The effect of minimum wages
on the employment rate of
low-educated workers
Bachelor Thesis Name: Lars Dedding Studentnumber: 10214925 Supervisor: A. Kamm MSc Date: 26-06-2015
1
Abstract
In this study, the effects of the minimum wage on employment are studied. The effects on low-educated workers are of particular interest, however the effect on the total labor force is also investigated. To study this effect, there is made use of a longitudinal dataset consisting of employment rates, real minimum wages and several control variables of 51 U.S. states
ranging from 2005 to 2013. The results show that the real minimum wage had a significant negative impact on the employment rate for the low-educated workers as well as for the entire labor force. A ten percent increase of the minimum wage is predicted to lead to a 0.59 percent decrease of employment of lower educated workers. For the total labor force the model predicts a 0.29 percent decrease. From this one can conclude that the minimum wage has a relatively small yet significant effect on employment. The effect on low-educated workers is larger than the effect on the total labor force.
2
Table of contents
List of tables and figures ... 3
Introduction ... 4 Literature review ... 5 Theory ... 5 Empirical Results ... 7 Methodology ... 10 Variables ... 10 Econometric Model ... 13 Results ... 14 Discussion ... 15 Conclusion ... 17 Recommendations ... 18 Reference List ... 19 Appendices ... 22
3
List of tables and figures
Description Page number
Table 1. Random effects panel data models results 14 Figure 1. Simple supply and demand model with imposed minimum wage wm. 5
Figure 2. Standard monopsony model with imposed minimum wage w1 7
Figure 3. Annual U.S. Inflation Rate from 2005 to 2013 11 Figure 4. Terms of trade ratio from 2005 to 2013 12 Figure 5. Annual long-term annual interest rate 12
4
Introduction
“When you raise the price of employment, guess what happens? You get less of it[…]”. That is the response John Boehner gave, on 13 February 2013, to president Barrack Obama. He called for raising the federal minimum wage from $7.25 to $9.00 on his State of the Union the night before (Thinkprogress, 2013). According to president Obama, the current federal
minimum wage is not sufficient to keep families out of poverty and should therefore be raised to a socially acceptable $9.00. The Democrats pointed out that raising the minimum wage would lead to families earning more money and thus would boost the economy by raising the aggregate expenditure (Thinkprogress, 2013). On the other hand, Republicans argued that by raising the minimum wage, labor gets more expensive which would lead to less demand for labor so that unemployment would increase (Thinkprogress, 2013). Boehner also pointed out “A lot of people who are paid the minimum wage, are being paid that because they come to the workforce with no skills.” (Wonkette, 2013). This research tries to aid the economic literature by investigating the effects of the minimum wage on just that group, the low-skilled workforce. This is accomplished by targeting the workforce with a relatively low level of education.
Research into the effects of the minimum wage on employment has been vastly popular under economists as summarized in a recent study from the Center for Economic and Policy
Research (2013). A true consensus on the impact has never been achieved, nor in the political world, nor in the economic empirical literature (Center for Economic and Policy Research, 2013). Studies into subgroups of the labor force, mainly young workers, have presented results that verify that there is an effect on the employment of these groups from changes in the minimum wage such as the study from Sen, Rybczynski and Van de Waal (2011). Thus far, little research has gone into verifying the effect on the subgroup of low-educated workers while they are more likely to work for minimum wage than workers with higher attained education as found by the U.S. Bureau of Labor Statistics (2014). The goal of this research is to aid the available information on minimum wage effects on employment, by expanding the amount of researched subgroups of the labor force.
The goal of this research is investigating whether the minimum wage has a significant empirical effect on the employment of the low-educated workforce. This is achieved by formulating a longitudinal econometric model from 50 US states plus the District of Columbia from 2005 until 2013. A regression will be performed to assess the effect of the minimum wage per state on the low-educated workers in that state, besides several control variables. This same regression is also performed on the total workforce, to simplify a comparison between the effect of the minimum wage on the total workforce versus the low-educated workforce.
This research takes the following form. It will commence with a literature review, that includes several theories and a summary of empirical results. It will then proceed with a methodology in which the implemented variables and the way of research is clarified. Here after, the results will be presented, accompanied by a discussion and a conclusion. Finally, recommendations for further research are given.
5
Literature review
As pointed out in the introduction, the effects of the minimum wage on employment have been of great interest for economists ever since the introduction of the Fair Labor Standards Act in 1938 (U.S. Department of Labor). Over the years, several hundreds of researches have been done within this field of labor economics. Summarizing these researches, we can pick out several classes that differ in their research methods as well as in their results (Center for Economic and Policy Research (2013). This literature review will commence with a
presentation of available theories and will then present a chronological summary of empirical results with their respective methodologies.
I Theory
First the simple supply-demand model is described, for which several assumptions must apply. There needs to be a perfectly competitive market which contains a large number of
homogeneous workers as well as a large number of homogeneous employers. Furthermore, the assumption is made that there is perfect information and free entry and exit out of the market (Econport, n.d.). As shown in figure 1 this model contains of a supply and demand model for labor, with wages on the vertical axis and employment on the horizontal axis. If a minimum wage were to be selected at a level above equilibrium, this would lead to a higher supply of labor and a smaller demand of labor. This would effectively decrease employment by E0- Em.
Figure 1. Simple supply and demand model with imposed minimum wage wm.
Brown, Gilroy and Kohen (1982) note that the excess supply of labor that is shown in figure 1 as Sm - Em does not have to equal the unemployment rate. The increase of supply combined
with the decrease of demand might lead to the withdrawing of workers from the labor market These people, who are not actively searching for work are not incorporated in the official
6
unemployment rate (Brown et al. 1982). Here fore it is superior to use the employment level in statistical research instead of the unemployment level. Another possibility that Brown et al. (1982) mention is that the effect of the minimum wage does not have to be observed as reduction in employment. Instead, it could also lead to a decline in employment growth or, if employment does drop, the effect may be lagged since employers might choose to stop hiring new personnel as an alternative for laying off their current workers.
Besides the simple supply-demand model, there are other theories that look into the effects of a minimum wage on employment, for instance the “shock effects” theory (Brown et al. 1982). This theory assumes that producers do not fully minimize costs. A raise of the minimum wage (which acts as a shock) can lead them to do so, because by raising the productivity they compensate for the more expensive labor. This might reduce the negative employment effects, since the same amount of workers (who are then paid a higher wage) could be used, for the same total costs. Problematic with this theory is that it might be possible that firms were not working as productive because they were using relatively too much labor. Efforts to raise productivity, created by the shock, would then lead to lower productivity. Stigler (1946) criticizes the shock theory with the aid of two arguments. First, he points out that
implementing the shock theory might be principally unfitting for low wage industries. He reviewed the large manufacturing industries which paid low wages in 1939. Overviewing these industries he noted that they were relatively competitive and also had a relatively high ratio of wages to total-processing-cost-plus-profit (Stigler, 1946). From this, we can
determine that the low wage paying industries were already working at highly productive levels, so that a higher minimum wage could not lead to a shock that would boost productivity. A third theory that Brown et al. (1982) discuss is the two-sector model. The two sectors
contained within this model are the sector that is covered by the minimum wage law and the sector that is not covered by the minimum wage law. In 1982, at the time this paper was written, coverage of the minimum wage law was not as abundant as it is nowadays. After the Fair Labor Standards Act was introduced in 1938, adaptions have been made as early as 1947 (U.S. Department of Labor, n.d.). Ever since, coverage kept increasing, since smaller
companies were also obliged to pay minimum wages. There is still no full coverage in the United States labor market, since some states still have protective measures for small enterprises and also because of the sub-minimum wage that is paid to young workers. The reasoning behind the employment effects of the two sector model is as follows (Brown et al. 1982). The raise of the minimum wage in the covered sector will decrease the demand of labor in that market. This decrease in demand will lead to lower employment within this sector, so the unemployed from the covered sector will switch to the uncovered sector, increasing the supply of labor. This will result in lower wages and thus higher employment in the uncovered sector. Not everybody who is laid off in the covered sector is willing to work in the uncovered sector for a lower wage. The overall effect on employment therefore will be dependent on the elasticity of labor demand, the size of the covered sector and the reservation wages (the lowest wages people are willing to work for) (Brown et al. 1982).
Related to the two sector model, is the two skills model which incorporates heterogeneous workers (Brown et al. 1982). In this model, there is accounted for specific groups of workers instead of assuming all workers are equal in skills for instance. By checking specific groups of workers which are more likely to work for minimum wage, for instance young workers, effects of minimum wage workers are more likely to be significant. On other words, the elasticity of employment to the set minimum wage is likely to be lower for the entire population of workers than for a subgroup of workers who are likely to work for minimum
7
wage. This method of reasoning is also implemented in the research question of this study by checking for the subgroup of the population with a relative low education.
The last model that will be discussed is the monopsony model, which includes a market in which only one party demands a good and several parties supply it (Maurice, 1974). In this case, the relevant good is labor and the role of monopsonist is played by an employer. Without a minimum wage, the employer will hire labor until marginal cost equals labor. He will do so since its marginal cost of labor always exceeds the supply price (Brown et al. 1982), as is shown in figure 2. The imposing of a minimum wage makes the employer a price taker, and if chosen correctly, it will lead to higher employment. Raising the minimum wage above the equilibrium wage w1 will lead to a decrease in employment from E1.
Figure 2. Standard monopsony model with imposed minimum wage w1
II Empirical results
As is clear from this summary of existing theories, there is no consensus on the magnitude or the direction of the effect of a minimum wage on employment. This is reflected in the results of relevant empirical studies which have been performed with data as early as from 1948 (Gramlich, Flanagan, & Wachter, 1976).
Brown et al. (1982) concluded from their study, that looks into numerous related empirical studies, that the effect of an increase of the minimum wage on employment of teens was significant. More precisely, they found elasticities ranging from -0.1 to -0.3, meaning that a 10 percent increase of the minimum wage would lead to a decline of employment ranging from 1 to 3 percent. They got to this result by balancing 24 separate studies into the
employment effects of the minimum wage on teenagers, ranging from 1970 to 1981. Common in these studies is the use of a time-series estimation model, in the form of:
8
Where, Y represents a measure of the employment rate. MW represents a measure of the minimum wage. D represents a business cycle variable to account for changes in the
wellbeing of the economy. X1 to Xn represent a number of other relevant explanatory factors
that might be helpful explaining the variance of the employment level. There was a wide variation among the dependent and independent variables of the contributing studies. Amid the choices of dependent variable were the labor force participation rate, the employment-population ratio and the unemployment rate. Among the choices of independent variable were the amount of persons in the armed forces, the school enrollment and the amount of
employment programs. Testing for the most plausible results, Brown et al. (1982) concluded that the lower estimates of elasticity were to be preferred. To be more precise, a ten percent increase of the minimum wage is more likely to lead to a decrease of employment of 0 to 0.75 percent. Besides the effects on teenage employment, effects on other age groups were also studied. The results for young adults were estimated to be smaller than for teens, but still negative. The effects of an increase in the minimum wage on employment for adults were ambiguous. This is also reflected in the theory, in which adult workers can become
unemployed since the minimum wage makes them too expensive for employers but there is also a possibility that they will become employed since they are no longer competing with teenagers (Brown et al. (1982).
Results from Brown et al. (1982) were held valid for a decade until a new form of minimum wage research arose. This research was based on natural experiments, in which single events were investigated in a setting similar to a laboratory experiment. The “new minimum wage research”, as it became known, looked at particular events, such as an increase of the minimum wage in a one separate state, and compared this to a state that did not increase its minimum wage. The research from David Card and Alan Krueger (1994) into the
employments effects of the New Jersey’s minimum wage increase from $4.25 to $5.05 per hour is perhaps the best know natural experiment within this field. The comparison that is made is in this case between New Jersey and Pennsylvania with regard to the employment of the fast food industry. For this, Card and Krueger surveyed 410 fast food restaurants in the concerning states before and after the rise in the minimum wage (1994). From this they did not find the hypothesized decline in employment but instead found that the increase of the minimum wage had led to an increase of 13 percent in employment. Card and Krueger concluded (1995, pp. 389-390), “The weight of this evidence suggests that it is very unlikely that the minimum wage has a large, negative employment effect.”
The uncommon results from Card and Krueger led to critiques from several economists (Center for Economic and Policy Research, 2013). David Neumark and William Wascher (2000) responded by investigating the same rise of the minimum wage in the same states by using payroll data instead of surveys. Their result was more in line with economic theory, they found a statistically significant decrease of employment in New Jersey of 4.6 percent. Neumark and Wascher (2000) preferred to make use of panel data analysis over natural experiments regarding testing the effects of the minimum wage on employment. Using panel data analysis, and specifically on U.S. states, offers several advantages over time-series analysis. Firstly, the federal minimum wage does not change a vast amount over time, which makes it more difficult to establish a statistical relation. Also, the changes that present in the minimum wage legislation are correlated with social welfare programs, which makes it even more complicated to isolate the minimum wage effects (Neumark and Wascher, 1992). The practice of using longitudinal data helps with isolating the minimum wage effect since it merely needs to explain the differences between the data over time. The second advantage
9
presented is that minimum wages often differ per state. This is for the reason that the federal minimum wage functions as an absolute minimum, states are allowed to raise their state minimum wage above the federal minimum wage which regularly happens. 29 states
including D.C. have minimum wages above the federal minimum wage at the time of writing (National Conference of State Legislatures, 2015). Using time-series analysis, this is
systematically not taken into account which can lead to errors in estimation. The third advantage mentioned is that states are allowed to legislate subminimum wages, to facilitate small businesses for example, which can also be taken into account when analyzing the data. Neumark and Wascher (1992) made use of panel data analysis by reexamining the data from the preceding studies. They provided more statistical evidence for relatively small negative effects of employment caused by the minimum wage with findings in the same range of elasticity as Brown et al. (1982). Elasticities for teenagers (aged 1619) ranged from 0.1 to -0.2, elasticities for young adults (16-24) ranged from -0.15 to -0.2.
The most recent relevant studies are either in the form of a natural experiment or in the form of a longitudinal analysis. Several meta-studies have been performed to examine the results of these studies. Among them is the meta-study Doucouliagos and Stanley (2009) performed containing over 1000 relevant studies. They found that the majority of the results clustered at zero or near zero effects of minimum wages on employment. From this they concluded that “Two scenarios are consistent with this empirical research record. First, minimum wages may simply have no effect on employment. Second, minimum-wage effects might exist, but they may be too difficult to detect and/or are very small.” (Doucouliagos and Stanley, 2009, p. 422)
10
Methodology
This section contains the econometric analysis of the research question. It will commence with an overview of the concerning variables with additional information on the dataset and the reasoning for the implemented variables. After this it will explain the econometric model used and the structure of the analysis.
I Variables
Herein the variables used to assess the effect of the minimum wage on the employment rate of low-educated workers are elucidated. The assembled dataset contains information on 51 states including the District of Columbia ranging from 2005 to 2013. This research looks into
separate U.S. states, for the same reasons David Neumark and William Wascher (1992) provided. Namely, the federal minimum wage does not change as often as state minimum wages do, furthermore most states have a minimum wage that differs from the federal
minimum wage. The time period over which this analysis has been performed is based on the available statistics of low-educated workers on a state level. Of the variables that were used, a distinction can be made between three classes: the dependent variables, the key dependent variable and the control variables. A descriptive summary of the chosen variables are given in the appendices, for a better overview of the variables.
In this research two dependent variables have been analyzed, the employment rate of lower educated workers and the employment rate of the total workforce. Doing so enables us to compare the effects of the two groups. To determine the employment rate of lower educated workers the dataset from the American Community Survey has been consulted. The ACS provided employment records per attained education group per state for the age of 25-64. Persons without a high school diploma were selected to define the ‘low-educated workers’. Data from the Bureau of Labor Statistics reported that this subgroup of workers were paid the minimum wage relatively the most, out of subgroups of attained education (2013). To define the employment rate of low-educated workers, the number of employed workers without a high school diploma were divided by the total number of persons without a high school diploma. To define the overall employment rate, the number of all employed workers was divided by the total number of persons.
The key dependent variable in this research is the real minimum wage. The United States Department of Labor supplied a historical overview of the basic minimum wage per state. From this overview, is was notable that there were relative large differences between states but also within states. The differences within states were explained by noticeably lower subminimum wages that were applied to protect relative small businesses. To attain a single value for the minimum wage for data analysis, the highest value was chosen, similar to the approach of Neumark and Wascher (1992). Doing so creates the assumption that all workers who get paid the minimum wage, get paid the highest available minimum wage in that state, which is not likely. However, statistics from the United States Census Bureau (2008) show that the multitude of workers do not work in a small business, which makes the assumption a reasonable proxy. To account for changes in price level, the real minimum wage is used instead of the nominal minimum wage. This is converted by the use of the Consumer Price Index method (U.S. Inflation Calculator, 2015).
In order to diminish omitted variable bias, control variables are added to the equation to account for other factors that influence the employment rate. Choudhry, Marelli and
11
Signorelli (2012) provided several causes of unemployment, under which macroeconomic conditions, demographic conditions and institutional variables. For the macroeconomic conditions gross domestic product growth, inflation, terms of trade and real long-term interest rates are given as causes that have a significant impact on unemployment. The structural conditions that are explained are primarily the degree of competitiveness and the degree of economic freedom. Choudry et al. (2012) focused on multiple countries for a longer time period. But in this short time period with data from only the United States, that had not seen large changes in these variables (Heritage.org, 2015), the choice is made not to include
structural conditions. For the institutional variables that can have an effect on the employment rate, unemployment benefits and labor taxes are given.
Macroeconomic conditions
The positive correlation of gross domestic product growth with unemployment is described in Okun’s law. This correlation has been examined by Lee (2000), and though he described the relationship as instable over time, he concluded that the link between growth and
unemployment is valid after all. In our observed time period, gross domestic product growth has seen relative severe changes, because of the financial crisis starting in 2008. The effects of a financial crisis on employment are studied upon by Bartolucci, Choudhry, Marelli and Signorelli (2011). He concluded that a financial crisis led to additional unemployment effects caused by increased uncertainty, besides the effects that one would consider by looking at the gross domestic product growth decrease on itself. To take this additional effect into account, a dummy variable is created that has the value ‘0’ up to 2008 and the value ‘1’ from thereon. This is based on the assumption that the worldwide financial crisis had long term employment effects. This assumption is validated by the height of the overall employment rates from 2008 to 2013 (Bureau of Labor Statistics, 2015) compared to the employment rate before the crisis. The trade-off of inflation and unemployment is described in the Philip’s curve. Herein excess demand on the goods market leads to inflation, and since excess demand would lead to more demand of labor to produce more goods, there must be a negative correlation between
inflation and unemployment (Santomore and Seater, 1978). The pattern of the inflation rate in the studied time period is presented in figure 3. This data show that the inflation rate varied a relative large amount and even experienced negative rates.
Figure 3. Annual U.S. Inflation Rate from 2005 to 2013 (U.S. Department of the Treasury).
-1.0% -0.5% 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% 4.5% 2005 2006 2007 2008 2009 2010 2011 2012 2013
12
A decrease in the terms of trade is stated to reduce national income and private savings, thereby generating a deficit on the current account, by the Laursen-Metzler effect (Laursen and Metzler, 1950). Feliz (1994) proceeds on this result by examining the effects of terms of trade on labor supply. His findings show that a decrease in the terms of trade leads to a decrease of labor supply. Figure 4 presents the terms of trade ratios.
Figure 4. Terms of trade ratio from 2005 to 2013 (OECD).
The effects of the real long-term interest rate on employment are described by the classical economic view. In increase in the real long-term interest rate will lower investments, decreasing aggregate demand. This will lead to a decline in the employment rate. Figure 5 illustrates the pattern of the real long-term interest rates.
Figure 5. Annual long-term annual interest rate (%) (World Bank).
90 91 92 93 94 95 96 97 98 99 100 101 2005 2006 2007 2008 2009 2010 2011 2012 2013
Terms of trade ratio
0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 2005 2006 2007 2008 2009 2010 2011 2012 2013
13
Institutional variables
The choice of institutional variables has been retained to the unemployment benefits. The theoretical reasoning for the effect of unemployment benefits is as follows, the higher the benefits, the less motivation one has to enter the labor market. Therefore, the higher the amount of unemployment benefits, the lower the employment will be (Choudhry et al. (2012). Other factors Choudhry et al. (2012) discuss such as the height of the labor are kept out of the equation. For the labor tax it proved to be too difficult to attain a single value for the
regression, because of the vast differences in taxation between states but also within states (Tax Foundation, 2013). The large amount of differences in personal exemptions, tax rates for singles and couples and so on, led to the choice of excluding the tax rate from the equation.
II Econometric model
For this analysis, an unbalanced random-effects panel data model is used, in order to fully make use of the available information. In order to make an easier comparison with previous research, in which results were provided in the form of elasticities, there is made use of log-log form. Since there are two dependent variables, we analyze two separate models. The model for the employment rate for low-educated workers is as follows:
lemplowit = βo + β1lminwageit + β2benefitit + β3gdpgrowthit (1)
+ β4inflationt + β5crisist + β6interestt + β7tott + εit
Herein lemplowit represents the logarithm of the employment rate for low-educated workers in
state i and time t. To alternatively investigate effects on the entire labor market, we make use of lemptotit as dependent variable, to represent the logarithm of the total employment rate.
This model is as follows:
lemptotit = βo + β1lrminwageit + β2benefitit + β3gdpgrowthit (2)
+ β4inflationt + β5crisist + β6interestt + β7tott + εit
To represent a constant in this model, we use βo. The key dependent variable, the logarithm of
the real minimum wage, is represented by lminwageit.. The control variables that vary per
state as well as per year, the average weekly unemployment benefit amount and the growth rate of the gross domestic product, are represented by benefitit and gdpgrowthit, respectively.
The control variables that vary over time but not per state, the level of inflation, the dummy variable that represents the financial crisis, the real interest rate and the terms of trade index level are represented by inflation, crisis, interest and tot respectively. The error term is represented by εit.
The choice for a random effects panel data model was made based on the significant outcome of a Breusch-Pagan Lagrange Multiplier test, which showed that analyzing the data with ordinary least squares was not in order. A random effects model was preferred over a fixed effects panel model based on the insignificant outcome of a Hausman-test. The test statistics with related significance are reported in the results table.
14
Results
The results of model (1) and (2) are shown below in table 1. Included within this table are the test statistics of the Wald Chi-Square test, the Breusch-Pagan Lagrange Multiplier test and the Hausman test.
Table 1: Random effects panel data models results
Models
Independent variables (1) Dependent variable =
lemplow (2) Dependent variable = lemptot Constant 5.943447** (0.9530771) 2.834356*** (0.3008707) Lminwage -0.0656113*** (0.0222007) -0.0321790*** (0.0070701) Benefit 0.0005219*** (0.0001833) 0.0001496** (0.0000614) Gdpgrowth -0.0934361 (0.0800326) 0.0045681 (0.0253154) Inflation -0.0531821*** (0.0102077)) -0.0266155*** (0.0032224) Crisis -0.3624473*** (0.0496708) -0.1499176*** (0.0156861) Interest -0.0316741*** (0.0064895) -0.009286*** (0.0020485) Tot -0.0629409*** (0.0090218) -0.0301104*** (0.002847) R2 (overall) 0.1516 0.1527 Number of observations 408 408
Wald Chi-Square test-statistic 309.30*** 645.44*** Breusch-Pagan Lagrange Multiplier test-statistic 974.70*** 1173.21*** Hausman test-statistic 2.47 7.49
1. Standard errors are reported in parentheses
15
Discussion
The results from our regressions show that the minimum wage has a significant negative effect on the employment rate of the low-educated workers as well as on the employment rate of the total workforce. As hypothesized, the effect on low-educated workers is considerably larger than on the total workforce. To make for an easier comparison with previous studies, the logarithms of the variables has been used. According to these statistics, an increase of the minimum wage with 10 percent would be accompanied with a decrease of 0.59 percent of the employment rate of low-educated workers. For the total workforce, this would be
accompanied with a decrease of 0.29 percent of the employment rate. When examining the 95-percent confidence interval of the models (for (1):0.1091239;-0.0220988], for (2): [-0.0460362;-0.0183218]), it is noticeable that these all these values are negative. Because of this we can state with 95 percent statistical certainty that the minimum wage has a significant negative effect on employment. The direction as well as the magnitude of the effect is in line with previous results from Brown et al. (1982) and Neumark and Wascher (1992).
This paper argues that the correlation that is found between the minimum wage variables and the employment rates of the low-educated workers as well as of the total labor force is not mere correlation but is also a reflection of a causal relationship. A regression is a useful tool for econometric research yet without theoretic foundation it is meaningless. In this study, the theoretical foundation is given in the form of multiple models which explain the analysis behind the effect of the minimum wage on the employment rate. It is however possible that a key determinant of the employment rate has been left out in this research. This would cause the established relationship to be mere correlation, and no causality. Nevertheless, numerous related researches on this topic have been performed in a similar manor. For example,
multiple regression models with the unemployment or employment rate as dependent variable and a measure of the minimum wage plus additional control variables as independent
variables have had many forms (Brown et al. (1982). A log-log regression model that uses the real minimum wage as independent variable, similar to this research, has been implemented by several studies such as Gramlich et al. (1976) and Adie (1973). Also, the particular dataset that has been studied, the separate employment rates of the states over a time-period of several years, has been investigated by Neumark and Wascher (1992). From the theoretical base combined with the relative large amount of similar studies with comparable
methodologies, one can conclude that there is a high possibility of a causal relationship between the minimum wage and the employment rate of low-educated workers as well as the total workforce.
When the control variables are studied, it is notable that some of these variables have a coefficient that indicates an effect that is opposite of what theory suggests, even though they are significant. For instance, the weekly average unemployment benefit amount variable has a small, yet positive coefficient. From a theoretical viewpoint one would expect this to be negative, since higher unemployment benefits would lessen the stimulant to enter the labor market. Another significant control variable that contradicts current theories is the terms of trade. An increase in the terms of trade would suggest an increase in national income, which would boost aggregate demand and thereby the demand for labor. In other words, a positive coefficient would be expected for this variable, yet this variable has a negative coefficient. The inflation rate variable has a negative coefficient which contradicts the predictions one would hypothesize from a theoretical perspective. However, the studied time period did include a relatively large economic crisis. This had led to unusual figures for these variables. For instance, figure 4 shows that the terms of trade variable showed significant large
16
decreases. Also, for the first time since 1955 (U.S. Inflation Calculator, 2015), the United States reported deflation as can be seen in figure 3. These abnormal occurrences may not have been tested thoroughly as of yet, leading to imprecise expectations of the direction of the coefficient.
The growth rate of the gross domestic product proves to be highly insignificant, and whether the coefficient is positive or negative stays unclear from the 95-percent confidence interval (for (1):[-0.2502971;0.0634248], for (2): [-0.0450493;0.0541854].
The crisis-dummy is one of the significant control variable of which the coefficient is as hypothesized. According to this dummy, the employment rate was significantly lower in the years in which the financial crisis occurred. The negative coefficient of the interest rate also coincides with theoretical predictions. The coefficient is significant and larger for the low-educated workers than for the total workforce.
17
Conclusion
The literature of minimum wage effects on employment has so far focused mainly on the subgroup of young workers. The reason here for is that this subgroup of the labor supply is most likely to work for minimum wage. It is however not the only subgroup of the labor force of which a substantial part receives the minimum wage, low-educated workers also satisfy this criteria (U.S. Bureau of Labor Statistics, 2014)
This research makes use of a random effects panel data model to determine the effects of the minimum wage on the employment rate of low-educated workers, as well as on the
employment rate of the total labor force. Doing so, enables one to compare the effects on the separate groups. For this, longitudinal statistics are used from 51 states ranging from 2005 to 2013.
The results show that the minimum wage has a significant negative impact on the
employment rate for the low-educated workers as well as for the entire labor force. A ten percent increase of the minimum wage is predicted to lead to a 0.59 percent decrease of employment of low-educated workers, referring to the approach Brown et al. (1982) used to report their results. As hypothesized, the same increase of the minimum wage leads to a smaller decrease in employment for the total labor force, namely a 0.29 percent decrease. The direction and the scale of the effect is in line with previous results from Brown et al. (1982) and Neumark and Wascher (1992). To minimize omitted variable bias, other relevant factors that may had an influence on the employment rate were taken into account. Out of the implemented control variables, most had a significant effect. However, the coefficients of some of the control variables did not point in the direction that theory would predict. This may have been the result of the tumultuous time period, yet it may have decreased the predictive power of the model.
To summarize, the key findings of this thesis are that a minimum wage has a significant negative effect on the employment rate. This effect is greater for low-educated workers than for the total labor force.
18
Recommendations
The following recommendations are proposed for future research into this specific field of research.
1. Improve the quality of the dependent variable. The used data for the employment rate of low-educated workers in this research is the only one that was available, but has flaws. For one, it only contains workers out of the age group of 25 to 64. Previous research has mainly focussed onto youth workers, since that subgroup contains a relative large amount of minimum wage workers.
2. Look into the enhancing of the key explanatory variable, the minimum wage. Assumptions were made to create a variable that had one value for all periods, while subminimum wages exist in several states. One can look into creating a weighted average minimum wage based on the amount of workers per state that work for different subminimum wages.
3. Make improvements in the use of control variables. In this research, several control
variables were insignificant or had a significant effect that could not (simply) be explained by theory. Also, the results might be deteriorated by omitted variable bias, since not all control variables that were suggested in the literature were taken into this model. Additional effort to find a method of adding the relatively complicated labour taxes into the equation might me necessary. Another form of making improvements in the use of control variables is by adding lagged effects. Brandt et al (2005) found effect of labour market reforms on employment that had a five year lag (bron 15). For now, this factor was too complicated to add.
19
Reference List
Bartolucci, F., Choudhry, M. T., Marelli, E., & Signorelli, M. (2011, July). Financial Crises and Unemployment: Beyond the Okun’s Law. In WP presented at Sixteenth World Congress of the International Economic Association,(Tsinghua University, Beijing, China 01/2011). Brown, C., Gilroy, C., & Kohen, A. (1982). The effect of the minimum wage on employment and unemployment. Journal of Economic Literature, 20(2), 487-528. Retrieved from
http://www.researchgate.net/profile/Charles_Brown24/publication/4981356_The_Effect_of_t he_Minimum_Wage_on_Employment_and_Unemployment/links/54200d4d0cf2218008d437 8e.pdf
Card, D., & Krueger, A. B. (1994). Minimum wages and employment: A case study of the fast food industry in New Jersey and Pennsylvania. The American Economic Review, 84(4), 772-793. Retrieved from http://www.mattblackwell.org/files/teaching/CarKru94.pdf
Choudhry, M., Marelli, E., & Signorelli, M. (2012). Youth and total unemployment rate: the impact of policies and institutions. Rivista Internazionale di Scienze Sociali. Retrieved from http://www.eco.unibs.it/~emarelli/CMS_unemp_instit.pdf
Heritage.org. (2015). 2015 Index of Economic Freedom. Retrieved from http://www.heritage.org/index/visualize
Center for economic and policy research. (2013). Why Does the Minimum Wage Have No Discernible Effect on Employment?. Retrieved from http://dev.takeactionminnesota.org/wp- content/uploads/2013/10/Why-Does-the-Minimum-Wage-Have-No-Discernible-Effect-on-Employment.pdf
David, C., & Krueger, A. (1995). Myth and Measurement: The New Economics of the Minimum Wage. Princeton, United States: Princeton University Press.
Doucouliagos, H., & Stanley, T. D. (2009). Publication Selection Bias in Minimum-Wage Research? A Meta-Regression Analysis. Retrieved from
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.398.2297&rep=rep1&type=pdf Econport. (n.d.). Competitive Market Model Assumptions. Retrieved June 8, 2015, from http://www.econport.org/content/handbook/industrialorg/Competitive/Assumptions.html Feliz, R. A. (1994). Terms of trade and labour supply: a revision of the Laursen-Metzler effect (No. 6). Centro de Estudios Económicos, El Colegio de México.
Gramlich, E. M., Flanagan, R. J., & Wachter, M. L. (1976). Impact of Minimum Wages on Other Wages, Employment, and Family Incomes. Brookings Papers on Economic
Activity, 1976(2), 411-420. Retrieved from http://www.jstor.org/stable/2534380 Laursen, S., & Metzler, L. A. (1950). Flexible Exchange Rates and the Theory of Employment. The Review of Economics and Statistics, 32(4), 281-299. Retrieved from http://www.jstor.org/stable/1925577
20
Lee, J. (2000). The robustness of Okun's law: Evidence from OECD countries. The American Economic Review, 22(2), 331-356. doi:10.1016/S0164-0704(00)00135-X
Mankiw, N. G. (1989). Real business cycles: A new Keynesian perspective(No. w2882). National Bureau of Economic Research.
Maurice, S. C. (1974). Monopsony and the Effects of an Externally Imposed Minimum Wage. Southern Economic Journal, 41(2), 283-287. Retrieved from
http://www.jstor.org/stable/pdf/1056739.pdf?acceptTC=true
National Conference of State Legislatures. (2015, January 6). STATE MINIMUM WAGES | 2015 MINIMUM WAGE BY STATE. Retrieved June 14, 2015, from
http://www.ncsl.org/research/labor-and-employment/state-minimum-wage-chart.aspx Neumark, D., & Wascher, W. (2000). Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Comment. The American Economic Review, 90(5), 1362-1396. Retrieved from http://www.jstor.org/stable/2677855
Neumark, D., & Wascher, W. (1992). Employment Effects of Minimum and Subminimum Wages: Panel Data on State Minimum Wage Laws. The American Economic Review, 46(1), 55-81. Retrieved from http://www.jstor.org/stable/2524738
Sen, A., Rybczynski, K., & Van de Waal, C. (2011). Teen employment, poverty, and the minimum wage: Evidence from Canada. Labour Economics, 18(1), 38-42. Retrieved from http://www.sciencedirect.com/science/article/pii/S0927537110000771
Santomero, A. M., & Seater, J. J. (1978). The Inflation-Unemployment Trade-Off: A Critique of the Literature. Journal of Economic Literature, 16(2), 499-544. Retrieved from
http://www.jstor.org/stable/2722879
Stigler, G. J. (1946). The Economics of Minimum Wage Legislation. The American
Economic Review, 36(3), 358-360. Retrieved from http://www.jstor.org/stable/pdf/1801842 Tax Foundation. (2013, April 01). State Individual Income Tax Rates, 2000-2014. Retrieved from http://taxfoundation.org/article/state-individual-income-tax-rates
U.S. Bureau of Labor Statistics. (2014). Characteristics of minimum wage workers, 2013 (REPOR T 1048). Retrieved from http://www.bls.gov/cps/minwage2013.pdf
U.S. Department of Labor. (n.d.). Wage and Hour Division (WHD). Compliance Assistance - Wages and the Fair Labor Standards Act (FLSA). Retrieved June 3, 2015, from
http://www.dol.gov/whd/flsa/
U.S. Department of Labor. (n.d.). Wage and Hour Division (WHD) History of Changes to the Minimum Wage Law. Retrieved June 8, 2015, from
http://www.dol.gov/whd/minwage/coverage.htm
U.S. Inflation Calculator. (2015, June 18). Consumer Price Index Data from 1913 to 2015. Retrieved June 19, 2015, from http://www.usinflationcalculator.com/inflation/consumer-price-index-and-annual-percent-changes-from-1913-to-2008/
21
U.S. Inflation Calculator. (2015, June 18). Historical Inflation Rates: 1914-2015. Retrieved from http://www.usinflationcalculator.com/inflation/historical-inflation-rates/
United States Census Bureau. (2008). Statistics about Business Size (including Small Business) from the U.S. Census Bureau. Retrieved June 16, 2015, from
http://www.census.gov/econ/smallbus.html
Van de Ven, A. H. (2007). Engaged scholarship: a guide for organizational and social research: a guide for organizational and social research. Oxford University Press.
22
Appendices
Table A1: Data descriptions and sources
Variable Definition Source
lemplow Employment rate of workers
with less than a high school diploma per state
American Community Survey
lemptot Employment rate of workers
per state
American Community Survey
lminwage Logarithmic function of the real minimum wage per state
United States Department of Labor
benefit Average weekly
unemployment benefit amount
United States Department of Labor
gdpgrowth Yearly growth rate of Gross Domestic Product per state
Bureau of Economic Analysis
inflation Yearly inflation rate United States Department of the Treasury
crisis Dummy variable which
contains the value 0 for 2005up to 2007 and the value 1 for all years after that
Not applicable
interest Real interest rate World Bank
tot Terms of trade ratio OECD
