3 0.6
7.6 2.7 0.6 0.5 0.5 0.8 1.3 3.0 12.2 45.3 9.1 1.2
6.2 2.4 0.3 0.2 0.3 0.4 0.6 0.8 2.0 10.3 56.4 9.2
5.7 2.5 0.6 0.6 0.5 0.5 0.6 0.7 0.8 1.6 9.1 67.0
Source: Computations based on Portugal, MTSS (1991–1992). Notes: (*) out db- out of the database; miss wmissing wage (workers who are not wage-earners, but instead owners of the company, unpaid family members or members of cooperatives).
References Aaberge, R., Bjorklund, A., Jantti, M., Palme, M., Pedersen, P.J., Smith, N., Wennemo, T., 2002. Income inequality and income mobility in the Scandinavian countries compared to the United States. Review of Income and Wealth 48, 443 – 469. Abraham, K.G., Houseman, S.N., 1994. Does employment protection inhibit labor market flexibility? Lessons from Germany, France, and Belgium. In: Blank, R.M. (Ed.), Social Protection Versus Economic Flexibility. University of Chicago Press, Chicago, pp. 59 – 93. Bertola, G., 1990. Job security, employment and wages. European Economic Review 34, 851 – 886. Bertola, G., Rogerson, R., 1997. Institutions and labor reallocation. European Economic Review 41, 1147 – 1171. Blanchard, O., Portugal, P., 2001. What hides behind an unemployment rate: comparing Portuguese and U.S. labor markets. American Economic Review 91, 187 – 207. Buchinsky, M., Hunt, J., 1999. Wage mobility in the United States. Review of Economics and Statistics 81, 351 – 368. Burkhauser, R.V., Holtz-Eakin, D., Rhody, S.E., 1997. Labor earnings mobility and inequality in the United States and Germany during the growth of the 1980s. International Economic Review 38, 775 – 794. Cardoso, A.R., 2000. Wage differentials across firms: an application of multilevel modelling. Journal of Applied Econometrics 15, 343 – 354. Cardoso, A.R., 2004. Wage mobility: do institutions make a difference? A replication study comparing Portugal and the UK. Institute for the Study of Labor discussion paper 1086, Bonn. Cardoso, A.R., Portugal, P., 2004. Bargained wages, wage drift and the design of the wage setting system. Centre for Economic Policy Research discussion paper 4405, London. Dickens, R., 2000. Caught in a trap? Wage mobility in Great Britain: 1975–1994. Economica 67, 477 – 497. European Commission, 2002. Industrial Relations in Europe. Office for Official Publications of the European Communities, Luxembourg. Gottschalk, P., Moffitt, R., 1994. The growth of earnings instability in the U.S. labor market. Brookings Papers on Economic Activity 2, 217 – 272. Groshen, E.L., Schweitzer, M.E., 1997. The effects of inflation on wage adjustments in firm-level data: grease or sand? Mimeo, Federal Reserve Bank of New York. Hartog, J., Pereira, P.T., Vieira, J.A.C., 2001. Changing returns to education in Portugal during the 1980s and early 90s: OLS and quantile regression estimators. Applied Economics 33, 1021 – 1037.
404
A.R. Cardoso / Labour Economics 13 (2006) 387–404
Lazear, E.P., 1990. Job security provisions and employment. Quarterly Journal of Economics 105, 699 – 726. Lindbeck, A., Snower, D.J., 1989. Macroeconomic policy and insider power. American Economic Review 79, 370 – 376. Machin, S., 1998. Recent shifts in wage inequality and the wage returns to education in Britain. National Institute Economic Review 166, 87 – 96. OECD [Organisation for Economic Co-operation and Development], 1992. OECD Economic Surveys, Portugal 1991/92. OECD, Paris. OECD [Organisation for Economic Co-operation and Development], 1994. OECD Employment Outloook 1994. OECD, Paris. OECD [Organisation for Economic Co-operation and Development], 1997. Earnings mobility: taking a longerrun view. OECD Employment Outlook 1997. OECD, Paris, pp. 27 – 61. OECD [Organisation for Economic Co-operation and Development], 1999. OECD Employment Outloook 1999. OECD, Paris. Portugal, Ministe´rio do Trabalho e da Seguranc¸a Social, 1986 to 1999. Quadros de Pessoal. Data in magnetic media.