Introduction
Over the past several decades, numerous countries have implemented minimum wage policies to ensure an adequate standard of living for their citizens. For example, as of now, 22 out of the 27 European Union member states have established some form of statutory minimum wage. Consequently, researchers have extensively investigated the impact of minimum wages on diverse labor market outcomes, such as hourly wages (Dickens & Manning,
2004), total income (Neumark & Wascher,
2007), and employment (Zavodny,
2000). In light of these findings, recent research has begun to explore additional domains potentially influenced by a minimum wage. One emergent strand of literature has examined the effects on physical and mental health. However, the evidence thus far has been inconclusive, with studies reporting positive (Hafner & Lochner,
2022), negative (Horn et al.,
2017), and non-significant effects (Maxwell et al.,
2022). As a consequence, researchers call for a better understanding of the mechanisms through which minimum wage policies might affect individual health outcomes (Leigh et al.,
2019).
One potential mechanism through which minimum wage policies might influence health outcomes is physical activity. The health benefits of physical activity are well-documented (e.g., Warburton et al.,
2006), and prior research has demonstrated that socioeconomic disparities at both individual and state levels contribute to inequalities in physical activity participation within populations (Cerin & Leslie,
2008; Pabayo et al.,
2018; Petersen et al.,
2010). Specifically, individuals with lower incomes, particularly women, have been found to engage in less physical activity (Humphreys & Ruseski,
2015; Kari et al.,
2015). Accordingly, with governments worldwide looking for policy interventions to promote health through physical activity participation, understanding the effect of reducing income inequality by implementing a minimum wage may provide valuable insights for policymakers.
The existing literature on this topic is, however, relatively limited and has primarily focused on incremental increases in minimum wages in the United States (Horn et al.,
2017; Lenhart,
2019).
This study aims to shed light on this relationship by utilizing the quasi-experimental design of Germany’s introduction of the minimum wage in 2015. Therefore, data from the German Socio-Economic Panel for the years 2013–2017 is used, and a difference-in-difference (DiD) estimator with matching between control and treatment groups is employed to identify the causal effect of the minimum wage on physical activity participation.
Studies examining increases in minimum wages have reported positive effects on self-reported health (Andreyeva & Ukert,
2018), the prevalence of various health conditions (Lenhart,
2017), and mental health (Lebihan,
2023) and mental distress and depression (Bai & Veall,
2023; Kuroki,
2021). Horn et al. (
2017) found a significant negative effect on self-reported health but a positive effect on mental health. Maxwell et al. (
2022) observed no significant effect. Regarding the specific introduction of a minimum wage, Kronenberg et al. (
2017) found no significant effect on mental health in the UK, while Hafner and Lochner (
2022) reported a small positive effect on self-rated health in Germany. In their review, Leigh et al. (
2019) argued that further research is needed to explore potential pathways between minimum wage and individual health outcomes.
From a theoretical perspective, the relationship between minimum wage and physical activity participation can be described by applying the Grossman model (
1972). This economic model on demand for health and healthcare assumes that every individual has a durable capital stock of health which depreciates over time at a certain rate and by investing in either market goods (e.g., health care) or non-market goods (e.g., physical activity), individuals can restore their health. According to the Grossman model, individuals allocate their time between market and non-market activities, aiming to maximize their utility, which is influenced by health status, consumption of goods, and leisure. The model suggests that individuals will invest in health-enhancing behaviors if the marginal benefits, such as improved health, longevity, and higher productivity, outweigh the marginal costs, including time, effort, and resources required for these investments.
An increase in the minimum wage can impact an individual’s investment in health-enhancing behavior through the channels of income and time costs. When health is considered a normal good, an increase in the minimum wage should lead to individuals increasing their health inputs which would improve the individual’s health status. However, it has to be considered that the consumption of unhealthy goods (e.g., alcohol, cigarettes) might increase as well (e.g., Huang et al.,
2021). On the other hand, an increase in the hourly wage also increases the opportunity costs of leisure time, potentially leading individuals to substitute physical activity with work or other leisure activities and as a consequence, reduce the time allocated to physical activity. However, potential adverse effects on employment might result in reduced working hours, allowing for more time to engage in the consumption of non-market goods (Burauel et al.,
2020).
Existing research has identified both higher income and time costs as important determinants of participation in physical activity. Humphreys and Ruseski (
2011) found that a higher income has a significant positive effect on participation in physical activity. This finding is in line with numerous studies which were able to identify this effect consistently (e.g., Downward et al.,
2014; Kari et al.,
2015). However, regarding the increase in time costs, Humphreys and Ruseski (
2011) also showed that higher levels of income had a significant negative relationship with the duration of participation, likely due to higher opportunity costs. In another study, Humphreys and Ruseski (
2015) revealed that this two-fold effect is likely depending on the type of physical activity. For example, a higher income is positively associated with participation in swimming or golfing, whereas a negative effect was found for walking or exercising at home.
The relationship between working hours and participation in physical activity is complex, as the time and energy individuals can allocate to leisure activities are influenced by their work schedule and the demands of their job (Kirk & Rhodes,
2011). Numerous studies have found that longer working hours can negatively impact physical activity levels, as individuals may have limited time and energy available for exercise (Fransson et al.,
2012). Specifically, employees working extended hours or in jobs with high physical or mental demands may be more likely to experience fatigue and time constraints, limiting their ability to engage in regular physical activity (Kirk & Rhodes,
2011; Schneider & Becker,
2005).
Research examining the impact of minimum wage on physical activity is limited. Horn et al. (
2017) explored this effect as part of a supplementary analysis, utilizing a broad measure of physical activity (exercise participation in the last 30 days, yes/no). Their findings suggest that women are more likely to engage in physical activity following a minimum wage increase whereas for men, no effect was found. Lenhart (
2019) investigated the association between minimum wage increases and the time individuals allocate to physical activity participation. Employing a DiD approach, the study reveals that a $1 increase in the minimum wage leads to a 20-minute reduction in weekly exercise time. The author posits that this negative relationship may arise from an increase in time dedicated to other leisure pursuits.
This study aims to contribute to this small body of research in several ways. First, as highlighted by Caliendo et al. (
2019), the introduction of the minimum wage in Germany presents a compelling case due to the relatively high-income floor compared to other countries, consequently affecting an unusually large proportion of the population. In addition, this case provides an opportunity to explore novel insights into the non-labor market outcomes associated with the introduction of a minimum wage compared to raises of minimum wage levels. Second, by employing a regression-adjusted DiD model with matching, this study follows Leigh et al.‘s (
2019) recommendations for addressing the methodological challenge of ensuring comparability between treatment and control groups. Lastly, by utilizing an ordinal measure of participation frequency, this research contributes new evidence, as prior studies have primarily employed binary measures or count measures of duration.
Research context
After extensive debates and negotiations among policymakers, trade unions, and employer associations, on January 1, 2015, the German government introduced the first statutory uniform minimum wage in the history of the country. The initial minimum wage was set at €8.50 per hour, applicable to all adult employees across the country, regardless of the sector or region. Exempt from the minimum wage policy were minors, interns, and apprentices. Since the implemented floor was relatively high in comparison to other countries, overall, 4.0 million employees, which represent between 10 and 14% of the total workforce, were affected by the regulation (Lesch & Schröder,
2016). Throughout the years, the German government has progressively elevated the minimum wage. In 2017, it was increased to €8.84, followed by subsequent adjustments to €9.19 in 2019, and €9.60 in 2021. On October 1, 2022, the recently elected government enacted the most substantial increase to date, raising the minimum wage to €12.00.
The introduction of the minimum wage in Germany has generated significant interest among researchers, who have examined various aspects of its impact on the labor market and economy. In terms of personal income, the evidence suggests a notable increase in the hourly wage by €0.50 per hour (Burauel et al.,
2020). However, Caliendo et al. (
2018) demonstrated that this positive effect did not result in increased gross monthly earnings. This can primarily be attributed to significant reductions in contractual hours, as reported by studies from Caliendo et al. (
2018) and Pusch and Rehm (
2017). Additionally, research has identified no significant impact on poverty risk (Bruckmeier & Bruttel
2021) and only marginal short-term negative employment effects (Caliendo et al.,
2018).
Methods
Data sources
The analysis of the relationship between the introduction of a minimum wage and physical activity is based on the German Socio-Economic Panel (GSOEP) (GSOEP,
2019). The GSOEP is a German household panel survey conducted annually by the German Institute of Economic Research since 1984. In previous research, the survey data has already been utilized to examine determinants (Breuer & Wicker,
2008) and outcomes of physical activity (Lechner,
2009). With regard to the introduction of a minimum wage, the GSOEP has been used to investigate numerous employment effects (Caliendo et al.,
2018) and various well-being outcomes (Gülal & Ayaita,
2020).
The present study focuses on the immediate effect of the implementation of the minimum wage in 2015. Since the information on the participation frequency in physical activity is not available in every wave, it compares the pre-treatment period in 2013 with the post-treatment period in 2015 and 2017.
Given that the focus of the analysis is on the minimum wage, only individuals who are working full-time or part-time are included. Self-employed individuals, interns, and apprentices are not included since the minimum wage does not apply to them. Marginally employed workers were also excluded from the analysis, as comparing their working arrangements to standard contracts presents challenges. Since information on the hourly wage was also available in 2014, observations that showed a change in treatment status compared to 2013 were not considered. The final sample size of the study consists of n = 2,258 respondents before the introduction of the minimum wage in 2013 and n = 2,448 after the introduction in 2015 and 2017.
Measures and variables
The outcomes measuring participation frequency in physical activity are assessed as follows: Participation frequency was measured by a four-point scale with the categories at least once a week, at least once a month, less often, or never (
PA frequency). The four categories are mutually exclusive.
1 According to the recommendations by the WHO (
2010), a certain frequency of participation in physical activity of at least once a week is needed to yield the aimed health benefits. Hence, a dummy variable for the category of participation in physical activity at least once a week (
PA weekly) was used as a binary dependent variable, with 1 indicating weekly participation.
To determine an accurate hourly wage, the monthly gross income—excluding bonus payments—is divided by the number of working hours. Since working hours were provided on a weekly basis, they were multiplied by 4.33. This number is used because there are, on average, 4.33 weeks in a month (accounting for both 30 and 31-day months). The approach is consistent with the recommendation by Dütsch et al. (
2019). The
treatment status variable indicates whether someone was affected by the minimum wage (= 1) or not (= 0). The treatment group comprises individuals who earned an hourly wage below €8.50 prior to the introduction of the new minimum wage in 2013. The control group consists of individuals who earned an hourly wage greater than €8.50 but no more than €12.75 (50% higher). This upper limit enables the formation of socio-demographically similar control and treatment groups. This approach has been previously employed in similar contexts by Gülal and Ayaita (
2020) and Reeves et al. (
2017). Additionally, the analysis includes two time-specific dummy variables for the years 2015 and 2017 to identify the post-minimum wage periods. These variables are assigned a value of 1 for the corresponding post-minimum wage years and 0 for the baseline year of 2013.
Furthermore, the dataset offers comprehensive information on socio-demographic and job-related characteristics (Table
1). The following socio-demographic control variables that influence the frequency of physical activity are included:
Age and
Age2 to account for a non-linear relationship, a binary indicator for health problems (
Bad health), presence of children (
Children), five distinct dummy variables representing various educational levels (
No degree,
Main school,
Secondary school,
Field-specific, and
A-level), an indicator for a university degree (
University degree), household size (
HH size), gender (
Male), and marital status (
Married).
Table 1
Overview of conditional variables
Dependent variables | | |
PA frequency | Participation frequency in physical activity (4-point scale) | Ordinal |
PA weekly | Participation in physical activity at least once a week (1 = yes) | Binary |
Socio-demographic | | |
Age | Age (in years) | Metric |
Age2 | Age squared | Metric |
Bad health | Self-reported health status (0 = other; 1 = bad health) | Binary |
Children | Presence of a child (1 = yes) | Binary |
Education | | |
No degree | No degree (yes = 1) | Binary |
Main school | Main school degree (yes = 1) | Binary |
Secondary school | Secondary school degree (yes = 1) | Binary |
Field-specific | Field-specific degree (yes = 1) | Binary |
A-level | A-level (yes = 1) | Binary |
University degree | University degree (yes = 1) | Binary |
HH size | Number of persons living in the household | Metric |
Male | Gender (0 = female; 1 = male) | Binary |
Married | Marital status (0 = other; 1 = married) | Binary |
Job-related | | |
Tenure | Tenure (in years) | Metric |
Company size | | |
< 10 | Number of employees < 10 (yes = 1) | Binary |
10–100 | Number of employees 10–100 (yes = 1) | Binary |
101–2000 | Number of employees 101–2000 (yes = 1) | Binary |
> 2000 | Number of employees > 2000 (yes = 1) | Binary |
Education fit | Fit between education and match (yes = 1) | Binary |
Job change | Job change in the past (yes = 1) | Binary |
Job status | | |
Blue-collar | Blue-collar job (yes = 1) | Binary |
White-collar | White-collar job (yes = 1) | Binary |
Part-time | Part-time job (yes = 1) | Binary |
Temporal job | Temporal job (yes = 1) | Binary |
For job-related controls, the study incorporates the tenure with the employer (Tenure), the size of the employer based on the number of employees (< 10; 10–100; 101–2000; >2000), the compatibility of the job with the individual’s education (Education fit), whether there was a job change in the previous year (Job change), and whether the job is part-time (Part-time) or temporary (Temporal job). Information on the type of job (Blue-collar and White-collar) and based on the NACE branch codes, seventeen dummy variables controlling for the industry are considered. In addition, every model includes state fixed effects to control for regional differences.
Empirical analysis
The study applies a DiD approach to estimate the causal effect of the introduction of the minimum wage in 2015 on physical activity in Germany. The DiD captures how participation in physical activity in the treatment group changes in comparison to the control group. The following notation gives the econometric model:
$$\begin{aligned} {Y}_{it} & = {\beta }_{0}+{\beta }_{1}{Treatment}_{i}+ {\beta }_{2}{Post2015}_{t}+{\beta }_{3}{Post2017}_{t}+{\beta }_{4}{Treatment}_{i} \\ & \quad *Post{2015}_{t} +{\beta }_{5}{Treatment}_{i}*Post{2017}_{t}+{\beta }_{6}{X}_{it}+{\beta }_{7}{Z}_{it}+{\theta }_{i}+{\varepsilon }_{it}\end{aligned}$$
(1)
\({Y}_{it}\) describes the physical activity outcome variable varying by individual i and by time t. The treatment effect on the treatment group (ATT) is identified by the interactions between the treatment group indicator variable \({Treatment}_{i}\)and the variables reflecting the post-minimum wage years \({Post2015}_{t}\) and \({Post2017}_{t}\). To improve the precision of the model the vector \({\beta }_{4}{X}_{it}\) captures the effect of the time-variant socio-demographic control variables and \({\beta }_{5}{Z}_{it}\) the effect of job-related control variables. Individual fixed effects are included with \({\theta }_{i}\).
To ensure comparability between the control and treatment groups based on their observable characteristics, a regression-adjusted DiD matching strategy was employed, following the method proposed by Heckman et al. (
1997). In the first step, a probit model estimates propensity scores, representing the probability of being affected by the minimum wage, conditional on the observable characteristics. These include the aforementioned socio-demographic and job-related control variables. Next, kernel matching with the epanechnikov kernel function and with bandwidth of 0.06 was conducted. This method assigns weights to individuals based on their propensity scores and has been widely adopted in previous studies (Hafner & Lochner,
2022; Marcus,
2014). The results were robust for different specifications of the matching procedure such as a logit model in the first stage, different kernel functions, and bandwidths. The results are available upon request.
When analyzing both outcomes—frequency of physical activity (PA) and weekly PA participation—we employed individual fixed effects models. Lechner et al. (
2016) argue that incorporating individual fixed effects enhances the precision of DiD models, particularly when time-varying panel non-response might affect the assumption of parallel trends. Given the fluctuation in our sample size across the time periods, the adoption of individual fixed effects models is posited to yield more consistent estimations. Both outcomes yielded similar results in (ordered) logit models for the non-matching scenarios. All models are estimated with robust standard errors.
To identify the causal effect of the minimum wage on the frequency of physical activity participation, the assumption is made that the difference in the change in physical activity behavior between the treatment and control groups is only attributable to the introduction of the minimum wage. Consequently, the physical activity outcomes for both groups would have exhibited similar changes over time had the minimum wage not been implemented (common trend assumption). The inclusion of numerous time-variant socio-demographic, job-related control variables and individual fixed effects should increase the likelihood of the common trend assumption holding true. Additionally, an event study model spanning the years 2009–2017 is estimated. In an event study, lag and lead variables related to the event of interest (i.e., the introduction of the minimum wage) are incorporated, allowing for the examination of any significant pre-trends in terms of differences in physical activity participation between the control and treatment groups. In our model, the years 2009 and 2011 are included as pre-trend years.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.