As the IWFP methodology focuses on wage changes that are not influenced by worker mobility, the database was restricted to include only the individuals who worked for at least two consecutive years in the same firm (job stayers). This leads to a substantial reduction of the total number of workers. However, the main differentiating characteristics of natives and immigrants described in Section 3.1 remain valid (see Appendix). In the case of the total and natives samples, the number of job stayers is still very large, so, for computational convenience, a 10 percent random sample was selected for calculating the rigidity measures7. In the case of immigrants, the sizes of the samples of job stayers are much smaller. Thus, we used these samples directly in the estimation of the wage rigidity measures.
As expected, the incidence of measurement errors in the QP database is extremely limited and, hence, the differences between the observed distributions of wage changes and the “error-corrected” distributions are virtually nil. We opted to use the observed distributions, but this choice has no significant impact in the results8.
All estimates of nominal and real wage rigidity were done using a regular wage measure, which includes the base wage and all regular benefits. For the estimation of wage rigidity from the firm’s perspective, a more comprehensive analysis would focus on total labour costs per worker, that is, total compensation, which comprises also all non-regular benefits and overtime pay. However, given that we only have data for one month in each year, to use total compensation could result in spurious wage changes, reflecting the payment of non-regular benefits in different months from one year to the other. Another possibility would be to use only the base wage. As a robustness check, we replicated all estimations using the base wage and the main results are qualitatively similar9.
Figure 3 displays the histograms of the observed wage change distributions, i.e., the empirical distributions, and the notional rigidity-free distributions of wage changes for the total sample from 2003 to 2008. Contrasting with the notional wage distributions, negative nominal wage changes are very limited in the empirical distributions and there is very high concentration on the zero change. Between 2003 and 2006, the distribution of nominal wage changes has a second mode near the expected inflation rate (and bargaining and minimum wage reference values). In 2007 and 2008, the distribution of wage changes has three spikes - at zero, at the expected inflation rate value (and bargaining reference value) and at the rate of change of the minimum wage. Despite differences in the relative size of the spikes, the main features of the empirical distributions of nominal wage changes - near absence of negative changes and two- or three-spike distributions - are common to native and immigrant workers10. The fact that the empirical distributions of wage changes show only a small fraction of negative nominal changes and a very high concentration on the nil change suggests resistance to nominal wage declines (nominal rigidity). Furthermore, the existence of a second mode near the expected inflation rate and a smaller concentration in rates immediately below are evidence in favour of real wage rigidity.The hints obtained from the visual inspection of the wage change distributions are confirmed by the estimated measures of nominal and real wage rigidity displayed in Figure 4. The Appendix includes the yearly estimates of the notional and empirical average wage changes, the corresponding average wage sweep-ups, the expected inflation rate or real rigidity bound, and the nominal and real rigidity measures for the whole sample and for the separate samples for natives and immigrants. The IWFP methodology accounts for changes in the estimated moments of the notional distributions of wage changes over years and samples, as well as for changes in the focal points for the computation of real wage rigidity. Therefore, separate estimates for different sub-samples (natives and immigrants, in this case) do not necessarily average out to the estimate for the entire sample.
Let us now examine in more detail the percentage of workers affected by nominal and real wage rigidities. For the total sample, 47.1 percent of the workers who would have a nominal wage cut in the absence of rigidity, have instead nil wage changes in the period 2003-2008. This result not only is influenced by the legal framework associated with the existing barriers to nominal base wage cuts, but is also related to the fact that, even in the absence of legal constraints, firms tend to avoid nominal wage cuts for motivational reasons (see (Bewley 1998) and (Howitt 2002)). The results are relatively similar when native and immigrant workers are analysed separately. In some years, nominal wage rigidity is higher for immigrants than for natives and the opposite happens in other years. On average, in the 2003-2008 period, the difference between natives and immigrants in terms of the incidence of nominal wage rigidity is small (around 1 percentage point).
The differences between natives and immigrants in terms of real wage rigidity are higher. For the total sample, 19.7 percent of the workers who would face a decline in their real wages in a context of wage flexibility, see their wages increase in line with the expected inflation rate in the period 2003-2008. The average percentage of immigrant workers affected by real wage rigidity in this period reaches 22.7 percent, which compares to 18.1 percent for native workers.
The real wage rigidity measures have an irregular behaviour over the 2003-2008 period. The estimated measures tend to increase from 2003 to 2006, are virtually nil in 2007 and increase again in 2008, but to a level below the average of previous years. The reduction in the real rigidity measures in the last two years was common to native and immigrant workers.
Up to 2006, the real wage rigidity measure combined both “strictly” real and institutional rigidities, related to collective bargaining and mandatory minimum wages. In the distribution of wage changes, the share of individuals covered by collective bargaining and/or earning the minimum wage reinforced the spike at the expected inflation rate. In contrast, in 2007 and 2008, instead of having a two-spike distribution of wage changes, the distribution of wage changes has three spikes - at zero, at the expected inflation rate (and wage bargaining reference value), and at the rate of change of the minimum wage. In these years, the IWFP methodology only takes into account the non-zero spike that is more significant in the range of the estimated expected inflation rate. Hence, in both years, the focal points implicit in the real rigidity measures do not reflected the rate of change of the minimum wage and the corresponding real rigidity measures do not include the percentage of minimum wage earners11. Hence, the evolution of the minimum wage induced a break in the series of the real wage rigidity measure in 2007.
In the period until 2006, we obtained higher values of the real rigidity measure for immigrants. Why would immigrants be more prone to real wage freezes? There are several factors that can contribute to this result like differences in terms of tenure, the incidence of the different types of contract, the sectoral distribution of workers or the proportion of minimum wage earners. In 2008, where the estimated real wage rigidity measure captures only “strictly” real rigidity and not the impact of mandatory minimum wages, the percentage of workers affected by real wage rigidity is smaller for immigrants than for natives. So, the higher real wage rigidity for immigrants up to 2006 is likely to reflect the higher concentration of immigrant workers in the minimum wage.To better assess the impact of the minimum wage on the incidence of downward wage rigidity in Portugal, we re-estimated the rigidity measures using a sample that excludes all workers earning the minimum wage (Figure 5). The estimated measures of nominal wage rigidity are higher in the sample that excludes minimum wage earners, but the difference between natives and immigrants is similar. Since minimum wage earners are not subject to nominal wage rigidity, this increase in the measures has a simple statistical justification: the sample of workers is now smaller, but the workers affected by nominal wage rigidity are essentially the same, so the proportion of these workers increases.
The real wage rigidity measures estimated in the sample without minimum wage workers are smaller than the ones estimated in the full sample for both natives and immigrants. However, the exclusion of minimum wage earners from the sample leads to a sharper reduction of the proportion of immigrant workers subject to real wage rigidity (to 16.1 percent), which is now slightly below the one estimated for natives (17.6 percent). So, the higher real wage rigidity measure estimated for immigrants in the total sample appears to be linked to the higher share of minimum wage earners in immigrant employment.As suggested in Figure 3, the mean of the estimated notional wage change distribution is always smaller than that of the empirical wage change distribution, reflecting the existence of wage rigidities. The related wage sweep-ups, which measure the overall increase in the average wage growth due to the presence of rigidities, are higher for immigrants than for natives, in particular due to non-nominal wage rigidity (Figure 6). Recall that, even if the real wage rigidity measures do not capture the effect of legal minimum wages in 2007 and 2008, the average wage sweep-ups always include, by construction, the impact of both real and institutional rigidities on average wages. In the absence of downward real and institutional wage rigidity, the average wages of immigrants would have grown by less 1.9 percentage points per year on average during this period, which compares to 1.2 percentage points for native workers. In comparison, the average wage sweep-up caused by nominal rigidity for immigrant workers is smaller (1.2 percentage points per year) and similar to the one obtained for natives (1.0 percentage points).
5.1 Accounting for heterogeneity by immigrant origin
The previous estimates differentiate between natives and immigrants but assume that wage rigidity is homogeneous among immigrants’ nationalities. As described in Section 3.1, immigrant workers in Portugal are not a homogeneous group, so we re-estimated the model using separate samples for the major immigrant groups in Portugal (Brazil, PALOP and CEEC) and also for the EU15 and China. Immigrants from the EU15 are very different from the average immigrant worker, as they are much more qualified and earn much higher wages, on average. At the other extreme are the immigrants from China, which grew strongly in recent years: they are the least qualified and earn the lowest wages, on average. Table 2 shows the estimated nominal and real wage rigidity measures for each immigrant group, on average in the 2003-2008 period.
The measure of nominal rigidity estimated for immigrants from PALOP is the lowest of all nationality groups considered, including the natives. This result is in line with most labour market theories that predict a lower wage rigidity for low-skilled blue-collar workers. These workers have a lower probability of reducing effort or quitting the job in response to a wage cut. The cost of hiring and training replacements and the loss of firm-specific human capital if the workers quit are also less significant for these workers. In addition, more than 15 percent of immigrants from PALOP are employed in temporary work agencies and in cleaning services, typically jobs with low replacement costs and, hence, more prone to wage cuts. A symmetric reasoning for high-skilled white-collar workers applies to the high nominal rigidity obtained for EU15 immigrants. At odds with these theoretical predictions, the estimated measures of nominal wage rigidity of immigrants from CEEC, Brazil and China are above those of native workers. However, this result is in line with other studies that found that nominal wage rigidity tends to be high at the bottom of the earnings distribution (see (Du Caju et al. 2007) for Belgium).
The estimated measures of real wage rigidity of the different immigrant groups are above those of native workers, but the difference is very small in the case of PALOP workers. Chinese workers, which have the highest proportion of minimum wage earners, also have the highest value for the real wage rigidity measure. In fact, with the exception of EU15 and PALOP workers, the estimated higher real wage rigidity measure seems to reflect the greater concentration of immigrant workers in the minimum wage. Excluding minimum wage earners from the sample, we obtain a lower incidence of real wage rigidity for these immigrant groups than for native workers.
The combined average wage sweep-ups due to wage rigidity for the various immigrant groups are higher than those estimated for natives (Figure 7). This difference is higher in the case of EU15 and Chinese immigrants: the effect of sweeping up all nominal and real notional wage cuts for workers affected by wage rigidity raises average wage growth by around 3 percentage points per year in both cases. This increase in average labour costs due to downward wage rigidity can have an adverse impact in future labour market outcomes. (Barwell and Schweitzer 2007) found that British workers who are more likely to be protected from wage cuts are also more likely to lose their jobs and (Devicienti et al. 2007) concluded that downward wage rigidities in Italy have a positive impact on worker turnover at the firm level and on local unemployment rates. This fact can be especially important for immigrant groups with a higher share of young and poorly educated workers with fixed-term contracts. The higher educational level and incidence of permanent contracts of EU15 immigrants can shield them more as firms try to replace workers affected by downward rigidities in response to adverse demand shocks.