Abstract
This paper evaluates the short-run impact of the introduction of a statutory minimum wage in Germany on the hourly wages and monthly earnings of workers targeted by the reform. We first provide detailed descriptive evidence of changes to the wage structure in particular at the bottom of the distribution and distinguish between trends for regularly employed and marginally employed workers. In the causal analysis, we then employ a differential trend adjusted difference-in-differences (DTADD) strategy to identify the extent to which these changes in wages and earnings can be attributed to the minimum wage introduction. We find that the minimum wage introduction can account for hourly wage growth in the order of roughly 6.5 % or
1 Introduction
On January 1st, 2015, a coalition government in Germany introduced the country’s first national statutory minimum wage in history. In contrast to most evaluation studies that exploit marginal changes in the existing minimum wage laws in the United States or other countries, the German case study proves particularly interesting because it represents a high-impact, binding minimum-wage introduction with a large share of the population affected. With incremental changes, identification of effects has found mixed results, thus, yielding substantial uncertainty for the derivation of out-of-sample predictions with regard to this highly topical policy tool. This challenge is particularly problematic in light of the fact that substantial increases or new introductions of minimum wages have found their way into current debates in several countries (e.g. the US). Against this background, the German reform offers a unique opportunity to more clearly establish causality and contribute to the broader debate in Germany and around the world. The primary goal of the reform, which set a wage floor of
Due to the recency of its introduction, very few studies have investigated the impact of the statutory minimum wage reform in Germany on the wage and earnings distribution in a causal fashion using data from the post-reform period. Such studies include Bellmann et al. (2017), Caliendo et al. (2017) and Pusch and Seifert (2017). Using data from the IAB Establishment Panel for the state of Saxony, Bellmann et al. (2017) find a strong, positive effect of the reform on gross monthly earnings, not only for workers earning below
The present paper belongs to the first wave of evaluations using post-reform data and builds on these studies mentioned above by providing descriptive as well as causal evidence of wage changes around the time of, or on account of, the minimum wage. Our results also offer evidence regarding the distributional effects of the minimum wage which, with the exception of ex-ante evaluation studies, has been scarce.
Beyond the German context, a large literature has grappled with the distributional effects of minimum wages. In one of the earliest papers to address this question, DiNardo et al. (1996) study the importance of several institutional factors such as the decline in union coverage and the real minimum wage for explaining hourly wage inequalities in the United States. Using CPS data, they attribute 25 % of inequality growth among men and 30 % of inequality growth among women to the decreasing real value of the minimum wage over time. Lee (1999) likewise investigates the relationship between the real minimum wage and wage inequality using CPS data and corroborates this result. More recent papers by Autor et al. (2008) and Autor et al. (2016), however, attribute a much larger role to market factors such as skill-biased technological change rather than minimum wages in explaining wage inequality. Nevertheless, Autor et al. (2008) find that, in particular for women in the lower tail of the distribution, the intertemporal decline in the real minimum wage contributed meaningfully to wage inequality. Going beyond effects on the hourly price of labor (wages) and using the same data as these above mentioned studies, Neumark et al. (2004) explore several channels of minimum wage effects, including monthly earnings, employment probability and hours worked in addition to hourly wages. The authors find the largest increases to hourly wages in the bottom tail of the distribution, but they show that subsequent reductions in the hours worked and employment opportunities counteract the positive wage effect. Allowing for lagged responses to the minimum wage, they moreover find that the overall effect on monthly earnings becomes negative for low-wage workers. With the exception of some of the very recent substantial state hikes in minimum wages, minimum wage adjustments in the long history of the US minimum wage have predominantly been small and incremental in comparison to the bite of the German statutory minimum wage.
In Great Britain, where a national statutory minimum wage was introduced in 1998, several studies assess its impact on the wage and earnings distribution. Manning (2013), Low Pay Commission (2015); Low Pay Commission (2016) provide overviews of this work. Using several different data sources,[2] these studies predominantly conclude that the British minimum wage decreased wage inequality at the lower tail of the distribution (see for example Dickens/Manning 2004; Dolton et al. 2012; Butcher et al. 2012).
The institutional setting, design and bite of minimum wage reforms as well as the pre-reform wage distribution differ greatly from country to country and are likely to influence the effect of reforms in the United States, Great Britain and Germany. Moreover, the spectrum of measured compliance – the degree to which a wage floor is actually enforced – varies substantially across countries and groups of workers as well as over time. Furthermore, measured compliance rates differ depending on whether the data employed in the analysis is based on statements from employees or employers. Ashenfelter and Smith (1979) calculate differences as large as 13 percentage points. Metcalf (2008) arrives at similar results. Previous literature has established different non-compliance rates across groups of workers, with larger rates among workers in low-wage sectors and those with immigrant backgrounds (Cortes, 2004; Weil, 2005). Using rich survey data from the Socio-Economic Panel Survey (SOEP), we are able to quantify the degree of non-compliance with the German minimum wage on average as well as across different types of workers.
The remainder of the paper is structured as follows. Section 2 provides a brief background to the timeline and eligibility rules of the reform. Section 3 describes the data used in the analysis and introduces the econometric method applied to identify causal effects of the reform. Section 4 offers a detailed analysis of trends in wages and salary earnings at the mean as well as separately for the bottom wage segments. Section 5.1 presents the results for the entire sample while Section 5.2 examines heterogeneous treatment effects for individuals in socially insured regular employment and the marginally employed separately. Section 5.3 tests the robustness of our results and Section 6 concludes.
2 Institutional background
Following years of debating the introduction of a minimum wage in Germany, the debate became more concrete during the Federal elections in September 2013 and even more so by the end of November of that same year when an emerging Grand Coalition government announced the intention to implement a national, statutory minimum wage of
3 Methodology
3.1 Data and sample restrictions
Nationally representative data from the 2010 to 2016 waves of the Socio-economic Panel (SOEP) form the basis for our analysis. The SOEP is a panel survey conducted annually in Germany since 1984 and contains about 15,000 households (Goebel et al., 2018). It surveys households regarding their composition, income and relevant employment information, including gross monthly earnings and working hours. Individual hourly wages and gross monthly earnings form the central outcome variables of interest in the present analysis. Although the SOEP does not ask respondents their hourly wage directly, it can be calculated as the quotient of two variables ascertained in the survey, namely monthly earnings and usual weekly hours worked, with the denominator multiplied by 4.33 weeks/month.[4]
Because the field interviews predominantly end in the first half of the year, this time frame enables us to study pre-reform trends, anticipation effects and two years of post-reform effects. Furthermore, the survey asks respondents about their contractual as well as actual hours worked by asking them to report paid and unpaid hours usually worked in their main job.[5] This information allows us to investigate the possible adjustment channel of increased unpaid overtime work. In the following, we refer to the sum of paid and unpaid hours as ’actual hours worked’ and the number of paid hours as ’contractual hours worked’, the latter of which presents the primary measure for analysis, as it is less prone to measurement error.[6]
The SOEP consists of several subsamples that together (weighted) represent the entire population. In this paper, we utilize both the cross-sectional and longitudinal samples, as they possess different, complementary advantages. Central parameters representative of the entire population are constructed using the cross-sectional sample and weights. The measurement of individual changes in hourly wages and monthly earnings, however, requires that individuals were present and employed in at least two consecutive SOEP waves. Thus, for this latter analysis, we employ the panel sample and weights. Together, these two samples enable us to paint a full picture of the minimum wage effects. The following Table 1 summarizes the sample restrictions applied throughout the paper.
2012 | 2013 | 2014 | 2015 | 2016 | Total | |
---|---|---|---|---|---|---|
Employed | 16,155 | 18,199 | 16,066 | 15,822 | 14,895 | 81,137 |
Hourly wage undefined | ||||||
Exempt from minimum wage or | ||||||
has sector-specific minimum wage | ||||||
Cross-Sectional Sample | 9,899 | 11,059 | 10,216 | 9,542 | 9,003 | 49,719 |
Not observed in | -/- | -/- | ||||
Job loss | -/- | -/- | ||||
Missing information | -/- | -/- | ||||
[2mm] 2-Year Panel Sample | 6,133 | 6,703 | 6,475 | -/- | -/- | 19,311 |
Source: SOEP v33 2012-2016, own calculations.
On average, the SOEP contains about 16,000 annual observations of employed individuals above the age of 18. This number includes both full-time and part-time workers as well as the marginally employed and self-employed. We exclude roughly 16 percent of these observations from the sample due to their exemption status from the minimum wage, discussed in detail above. Making these exclusions ensures that treatment and control groups defined in the causal analysis remain comparable. The following analysis applies exclusively to this sample of the population.[7] The remaining 49,719 individuals form the sample population for the cross-sectional analysis.
Building upon the cross-sectional sample, we construct the sample for the longitudinal analysis, referred to subsequently as the panel sample. With a reference year in time t, the panel sample draws upon individual information from the wave two periods later, in time
3.2 Econometric specification
In order to distinguish a causal effect of the minimum wage introduction from secular wage trends that would have developed even absent the reform, we employ a differential trend adjusted difference-in-differences strategy (DTADD) (Blundell/Dias 2009). A difference-in-differences (DD) strategy would not suffice because hourly wages of low-wage workers do not exhibit a parallel trend[9] with any control group and, thus, such estimates would prove biased. For the DTADD strategy, average individual wage growth of the treatment and control group forms the foundation for the analysis. We define the treatment group as individuals earning below
The first difference of the DTADD estimator is defined through the four terms of the group-specific average individual wage growth between time t and
If, for instance, hourly wages in the treatment group increased on average by
In contrast to the DD analysis, this identification strategy does not require the common trend assumption to hold. Rather, it modifies this assumption to stating that existing differences between the treatment and control group would have remained unchanged on average over time. The assumption requires that business cycle effects equally impact the treatment and control group between 2012–2014 and 2014–2016. Given the consistently strong business cycle during these years, this assumption likely holds.[10] A further threat to identification could arise if the control group is affected by the minimum wage. Previous studies have established the existence of such “spillover” effects in the United States and Germany for earlier reforms, at least in the long run (Lee 1999; Neumark et al. 2004; Dickens/Manning 2004; Aretz/Gregory Autor et al. 2016, 2013). Data from the IAB Establishment Panel suggest that also for the statutory minimum wage reform, effects on wages may have also spilled over to higher segments of the wage distribution already in the short run (Mindestlohnkommission, 2016). For this reason, Section 5.3 tests the validity of this assumption and does not find evidence of any significant spillover effects.
To illustrate the identification strategy applied in this paper, Table 2 provides descriptive statistics of wage growth for the treatment and control groups. The sample includes only observations present and employed in both t and
Table 2 shows an average increase in hourly wages of
To place this effect in relation to the individual, pre-reform wage level, we consider the logged change in panel C, which indicates a relative change of 7.4 percentage points additional wage growth between 2014 and 2016 due to the minimum wage introduction. With a mean hourly wage of
DTADD | Placebo | |||||
---|---|---|---|---|---|---|
(1) | (2) | (3) | (4) | (5) | (6) | |
2014/16 | 2012/14 | Difference | 2012/14 | 2010/12 | Difference | |
(1)–(2) | (4)–(5) | |||||
Panel A: Observations | ||||||
Wage< 8.50 | 545 | 549 | 549 | 533 | ||
8.50 | 438 | 412 | 412 | 397 | ||
Panel B: Absolute | ||||||
Change (in Euro) | ||||||
Wage< 8.50 | 2.7 | 2.1 | 0.6 | 2.1 | 2.0 | 0.1 |
(3.8) | (3.9) | (3.9) | (3.3) | |||
8.50 | 1.4 | 1.5 | 1.5 | 1.1 | 0.4 | |
(3.5) | (3.2) | (3.2) | (2.4) | |||
DTADD | 0.7* | |||||
Panel C: Log Change | ||||||
( | ||||||
Wage< 8.50 | 28.8 | 22.5 | 6.3 | 22.5 | 22.0 | 0.5 |
(33.8) | (35.8) | (35.8) | (32.8) | |||
8.50 | 10.5 | 11.6 | 11.6 | 8.6 | 3.0 | |
(26.0) | (23.7) | (23.7) | (22.5) | |||
DTADD | 7.4*** |
Source: SOEP v33 2010-2016, own calculations. Results are based on contractual hourly wages and are Unweighted. Standard errors in parentheses. Significance levels: * p< 0.1, ** p< 0.05, *** p< 0.01.
To complete the analysis of the causal effect of the minimum wage reform on hourly wages, we use regression analysis to additionally control for differential characteristics of the treatment and control group that likewise influence hourly wages, independently of the minimum wage reform. The regression equation can be stated as follows:
4 Trends in wages and salary earnings
4.1 Wage growth throughout the distribution
Before turning our attention to the causal analysis, this section describes trends in gross hourly wages before and after the minimum wage reform. Table 3 presents descriptive statistics for the central variables of the cross-sectional analysis in each year between 2012 and 2016. Among the population of employees eligible for the minimum wage, mean gross monthly earnings amount to
2012 | 2013 | 2014 | 2015 | 2016 | |
---|---|---|---|---|---|
Monthly Gross Earnings | 2,622.77 | 2,649.42 | 2,703.05 | 2,818.06 | 2,846.49 |
in Euros | (1,534.92) | (1,577.11) | (1,639.14) | (1,684.00) | (1,685.27) |
Weekly Hours Worked | 34.18 | 34,10 | 33,75 | 34,03 | 33.98 |
Contractual | (9.81) | (9,64) | (9,96) | (9,78) | (9.77) |
Actual | 37.76 | 37.58 | 36.98 | 37.17 | 37.11 |
(11.72) | (11.65) | (11.81) | (11.59) | (11.51) | |
Hourly Wage in Euros | 17.22 | 17.41 | 17.88 | 18.54 | 18.74 |
Contractual | (8.51) | (8.72) | (9.06) | (9.25) | (9.24) |
Actual | 15.58 | 15.8 | 16.28 | 17.00 | 17.16 |
(7.22) | (7.52) | (7.80) | (8.05) | (8.06) | |
Observations | 9,899 | 11,059 | 10,216 | 9,542 | 9,003 |
Source: SOEP v33 2012–2016, own calculations. The table shows weighted averages based on the cross-sectional sample; standard deviations in parentheses.
A look at the evolution of contractual wages in Germany during the past decades helps to understand the role of the minimum wage introduction for the evolution of wages immediately following the reform. Figure 1 exhibits growth rates in decile-specific average contractual hourly wages throughout the wage distribution for two-year changes between 1998 and 2016. We denote these growth rates ‘anonymized growth rates’ because this procedure measures growth not at the individual level, but rather based on decile-specific averages, which may be comprised of a different pool of people from one year to the next.
The light-grey, dashed lines show the growth rates based on the two years’ difference during the pre-reform period. The black dashed line represents the average over these years before the minimum wage introduction. The black, solid line shows the two year difference between 2014 and 2016. As such, Figure 1 shows that the correlation between wage decile and wage growth systematically differs from the trend in the pre-reform period. The average pre-reform growth rate lies at around 2.5 %, with the upper wage deciles experiencing faster growth at about 3.5 % compared to the lower ones at below 2 %. Between 2014 and 2016 in contrast, wage growth in the lower deciles lay well above the decile-specific average of the past years, accelerating from a meager 1 % average growth to 15 %. At the same time, wage growth in the higher deciles continued at about the same rate after the reform compared to the average of the previous years.
4.2 Changes in wage inequality
Concerns regarding growing wage inequality in Germany motivated support for the minimum wage reform of 2015. For this reason, this section briefly discusses the evolution of wage inequality both prior to and after the minimum wage introduction. The mean log deviation (MLD) in wages serves as a standard measure of inequality, which we also utilize here. Two advantages of MLD make it particularly appropriate for our analysis: it is especially sensitive to changes at the bottom of the wage distribution where the minimum wage binds and it can be decomposed into wage difference between and within groups (see for example Cowell 2011). Table 4 displays the MLD for the entire cross-section. The first row shows the lower and the third row the upper limit of the 95 % confidence interval. The second row contains the point estimate.
2012 | 2013 | 2014 | 2015 | 2016 | |
---|---|---|---|---|---|
Lower Limit | 0.111 | 0.112 | 0.116 | 0.113 | 0.106 |
Point Estimate | 0.115 | 0.117 | 0.121 | 0.118 | 0.111 |
Upper Limit | 0.120 | 0.121 | 0.125 | 0.122 | 0.116 |
Source: SOEP v33 2012–2016, cross-sectional sample, own calculations. Results are weighted and lower and upper limits refer to a 95% confidence interval using a bootstrapping procedure with 200 iterations.
While inequality in the entire cross-sectional sample increased during the period from 2012 to 2014, this trend reversed by 2015, in the year immediately following the introduction of the minimum wage. Table 4 shows a statistically significant reduction in inequality of average hourly wages in 2016 compared to 2014 before the introduction of the minimum wage. This result should not be interpreted as a causal effect of the minimum wage, as any number of factors could have contributed to this evolution. Instead, results place the minimum wage reform in the context of increasing wage inequality that began to decrease during the same period as the minimum wage introduction. Section 5 builds on this descriptive evidence by exploring a causal link between wage growth and the reform.
4.3 Developments in the bottom wage segments
The previous section showed trends in average wages throughout Germany. This section focuses on developments in the bottom 40 % of the gross hourly wage distribution between 2012 and 2016. Figure 2 highlights these developments using Pen’s Parade. This graphical concept first sorts all workers in the given year according to their hourly wage, from lowest to highest. The next step entails plotting the average wage in each percentile against each consecutive percentile of workers. Plotting Pen’s Parade for between 2012 and 2016 together allows for a comparison of wage growth over several years. Figure 2 shows Pen’s Parade for workers earning in the bottom 40 % of the hourly wage distribution.
The left-hand panel of Figure 2 depicts Pen’s Parade of average hourly wages calculated on the basis of contractual hours worked and the right-hand panel the corresponding rates on the basis of actual hours worked. The figure confirms and shows in more detail the wage growth in the low-wage segments targeted by the reform. In addition, the image quantifies the share of workers earning below the minimum wage level of
4.4 Individual wage growth
Following the cross-sectional analysis of changes in (anonymized) wages throughout the distribution, this section utilizes the panel sample in order to focus on changes to the wages of individuals who earned below the minimum wage prior to the reform and remained employed after its introduction. Figure 3 illustrates these changes with a personalized wage growth curve. The personalized wage growth curve describes the relationship between average individual wage growth and the individual’s position in the wage distribution in the initial, pre-reform period. Whereas a Pen’s Parade depicts how the wage, for example, in the
Like the anonymized (cross-sectional) wage growth curve in Figure 1, Figure 3 exhibits the change in wages from 2014 to 2016 as a solid, thick black line and juxtaposes this growth to historical two-year changes in wages between 1998 and 2014. Light grey, dashed lines capture individual two-year changes and the thick black dashed line the average of these between 1998 and 2014.
In the bottom decile of the wage distribution for each initial period, wages grow by 30–40 % and then sink to a rate under 20 % by the second decile, indicating historically high growth rates of the average individual with the lowest wages in the initial period. High growth rates at the bottom demonstrate that, for many individuals, low wages represent a transitory phenomenon. Workers with wages in the lowest decile tend to be young with short work biographies who then gain human capital and work experience that subsequently promote them into higher wage categories. From 2014 to 2016, growth at the bottom increased even further, to about 50 %.
4.5 Mobility between wage segments
This section examines the transitions of individuals across wage segments of the distribution, as workers may occupy different positions throughout their working biography. For this exercise, we use transition matrices to illustrate mobility. The matrices describe the probability to transition from a wage segment in time t to another segment in time
Not | Below | EUR 8.50 | EUR 10.50 | Above EUR | |
---|---|---|---|---|---|
Employed | EUR 8.50 | 12.00 | |||
Wage Group in 2014 | |||||
Wage Group in 2012 | |||||
Not Employed | 0.922 | 0.021 | 0.015 | 0.007 | 0.035 |
(0.004) | (0.002) | (0.002) | (0.001) | (0.003) | |
Below EUR 8.50 | 0.274 | 0.379 | 0.198 | 0.078 | 0.07 |
(0.027) | (0.029) | (0.024) | (0.024) | (0.014) | |
EUR 8.50-10.50 | 0.137 | 0.073 | 0.369 | 0.185 | 0.236 |
(0.020) | (0.015) | (0.028) | (0.025) | (0.027) | |
EUR 10.50-12.00 | 0.132 | 0.011 | 0.069 | 0.314 | 0.474 |
(0.025) | (0.004) | (0.019) | (0.042) | (0.039) | |
Above EUR 12.00 | 0.093 | 0.006 | 0.01 | 0.013 | 0.879 |
(0.007) | (0.001) | (0.003) | (0.002) | (0.008) | |
Wage Group in 2016 | |||||
Wage Group in 2014 | |||||
Not Employed | 0.930 | 0.011 | 0.015 | 0.005 | 0.039 |
(0.004) | (0.002) | (0.002) | (0.001) | (0.003) | |
Below EUR 8.50 | 0.217 | 0.235 | 0.302 | 0.097 | 0.149 |
(0.026) | (0.024) | (0.030) | (0.023) | (0.027) | |
EUR 8.50-10.50 | 0.162 | 0.073 | 0.377 | 0.21 | 0.177 |
(0.025) | (0.013) | (0.029) | (0.023) | (0.023) | |
EUR 10.50-12.00 | 0.166 | 0.029 | 0.116 | 0.237 | 0.452 |
(0.032) | (0.016) | (0.021) | (0.031) | (0.037) | |
Above EUR 12.00 | 0.094 | 0.004 | 0.008 | 0.016 | 0.878 |
(0.006) | (0.002) | (0.002) | (0.003) | (0.007) |
Source: SOEP v33 2012-2016, panel sample. N = 14,538 in the sample 2012-2014 and N = 14,398 in the sample 2014-2016. All probabilities stated in decimal value (0.285 = 28.5%). Standard deviations in parentheses.
Each row describes a certain wage group status in the initial period, 2012 for the upper panel and 2014 for the lower panel. The columns represent the share of each group that transitions from the given wage group to the wage group denoted in the column title (each row adds up to one). The shares in the main diagonal correspond to the share of each wage segment that remained in that wage group two years later. The table shows that the share of individuals that remained in employment remunerated below the minimum wage level of
The descriptive evidence provided in this section paints a clear picture: following many years of low wage growth at the bottom of the wage distribution, the introduction of the statutory minimum wage is associated with significant growth in wage dynamics in the bottom decile of the distribution.[11] and, consequently, a compression of the wage distribution. Nevertheless, compliance with the minimum wage remains imperfect and many eligible workers still earn an hourly wage below
5 Results of the causal effects analysis
5.1 Main results
Table 6 summarizes the results from the regression analysis for changes in contractual hourly wages of all workers eligible for the minimum wage. In order to control for non-linear relationships, the dependent variable is defined in logarithmic rather than absolute terms. Therefore, coefficients should be interpreted as percentage changes. In addition to showing results for the two-year changes, (Columns (4)–(6)), Table 6 also provides results for one-year changes (Columns (1)–(3)) in order to describe potential differences in the effects across the time period following the reform. Columns (1) and (4) present results for the baseline specification using only treatment indicators and year fixed effects as control variables. Columns (2) and (5) additionally include sociodemographic and employment characteristic controls. Columns (3) and (6) also include controls for changes in employment.
One-Year Analysis | Two-Year Analysis | |||||
---|---|---|---|---|---|---|
(1) | (2) | (3) | (4) | (5) | (6) | |
One-Year Analysis | ||||||
Hourly wage < EUR 8.50 | 10.83*** | 12.57*** | 12.49*** | |||
(1.57) | (1.59) | (1.59) | ||||
4.01* | 4.13* | 3.96* | ||||
(2.14) | (2.12) | (2.11) | ||||
(2.19) | (2.17) | (2.17) | ||||
Two-Year Analysis | ||||||
Hourly wage < EUR 8.50 | 10.89*** | 12.59*** | 12.93*** | |||
(1.92) | (1.95) | (1.94) | ||||
7.44*** | 6.75** | 6.47** | ||||
(2.71) | (2.68) | (2.68) | ||||
2.49 | 2.29 | 2.07 | ||||
(2.63) | (2.61) | (2.59) | ||||
Constant | 6.62*** | 13.44*** | 7.31** | 11.58*** | 20.04*** | 10.59*** |
(0.92) | (2.77) | (3.45) | (1.17) | (3.55) | (4.09) | |
Control Variables | ||||||
Year fixed effects | yes | yes | yes | yes | yes | yes |
Socio-demographic info. | yes | yes | yes | yes | ||
Employment characteristics | yes | yes | yes | yes | ||
Changes in | yes | yes | ||||
Employment | ||||||
Observations | 3,523 | 3,523 | 3,523 | 2,874 | 2,874 | 2,874 |
Adj. R | 0.043 | 0.081 | 0.085 | 0.056 | 0.087 | 0.098 |
Source: SOEP v33 2010-2016, own calculations. Robust standard errors in parentheses, clustered at the individual level. Significance levels: * p< 0.1, ** p< 0.05, *** p< 0.01.
The first row of Table 6 quantifies the differential wage dynamics within the one-year panel sample between treatment and control groups. According to the one-year analysis, hourly wages of workers earning below
The two-year analysis shows that hourly wages of employees earning below
Finally, the third row of results for the one-year and two-year analyses lend credence to the validity of the crucial identifying assumption for the DTADD approach, namely time-persistent differences between treatment and control groups. Specifically, the placebo test examines whether wages of the treatment and control groups grew at different speeds during the period of 2012–2013 than in 2013–2014. The observed differences prove statistically insignificant in all specifications.
To answer the question of whether the minimum wage not only increased hourly wages, but also the overall labor income position of the target group, it becomes necessary to consider the effect on monthly earnings. After all, the goal of the reform was not just to increase wages per hour, but rather to improve the economic situation of low-income individuals. Monthly earnings combine two possible dimensions of adjustment: hours worked and hourly wages. In a related contribution Burauel et al. (2019) demonstrate that the minimum wage introduction not only increased wages, but also had a negative impact on average hours worked. Therefore, in the following, we investigate the net effect of these two opposing forces. Table 7 depicts the results of the DTADD estimation from eq. (3) where the change in gross monthly earnings replaces the change in wages as the dependent variable.
One-Year | Two-Year | |||
---|---|---|---|---|
Analysis | Analysis | |||
(1) | (2) | (3) | (4) | |
Hourly wage < EUR 8,50 | 10.98*** | 10.54*** | 8.91*** | 8.24*** |
(1.91) | (1.90) | (2.42) | (2.42) | |
1.39 | 1.09 | |||
(2.69) | (2.63) | |||
8.70** | 6.58* | |||
(3.54) | (3.40) | |||
(2.71) | (2.64) | |||
4.15 | 3.08 | |||
(3.53) | (3.46) | |||
Constant | 7.79*** | 17.68*** | 12.69*** | 11.63** |
(1.25) | (4.24) | (1.57) | (5.23) | |
Control Variables | ||||
Year fixed effects | yes | yes | yes | yes |
Sociodemographic information | yes | yes | ||
Employment characteristics | yes | yes | ||
Changes in employment | yes | yes | ||
Observations | 3,523 | 3,523 | 2,874 | 2,874 |
Adj. R | 0.022 | 0.071 | 0.027 | 0.122 |
Source: SOEP v33 2010–2016, own calculations. Robust standard errors in parentheses, clustered at the individual level. Significance levels: * p< 0.1, ** p< 0.05, *** p< 0.01.
The one-year analysis of Table 7 reveals that the effect of the minimum wage on the gross monthly earnings of the treatment group could not be statistically distinguished from zero. Similar to the results for changes in (log) wages, gross monthly (log) earnings experienced a higher growth rate in the treatment group than in the control group during the period under investigation (2012–2015): the first row of the first two columns indicates that earnings grew by roughly 10.5% more in the treatment than in the control group. The minimum wage, however, did not affect this relationship. Although the minimum wage led to a rise in hourly wages, it also lowered hours worked for this same group. In sum, the net impact of the minimum wage on monthly earnings of the target group can not be distinguished from zero in the one-year horizon.
5.2 Heterogeneity of effects by employment type
As argued in Burauel et al. (2018), one would expect the introduction of the minimum wage to differentially impact employees with socially insured positions and the marginally employed. Marginally employed individuals have an incentive to reduce their hours worked in order to remain below the threshold of
The lower panel of Figure 4 shows the evolution of monthly gross earnings for the socially insured compared to the marginally employed workers between 2012 and 2016. The left panel reveals only a slight improvement for the socially insured in terms of gross monthly earnings. For the marginally employed, in contrast, the share of the group earning below the tax threshold of
Turning to the causal effects for regularly employed and marginally employed workers separately, Table 8 shows results based on the two-year panel sample. The subgroup analysis further splits the sample of regularly employed individuals into full- and part-time categories to examine potential heterogeneous treatment effects.[12] This further partition reduces the sample size and, thus, the power of the separate regressions compared to using the full sample. Analogously to the results of the full sample, the column titled “Hourly Wage < 8.50” reflects the different wage dynamics between treatment and control groups. “DTADD 2014–2016” identifies the change in hourly wage attributable to the minimum wage introduction. “Placebo 2010–2012” tests the critical identification assumption for the DTADD, namely whether wage differences between treatment and control groups can be considered time-constant. All regressions consider the full set of controls.
Observations | ||||||||
---|---|---|---|---|---|---|---|---|
2014 | ||||||||
Treatment | Control | Hourly Wage | DTADD | Placebo | ||||
Group | 2014-2016 | 2010-2012 | ||||||
Panel A: Hourly Wages | ||||||||
Entire Sample | 545 | 438 | 12.93*** | (1.94) | 6.47** | (2.68) | 2.07 | (2.59) |
Socially Insured Workers | 382 | 383 | 11.79*** | (2.15) | 4.57 | (2.99) | 2.36 | (2.85) |
Full-Time Regularly Employed | 270 | 303 | 10.73*** | (2.46) | 7.79** | (3.44) | 3.64 | (3.19) |
Part-Time Regularly Employed | 112 | 80 | 14.53*** | (4.52) | (6.07) | (6.65) | ||
Marginally Employed | 163 | 55 | 17.40*** | (4.80) | 15.51** | (6.90) | 2.43 | (6.66) |
[2mm] Panel B: Monthly Earnings | ||||||||
Socially Insured Workers | 382 | 383 | 9.60*** | (2.56) | 3.54 | (3.51) | 5.29 | (3.56) |
Marginally Employed | 163 | 55 | 3.85 | (6.37) | 13.14 | (9.21) | (12.01) |
Source: SOEP v33 2012–2016, own calculations. Robust standard errors in parentheses, clustered at the individual level. All regressions include the full set of controls, including demographic and employment characteristics as well as information regarding changes in employment. Individuals will only appear in the sample in those years for which the row-specific condition is fulfilled. To deal with changes in employment, e.g. from marginal employment to part-time employment, we control for changes in eligibility, job, contract term, company size and sector. Significance levels: * p< 0.1, ** p< 0.05, *** p< 0.01.
The differential analysis according to regular employment status reveals substantial heterogeneity in the treatment effect. During the period under investigation, wage growth was the most dynamic for the marginally employed, followed by the part-time regularly employed, increasing 17.4 % and 14.5 %, respectively, more in the treatment compared to the control group. In contrast, wage growth for full-time employees in the treatment group surpassed that of the control group by 10.7 %. Despite the high growth rates of the part-time regularly employed, this growth cannot be attributed to the minimum wage. For this group, the effect of the reform is negative, but statistically insignificant. The reform did, however, positively impact hourly wages of full-time employees by 7.8 percentage points. According to the subgroup analysis, the minimum wage introduction had the largest, positive effect on the hourly wages of the marginally employed, who experienced a growth rate 15.5 percentage points higher in the treatment than in the control group. Finally, Panel B of Table 8 considers the net effect on gross earnings, which results from changes in hourly wages and in hours worked, for the socially insured and marginally employed separately. Despite positive treatment effects on hourly wages of the marginally employed, the reduction in hours worked counteracts the wage effect compare (Burauel et al., 2019). Neither for socially insured nor for the marginally employed can any positive impact of the minimum wage reform on gross monthly earnings be detected. At the same time, partitioning workers into these categories renders the sample sizes smaller than in the entire sample and the sample size may simply become too small to detect an effect.
5.3 Robustness analysis - Spillover effects
As discussed in Section 3.2, the causal identification of the DTADD treatment effect relies on the assumption that the introduction of the minimum wage did not affect the selected control group. A priori, the direction of potential spillover effects is unclear. On the one hand, rising wage costs in the lower segment of the distribution could cause employers to decrease wages of higher earners in order to pass along the additional costs of the reform. However, in reality, wages tend to be sticky and long-term contracts as well as social norms may prevent employers from doing so. Negative spillover effects in the form of wage compression tend to be associated with new hires rather than the current stock of employees, rendering this type of spillover a predominantly long-term phenomenon. On the other hand, wages may also rise for workers previously earning just above the minimum wage if employers wish to retain the wage structure within their establishment. Data from the IAB Establishment Panel Survey suggest this latter direction is more likely: 14 percent of all responding establishments report increasing wages not only for those previously earning below the minimum, but also for those earning above the mandated threshold (Mindestlohnkommission, 2016). The presence of positive spillover effects would bias the estimates of wage growth downward if the control group does not correctly reflect the counterfactual situation.
The existing literature finds that spillover effects appear mostly in groups earning close to the minimum wage cutoff. For this reason, we test the existence of possible spillover effects by comparing the results of our main specification from Table 6 with a robustness estimation in which we employ an alternative control group consisting of workers earning between
Change in Contractual Hourly Wages | ||||
---|---|---|---|---|
One-Year | Two-Year | |||
Analysis | Analysis | |||
(1) | (2) | (3) | (4) | |
One-Year Analysis | ||||
Hourly wage < EUR 8,50 | 10.83*** | 12.56*** | ||
(1.57) | (1.58) | |||
4.01* | 4.00* | |||
(2.14) | (2.11) | |||
(2.19) | (2.17) | |||
EUR 10 | ||||
(1.43) | (1.42) | |||
(1.92) | (1.88) | |||
1.28 | 1.96 | |||
(1.95) | (1.94) | |||
Two-Year Analysis | ||||
Hourly wage < EUR 8,50 | 18.67*** | 21.41*** | ||
(3.11) | (3.10) | |||
8.28* | 6.48 | |||
(4.53) | (4.44) | |||
1.94 | 1.25 | |||
(4.10) | (4.02) | |||
EUR 10 | ||||
(2.17) | (2.22) | |||
0.56 | ||||
(3.15) | (3.16) | |||
3.71 | 3.05 | |||
(2.88) | (2.88) | |||
Constant | 7.16*** | 6.94** | 15.79*** | 21.59*** |
(1.02) | (2.88) | (1.69) | (4.71) | |
Control Variables | ||||
Year fixed effects | yes | yes | yes | yes |
Sociodemographic information | yes | yes | ||
Employment characteristics | yes | yes | ||
Changes in employment | yes | yes | ||
Observations | 4,927 | 4,927 | 4,036 | 4,036 |
Adj. R | 0.052 | 0.089 | 0.061 | 0.103 |
Source: SOEP v33 2010-2016, own calculations. DTADD regressions, robust standard errors in parentheses, clustered at the individual level with * p< 0.1, ** p< 0.05, *** p< 0.01. Results are unweighted and based on the panel sample.
Table 9 summarizes results for the one-year and two-year comparison and demonstrates that results remain robust to this alternative control group. It shows a positive and statistically significant treatment effect for the original treatment group between 2014 and 2015 (Columns (1) and (2)). For the two-year analysis, the treatment effect has a similar magnitude, although it loses its significance (Columns (3) and (4)). Only a marginal difference exists in the general wage dynamics between workers earning between
6 Conclusion and discussion
In this paper, we descriptively examine the evolution of the wage and earnings structure of German workers around the time of the introduction of the minimum wage reform and causally identify the impact of the reform on the wage and income distribution. The descriptive analyses illustrate an acceleration of wage growth for workers earning below
As a complement to the cross-sectional analysis, the panel allowed for an investigation of individual wage growth and mobility. Particularly high growth rates in the bottom decile of the distribution indicate that very low wages represent a transitory phenomenon for many workers. This group tends to consist of young workers with short employment biographies, who gain experience and quickly transition into higher wage segments. The panel analysis further finds that workers earning below the minimum wage before the introduction had a higher probability of transitioning into higher wage segments than had been the case in previous years: the probability of transitioning into the segment between
Moving beyond the description of trends to the causal analysis, we employ a DTADD strategy to establish the extent to which increases in wages and earnings can be ascribed to the reform of 2015. We find that the minimum wage introduction can account for hourly wage growth in the order of 6.47 percentage points, or
Subgroup analysis according to type of employment (socially insured vs. marginally employed) revealed that the minimum wage had the highest positive impact on the wages of marginal workers, who experienced a 15.5 percentage points higher growth rate on account of the reform, followed by the full-time regularly employed with an additional increase in hourly wages of 7.8 percentage points. Despite positive treatment effects for hourly wages in both of these groups, however, no impact of the minimum wage reform on monthly earnings could be detected when estimating the effect for these groups separately. The absence of an effect may be attributed to a reduction in power (small sample size) after partitioning the sample into the socially insured and marginally employed.
The introduction of the statutory minimum wage in Germany presents a substantial intervention into the labor market. This paper investigated its short-term impacts on the wage and earnings distribution, accounting for detectable effects through the second quarter of 2016. Evaluations of minimum wages in other countries have established that the full implementation of national, statutory minimum wages tend to experience a delay due to lags in wage and salary policy responses or adjustments to production processes of employers and/or time needed to clarify legal details. Therefore, continued evaluation of the medium to long-run effects will prove indispensable for understanding the full impact of the reform. Going forward, it remains to be seen whether the positive treatment effect will persist or even grow over time and whether the compliance gap will close, for instance due to stronger sanctions for non-compliance or to social pressure. More compliance, on the other hand, could induce stronger negative employment effects, which would carry further repercussions for the wage distribution. Moreover, it is likely that the relatively favorable business cycle that accompanied the introduction prevented a larger negative employment reaction. This situation may change if faced with a future recession. Finally, substitution effects in the medium run cannot be ruled out. It is possible that firms begin to favor workers exempted from the minimum wage or that they alter their production processes to outsource work packages abroad or to the self-employed in order to cut costs. All of these adjustments could influence the long-run income distribution in Germany.
Future research could furthermore consider whether the increase in the initial minimum wage level to
References
Aretz, M., T. Gregory (2013), The Minimum Wage Affects Them All: Evidence on Employment Spillovers in the Roofing Sector. German Economic Review 14 (S): 282–315.10.1111/geer.12012Search in Google Scholar
Ashenfelter, O., R.S. Smith (1979), Compliance with the Minimum Wage Law. Journal of Political Economy 87 (2): 333–350.10.1086/260759Search in Google Scholar
Autor, D.H., L.F. Katz, M.S. Kearney (2008), Trends in U.S. Wage Inequality: Revising the Revisionists. The Review of Economics and Statistics 90 (2): 300–323.10.1162/rest.90.2.300Search in Google Scholar
Autor, D.H., A. Manning, C.L. Smith (2016), The Contribution of the Minimum Wage to us Wage Inequality Over Three Decades: A Reassessment. American Economic Journal: Applied Economics 8 (1): 58–99.10.1257/app.20140073Search in Google Scholar
Bellmann, L., M. Bossler, S. Dummert, E. Ostmeier (2017), Mindestlohn: Längsschnittstudie für sächsische Betriebe. IAB Forschungsbericht Nr. 7/2017.Search in Google Scholar
Blundell, R., M. C. Dias (2009), Alternative Approaches to Evaluation in Empirical Microeconomics. Journal of Human Resources 44 (3): 565–640.10.1353/jhr.2009.0009Search in Google Scholar
Bound, J., C. Brown, N. Mathiowetz (2001), Measurement Error in Survey Data. Handbook of Econometrics 5: 3705–3843.10.1016/S1573-4412(01)05012-7Search in Google Scholar
Burauel, P., M. Caliendo, A. Fedorets, M. Grabka, C. Schröder, J. Schupp, L. Wittbrodt (2017), Mindestlohn noch längst nicht für Alle – Zur Entlohnung anspruchsberechtigter Erwerbstätiger vor und nach der Mindestlohnreform aus der Perspektive Beschäftigter. DIW Wochenbericht 89 (49): 1109–1123.Search in Google Scholar
Burauel, P., M. Caliendo, M. Grabka, C. Obst, M. Preuss, C. Schröder (2018), Auswirkungen des gesetzlichen Mindestlohns auf die Lohnstruktur. Studie im Auftrag der Mindestlohnkommission, Projektvergabe durch die Bundesanstalt für Arbeitsschutz und Arbeitsmedizin (BAuA), Vergabe-Nr. 525549, Abschlussbericht.Search in Google Scholar
Burauel, P., M. Caliendo, M.M. Grabka, C. Obst, M. Preuss, C. Schröder (2019), The Impact of the Minimum Wage on Working Hours. mimeo.10.1515/jbnst-2018-0081Search in Google Scholar
Butcher, T., R. Dickens, A. Manning (2012), Minimum Wages and Wage Inequality: Some Theory and an Application to the UK. University of Sussex Economics Department Working Paper Series Nr. 45/2012.Search in Google Scholar
Caliendo, M., A. Fedorets, M. Preuss, C. Schröder, and L. Wittbrodt (2017), The Short-Term Distributional Effect of the German Minimum Wage Reform. IZA discussion Paper 11246.10.2139/ssrn.3105410Search in Google Scholar
Caliendo, M., A. Fedorets, M. Preuss, C. Schröder, L. Wittbrodt (2018), The Short-Run Employment Effect of the German Minimum Wage Reform. Labour Economics 53: 46–62.10.1016/j.labeco.2018.07.002Search in Google Scholar
Caliendo, M., W.-S. Lee, R. Mahlstedt (2017), The Gender Wage Gap and the Role of Reservation Wages: New Evidence for Unemployed Workers. Journal of Economic Behavior & Organization 136 (April): 161–173.10.1016/j.jebo.2017.02.011Search in Google Scholar
Caliendo, M., C. Schröder, L. Wittbrodt (2019), The Causal Effects of the Minimum Wage Introduction in Germany: An Overview. German Economic Review, forthcoming.10.1111/geer.12191Search in Google Scholar
Cortes, K.E. (2004), Wage Effects on Immigrants from an Increase in the Minimum Wage Rate: An Analysis by Immigrant Industry Concentration. IZA Discussion Paper No. 1064.10.2139/ssrn.524606Search in Google Scholar
Cowell, F.A. (2011), Measuring Inequality. Oxford: Oxford University Press.10.1093/acprof:osobl/9780199594030.001.0001Search in Google Scholar
Dickens, R., A. Manning (2004), Has the National Minimum Wage Reduced UK Wage Inequality? Journal of the Royal Statistical Society: Series A 167: 613–626.10.1111/j.1467-985X.2004.aeI2.xSearch in Google Scholar
DiNardo, J., N. M. Fortin, and T. Lemieux (1996), Labor Market Institutions and the Distribution of Wages, 1973–1992: A Semiparametric Approach. Econometrica 64 (5): 1001–1044.10.3386/w5093Search in Google Scholar
Dolton, P., C. R. Bondibene, J. Wadsworth (2012), Employment, Inequality and the UK National Minimum Wage over the Medium–Term. Oxford Bulletin of Economics and Statistics 74 (1): 78–106.10.1111/j.1468-0084.2011.00653.xSearch in Google Scholar
Fitzenberger, B., A. Doerr (2016), Konzeptionelle Lehren aus der ersten Evaluationsrunde der Brancenmindestloehne in Deutschland. Journal for Labour Market Research 49 (4): 329–347.10.1007/s12651-016-0214-0Search in Google Scholar
Garloff, A. (2016), Side Effects of the New German Minimum Wage on (un-)employment: First Evidence from Regional Data. IAB Discussion Paper No. 31/2016.Search in Google Scholar
Goebel, J., M.M. Grabka, S. Liebig, M. Kroh, D. Richter, C. Schröder, J. Schupp (2018), The German Socio-Economic Panel (SOEP). Journal of Economics and Statistics.10.1515/jbnst-2018-0022Search in Google Scholar
Lee, D.S. (1999), Wage Inequality in the United States During the 1980s: Rising Dispersion or Falling Minimum Wage? The Quarterly Journal of Economics 114 (3): 977–1023.10.1162/003355399556197Search in Google Scholar
Low Pay Commission (2015), National Minimum Wage. Low Pay Commission Report, September.Search in Google Scholar
Low Pay Commission (2016), National Minimum Wage. Low Pay Commission Report, Spring.Search in Google Scholar
Manning, A. (2013), Minimum Wages: A View from the UK. Perspektiven der Wirtschaftspolitik 14 (1–2): 57–66.10.1111/pers.12004Search in Google Scholar
Metcalf, D. (2008), Why Has the British National Minimum Wage Had Little or no Impact on Employment? Journal of Industrial Relations 50 (3): 489–512.10.1177/0022185608090003Search in Google Scholar
Mindestlohnkommission (2016), Erster Bericht zu den Auswirkungen des gesetzlichen Mindestlohnes, Bericht der Mindestlohnkommission an die Bundesregierung nach § 9 Abs. 4 Mindestlohngesetz.Search in Google Scholar
Mindestlohnkommission (2018), Zweiter Bericht zu den Auswirkungen des gesetzlichen Mindestlohnes, Bericht der Mindestlohnkommission an die Bundesregierung nach § 9 Abs. 4 Mindestlohngesetz.Search in Google Scholar
Neumark, D., M. Schweitzer, W. Wascher (2004), Minimum Wage Effects throughout the Wage Distribution. Journal of Human Resources 39 (2): 425–450.10.3368/jhr.XXXIX.2.425Search in Google Scholar
Pusch, T., H. Seifert (2017), Mindestlohngesetz. Für viele Minijobber weiterhin nur Minilöhne. WSI Policy Brief Nr. 9/2017.Search in Google Scholar
Schmitz, S. (2017), The Effects of Germany’s New Minimum Wage on Employment and Welfare Dependency.Search in Google Scholar
Stegmaier, J., S. Gundert, K. Tesching, S. Theuer (2015), Bezahlter Urlaub und Lohnfortzahlung im Krankheitsfall: In der Praxis besteht Nachholbedarf bei Minijobbern. IAB Kurzbericht 18.Search in Google Scholar
Weil, D. (2005), Public Enforcement/Private Monitoring: Evaluating a New Approach to Regulating the Minimum Wage. Industrial Labor Relations Review 58 (2), 238–257.10.1177/001979390505800204Search in Google Scholar
Article note
This paper is based on the research report “Auswirkungen des gesetzlichen Mindestlohns auf die Lohnstruktur” by Burauel et al. (2018) which was delivered to the German Minimum Wage Commission in January 2018. The paper was prepared for the special issue on ``Effects of the Introduction of the Statutory Minimum Wage in Germany“ in the Journal of Economics and Statistics. Access to further articles of this special issue can be obtained at www.degruyter.com/journals/jbnst.
© 2020 Burauel et al, published by De Gruyter
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.