Vol. 11, 2017-15 | May 29, 2017 | [Link]

2017-15

Minimum Wage and Employment: Escaping the

Parametric Straitjacket
Stefano Cabras, Jan Fidrmuc, and Juan de Dios Tena

Abstract
Parametric regression models are often not flexible enough to capture the true relationships
as they tend to rely on arbitrary identification assumptions. Using the UK Labor Force
Survey, the authors estimate the causal effect of national minimum wage (NMW)
increases on the probability of job entry and job exit by means of a non-parametric
Bayesian modelling approach known as Bayesian Additive Regression Trees (BART).
The application of this methodology has the important advantage that it does not require
ad-hoc assumptions about model fitting, number of covariates and how they interact. They
find that the NMW exerts a positive and significant impact on both the probability of job
entry and job exit. Although the magnitude of the effect on job entry is higher, the overall
effect of NMW is ambiguous as there are many more employed workers. The causal effect
of NMW is higher for young workers and in periods of high unemployment and they
have a stronger impact on job entry decisions. No significant interactions were found with
gender and qualifications.
JEL C23 C11 C14 J3 J4
Keywords BART; causal inference; regression approach; matching regression
Authors
Stefano Cabras, Universidad Carlos III, Spain, and Università di Cagliari, Italy
Jan Fidrmuc, Brunel University, London, UK; IES, Charles University, Prague,
Czech Republic; and CESifo Munich, Germany, [Link]@[Link]
Juan de Dios Tena, University of Liverpool, UK; Università di Sassari, Italy; CRENoS;
and Instituto Flores de Lemus, Universidad Carlos III, Madrid, Spain

Citation Stefano Cabras, Jan Fidrmuc, and Juan de Dios Tena (2017). Minimum
Wage and Employment: Escaping the Parametric Straitjacket. Economics: The Open-
Access, Open-Assessment E-Journal, 11 (2017-15): 1–20. [Link]
[Link].2017-15

Received April 8, 2016 Published as Economics Discussion Paper April 26, 2016
Revised March 27, 2017 Accepted April 17, 2017 Published May 29, 2017
© Author(s) 2017. Licensed under the Creative Commons License - Attribution 4.0 International (CC BY 4.0)
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

1 Introduction
The most characteristic feature of the literature on the causal impact of the minimum wage on
employment is the general lack of consensus. Neumark and Washer (2007) compile an extensive
survey of previous research and conclude that the minimum wage exerts an adverse impact on
employment of low-skilled workers and a non-significant impact on total employment. However,
other surveys on this issue, a meta-analysis by Card and Krueger (1995) and the subsequent
contributions by Doucouliagos and Stanley (2009) and De Linde et al. (2014) find that there is a
wide range of results in the previous research, and that once the publication selection bias is
accounted for the mean estimate is consistent with a non-significant impact of the minimum wage
on employment.
A possible reason for the wide range of findings is the fact that the results hinge dramatically
on ad hoc assumptions about the parametric specification of the empirical model and on the
definition of the control group in the analysis. This is corroborated in the insightful and interesting
discussion in a series of papers by Allegretto et al. (2011, 2013), Dube et al. (2010), and Neumark
et al. (2014) in a state-level panel analysis for the US. Dube et al. (2010) and Allegretto et al.
(2011) suggest that it is essential to control for spatial heterogeneity in order to estimate the impact
of the minimum wage in a panel data setting. In particular, they propose to include two types of
local controls consisting of: (1) jurisdiction-specific linear time trends; and (2) interactions
between time dummy variables for sets of neighboring states or neighboring counties so they
could be used as controls to determine the impact of the minimum wages. Subsequently, Neumark
et al. (2014) and Sabia et al. (2014) criticize these measures on the grounds that there are other
non-linear ways of controlling for unobserved trends and that this approach excludes other
potential controls apart from those for the neighboring regions. Crucially, the parametric form of
the model appears to be the critical determinant of whether a significant or insignificant impact
of minimum wage on employment is obtained. Hirsch et al. (2015), in turn, argue that the lack of
a significant effect can be driven by alternative channels of adjustment such as changes in prices,
profits, performance standards, and wage compression.
Another potential problem of the minimum-wage literature mentioned above is the fact that
many studies analyze this issue using aggregate data.1 Aggregation might mask the real effect of
minimum wage at the individual level. Moreover, the analyses based on aggregate data could be
affected by endogeneity as mimimum wage movements could be caused by regional or national
macroeconomic variables (Baskaya and Rubinstein, 2012; Sabia, 2014). While policy variables
can be endogenous to aggregate employment indicators, they are clearly exogenous with respect
to specific individuals and their outcomes.
In this paper, we use the UK Labor Force Survey to estimate the causal impact of the UK
national minimum wage (NMW) on employment using a non-parametric Bayesian modelling
approach known as the Bayesian Additive Regression Trees (BART henceforth) that was
originally developed by Chipman et al. (2010) and applied to the analysis of causal inference by
Hill (2011), Sparapani et al. (2016), Tan et al. (2016) and others. This procedure shares some
similarities with standard matching estimation strategies (see for example Abadie and Imbens,
2006), as it compares unemployment-to-employment and employment-to-unemployment
_________________________
1 See for example Lee et al (1990) for a discussion on aggregation bias.

[Link] 2
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

transitions of individuals affected by the NMW increase with similar individuals who are
unaffected by the increase but are sufficiently similar to the treatment group. The BART
procedure has important advantages over other more traditional parametric specifications. Among
them, it does not require any type of hypotheses or priors over the covariates to be included in the
model, it can consider a large number of regressors, and it can estimate any type of interactive
effects between the treatment variable and any other variable. Thus, under the BART model, the
definition of the closest untreated individual for each treated individual and the interactions
between the different clusters of individuals and time or and any other relevant covariate is not
constrained to follow any specific (and potentially ad-hoc) parametric function. Furthermore, and
more importantly, the parametric function need not be specified a priori.
Our paper is closely related to at least two previous works that estimate the impact of the UK
NMW on employment at the individual level using micro-data: Stewart (2004) and Dickens and
Draca (2005). Stewart (2004) analyzes how the introduction of the UK NMW in 1999 and its
subsequent changes in 2000 and 2001 affected the employment-to-employment transition.
Dickens and Draca (2005) follow a similar approach for the NMW increases in October 2003 but
they extend the analysis to consider the separate effect of the NMW on job entry and job exit
decisions. Both study the impact of the NMW by applying the difference-in-difference technique
to the UK Labor Force Survey data, and find that the NMW does not have a significant adverse
effect on employment. Unlike these papers, we do not consider a specific year’s increase in the
minimum wage but take into account all NMW changes since its introduction in 1999. Finally,
our approach allows us to identify the interactions of the NMW effect with other relevant variables
such as gender, age, qualifications and business cycle without the necessity of proposing a
parametric specification.
The contribution of our paper is twofold. First, we shed some new light on the relationship
between the minimum wage and employment. In particular, we find that the NMW exerts a
positive and significant impact on both the probability job entry and job exit. Although the
magnitude of the effect on job entry is larger, the overall effect of NMW is ambiguous as there
are many more employed than unemployed workers. This could explain the insignificant effect
found in the previous work based on aggregate macroeconomic estimations. We find also that the
effect is stronger for younger workers and in high unemployment periods. On the other hand,
gender and qualifications play little role in shaping the minimum wage effect.
Second, we demonstrate the applicability of the BART approach to analyses of economic
outcomes without imposing a specific parametric form a priori. While we chose the minimum
wage effect on employment, this method could be applied to a broad range of other contexts
equally well.
In the next section, we present the data used. Sections 3 and 4 discuss methodological
approaches used for analyzing the labor-market impact of the minimum wage and explain the
main features of the BART model, respectively. Empirical results are shown and discussed in
Section 5. The final section summarizes our findings and offers some conclusions.

[Link] 3
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

2 Data
Our analysis is based on the UK Labor Force Survey (LFS). The LFS is a quarterly nationally-
representative survey of households across the UK. Each quarter, approximately 60 thousand
households and over 100 thousand individuals aged 16 and above are surveyed. Each household
is retained in the survey for five consecutive quarters, with one-fifth of households replaced in
each wave. The survey contains detailed demographic and socio-economic information on the
respondents, including, importantly, their labor-market outcomes. Since the NMW was
introduced in April 1999, we use all quarterly datasets available from April–June 1999 to
October–December 2011, pooling all available LFS waves during this period. In order to have a
sufficient number of observations, we include all individuals aged between 16 and 40.
The UK NMW features three different age-dependent rates: the 16–17 years old rate, the
youth rate (applying to those aged 18–21)2, and the adult rate.3 Historically, the youth rate has
remained some 35% higher than the 16–17 rate while the adult rate has exceeded the youth rate
by around 20%. The LFS reports the date of birth of every respondent and also the date the survey
was carried out. By comparing these two dates, we can determine the precise age of each
respondent on the day of the survey.4 We therefore know whether a particular individual is below
or above the age threshold at which they become eligible for a different (higher) NMW rate.

3 Methodological considerations
We analyze the effect of the NMW increases on employment by going beyond standard regression
and matching estimation methodologies traditionally used for this purpose. Regardless of the
methodology, the analysis involves comparing the changes in labor-market outcomes (such as
employment) after a NMW change for the treatment and control groups.5 Consider the impact of
NMW on the probability of job loss. The treatment group comprises workers whose wages have
to go up in the wake of an annual NMW increase because the new NMW rate is higher than their
current wage. The wages of those in the control group should be close to but just above the new
rate so as not to have to change.
_________________________
2 The upper limit for the youth rate has been lowered to 20 from October 2010. Where relevant, our analysis takes this
change into account.
3 A fourth rate, for apprentice workers, was introduced in October 2010 (we do not consider those subject to this rate
in our analysis). No minimum wage applies to those who belong to one of the few exemptions such as members of the
armed forces, volunteers, students on work placements, workers living in the employers’ households, and (until 2010)
apprentices.
4 The precise date of birth is not available in the publicly released LFS datasets. We are grateful to the Office for
National Statistics for making the restricted release of the LFS available to us.
5 Note that our approach is similar in spirit to the difference-in-difference approach in Stewart (2004) and Dickens and
Draca (2005) who compare the average change in the employment status before and after the introduction of a very
specific minimum wage policy. Of course, as we show in Table 1, the vector Xit incorporates time-invariant
characteristics. Given that we consider a whole sample of minimum wage changes through thirteen years, our analysis
is based on the estimation of the effect on the change in employment status, before and after the policy application, for
the treatment and control groups.

[Link] 4
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

More specifically, the treatment group can be defined as the individuals whose wages meet the
following condition:
nmwt < wit < nmwt+1 (1)

Where nmwt is the (age-dependent) NMW rate in effect at time t while wit is the worker i’s wage.
The control group is defined as the workers whose wage before the increase is greater than the
new NMW rate but lower than some upper bound to ensure that we only consider workers earning
just above the minimum wage (who, therefore, are likely to display similar characteristics as those
earning the minimum wage). If we set the upper bound as a fraction c above the new rate. The
control group thus comprises workers meeting the following condition:
nmwt+1 ≤ wit < nmwt+1 ∗ (1 + c) (2)

We can then estimate the following equation

𝑃(𝑒𝑡+1 = 0|𝑒𝑡 = 1) = Φ(𝛼 ∗ 𝐷𝑖 + 𝛾 ∗ 𝑋𝑖𝑡 ) (3)

where the dependent variable is the probability that individual i is unemployed conditional on
being employed in the preceding quarter, Φ(. ) is the standard normal cumulative distribution
function, Di is a dummy variable denoting individuals belonging to the treatment group, included
on its own and in interaction with the gap between individual i’s wage and the new NMW rate,
and Xit collects all remaining covariates (individual socio-economic characteristics and time
effects). An analogous equation can be estimated for the probability of remaining employed
conditional on employment in the previous quarter. In line with the standard practice, equation
(3), and in particular the coefficient estimate of the first term, is interpreted as capturing the
differentiated effect of the minimum-wage increase on the probability of becoming unemployed
for the treated individuals relative to those in the control group.
A similar approach can be used to estimate the impact of NMW on the probability of job
entry. In this case the equation to estimate is
𝑃(𝑒𝑡+1 = 1|𝑒𝑡 = 0) = Φ(𝛼 ∗ 𝐷𝑖 + 𝛾 ∗ 𝑋) (4)

A particular problem presents itself here in the fact that we do not have any previous wage
information for those who enter employment only after the NMW increase. In other words, we
do not know whether those entering into employment after the increase would have earned more
or less than the minimum wage before the increase. Dickens and Draca (2005) resolve this by
defining the treatment group as those whose earnings are less than or equal to the (age-relevant)
new NMW rate and the control group as those who earn up to c percent above the NMW:
Treatment group: 𝑤𝑡+1 ≤ 𝑛𝑚𝑤𝑡+1 (5)
Control group: 𝑛𝑚𝑤𝑡+1 < 𝑤𝑡+1 < 𝑛𝑚𝑤𝑡+1 ∗ (1 + 𝑐) (6)

A somewhat uncomfortable implication of this specification is that the treatment group now
includes also those who earn less than the NMW (there are specific cases when this is allowed,
for example for apprentices or for those who receive employer-provided accommodation or other
in-kind payments). An alternative specification would entail constructing the treatment group as

[Link] 5
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

including only those who earn the minimum wage after the NMW increase. Using that
specification yields very similar results.
Note that our analysis could suffer from a potential endogeneity problem as in a non-
experimental sample, such as the Labor Force Survey, workers earning less than the new NMW
rate are more likely to lose their jobs even if NMW does not change because they are likely to be
less productive than workers earning higher wages.6 If so, it is the characteristics associated with
their lower wages (and not the minimum wage itself) that determine their higher probability of
job loss compared to other individuals with above-NMW wages. In other words, if wages are not
allocated randomly, the allocation of individuals into treatment and control groups is not random
either but depends on their characteristics.
In order to assess to what extent the two groups of individuals are similar, Table 1 presents
some basic descriptive statistics for the treatment and control groups, for the analyses of job exit
and job entry alike. There are some differences: the individuals in the control groups are slightly
more likely to have a university degree or higher education, less likely to have lower
qualifications, they are more likely to be white rather than black or Asian, less likely to be a full
time student, and they are more likely to live in the rest of South East and South West. However,
these differences are generally small and the two groups appear rather similar.
There are two standard approaches to estimate this causal impact. One is to compare the
outcome variable of a treated individual with that of one or several other individuals who are as
similar as possible to the treated individuals with respect to the values of covariates 𝑋𝑖𝑡 . A second
approach matches participants and nonparticipants based on their estimated propensity scores.
However, the application of these methodologies is only possible if there is a region of common
support between the treatment and control groups.
Regardless of the approach used, the average treatment effect is defined as
ATE = E[Y(1) − Y(0)], where the expected value is computed with respect to the probability
distribution of Y for all individuals. We focus on the causal effect for a given set of individuals,
for example those who have received the treatment, E[Y(1) − Y(0)|D = 1], that is, individuals
affected by NMW increases. In this case, the expected value is estimated with respect to the
conditional distribution of (Y|D = 1). Even more generally, if we have a set of covariates X we
can estimate the causal effect conditional on them, that is, conditional on X = x
However, this is not always possible because matrix 𝑋 typically has a very high
dimensionality and comprises a wide range of covariates, including qualitative and quantitative
variables, and some standard approaches such as, for example, the propensity score, cannot be
applied if the number of covariates is too high. This forces the analyst to consider a set of variables
of lower dimension, putting the strong ignorability assumption in doubt.7 Besides, the

_________________________
6 Note however that in expressions (2) and (6), small values of c would imply that the salary of the treated and control
groups could be deemed to be very similar.
7 See Caliendo and Kopeining (2008) and references therein for a discussion on this issue.

[Link] 6
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

Table 1. Descriptive Statistics.

Job entry Job exit
Treatment Control Treatment Control
Highest qualification
Degree or equivalent 3.29 4.31 5.04 6.02
Higher education 3.99 4.29 4.7 5.25
GCE A level or equivalent 25.92 25.59 22.58 22.92
GCSE grade A-C or equivalent 33.38 35.35 33.88 35.72
Other qualification 16.24 15.43 18.82 15.81
No qualification 14.55 12.29 12.67 12
Ethnic origin
White 90.17 91.09 91.91 93.07
Black 6.47 6.23 6.02 5.6
Asian 1.78 1.57 1.98 1.17
Region of usual residence
Tyne & Wear 3.7 2.36 3.18 2.46
Rest of Northern region 6.47 4.83 5.49 5.2
South Yorkshire 3.62 3.49 3.96 3.43
West Yorkshire 4.11 5.14 5.7 4.76
Rest of Yorks & Humberside 4.11 3.55 4.21 3.99
East Midlands 9.36 10.27 11.06 9.63
East Anglia 3.18 4.2 2.89 2.99
Inner London 1.34 0.92 0.87 1.11
Outer London 2.68 2.78 2.19 2.17
Rest of South East 11.69 14.05 10.44 13.84
South West 7.41 9.35 9 9.7
West Midlands (met county) 5.19 4.06 3.84 4.19
Rest of West Midlands 5.19 5.56 5.16 5.67
Greater Manchester 5.16 4.79 4.21 4.45
Merseyside 3.06 2.2 2.19 2.44
Rest of North West 4.29 4.43 4.46 3.92
Wales 6.82 6.02 6.03 5.67
Strathclyde 4.93 4.56 4.75 3.81
Rest of Scotland 5.45 5.71 5.57 6.69
Northern Ireland 2.24 1.74 4.8 3.9
Whether full time student
Full time student 19.27 15.03 10.03 8.88
Not full time student 80.35 84.76 89.97 91.12
Sex
Male 32.48 28.98 27.12 30.14
Female 67.52 71.02 72.88 69.86
Notes: Missing, ‘don’t know’ and ‘other’ responses are not reported.

[Link] 7
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

specification of regression models with many variables makes it not practical to consider all
possible interactions among the variables. Again, this forces the analyst to consider only
interactive effects among first or second order covariates or to use algorithms such as the forward
or backward variable selection that may provide locally optimal models. Unfortunately, there is
no theoretical justification, only empirical results, to guide us in assessing the scope of a local
instead of a global optimum.
Due to these drawbacks, we make use of a particular type of matching estimation based on
the BART model for the estimation of causal impact of NMW increases. Being a non-parametric
model, this frees us from being restricted by a given model specification. Furthermore, it allows
us to estimate with a satisfactory precision the response of the variable of interest to NMW
increases, and with that, the counterfactual result even for a high dimensional 𝑋. An additional
important advantage of this approach is that it allows for identification of the most significant
interactive effects between the treatment variable and any of the covariates without being
constrained to include these interactions in any parametric form.
In order to assume that the outcome is independent of the treatment, it is necessary to account
for all possible conditioning factors by including a broad range of covariates, 𝑋. More specifically,
the strong ignorability hypothesis with respect to the allocation of treatment states that 𝑌 is
conditionally independent of D given 𝑋 and that the probability of treatment allocation is always
positive regardless of the specific value of 𝑋. However, like in other matching methods, this does
not preclude the possibility of selection on unobservables. Under this hypothesis, the estimation
of the marginal effects associated to the treatment variable can be considered in general as a
consistent and unbiased estimation of the causal effect of NMW on the probability of job exit and
job entry: including a relevant set of covariates in equations (3) and (4) is a sufficient condition
to ensure an unbiased estimation. However, as argued by Morgan and Winship (2007), the
regression approach can be subject to two important drawbacks. The first relates to the fact that
the causal effect of NMW is not constant across individuals. In this case, the estimated causal
effect represent a conditional variance weighted estimate of causal effects of individuals and the
causal estimation is only unbiased and consistent for this particularly weighted average that is not
usually the parameter of interest. The second problem relates to the fact that the strong ignorability
condition does not necessarily imply that treatment is uncorrelated with the error term net of
adjustment for 𝑋 as this error term depends on the specification of covariates, 𝑋. Therefore, in
order to interpret the estimation of a regression strategy as a reliable causal effect, we require a
fully flexible parameterization of 𝑋.

4 BART model
In the following explanation of the model, we mainly follow the notation of Hill (2011) and
references therein (see Chipman et al., 2010, for details of the statistical model, and Leonti et al.,
2010, for an application of this model to the estimation of a causal effect of the use of medical
plants). Let Ɗ𝑎𝑡𝑎 be the available data, that is the set 𝑌, 𝑋, 𝐷 observed for 𝑁 individuals and
𝜋(∙ | ∙) the probability distribution of the left argument conditional to the right argument. The aim
of the analysis is to estimate the posterior probability distribution of the causal effect, that is

[Link] 8
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

𝜋(𝐴𝑇𝐸|Ɗ𝑎𝑡𝑎), or the same posterior distribution but conditionally on some covariates,

𝜋(𝐴𝑇𝐸|Ɗ𝑎𝑡𝑎, 𝑋 = 𝑥). In order to do this we use a non-parametric regression model. The novelty
in these types of causal inference analyses is the use of a Bayesian regression model known as
BART. As in all Bayesian models, we need a likelihood function defined for a set of parameters,
θ ∈ Θ ∉ ℝ, and a prior distribution π(θ), θ ∈ Θ. The likelihood function, L(Y|X, D, θ) , is obtained
from the following additive regression model, where the mean of 𝑌, 𝐸(𝑌) = 𝑃𝑖 (𝑌 = 1) = 𝑃, is
determined from the sum of estimated models for the response variable:

𝑃 = Φ[∑𝑚
𝑗=1 𝑔(𝑋, 𝐷; 𝑇𝑗 , 𝑀𝑗 )] (7)

where 𝑔(𝑋, 𝐷; 𝑇𝑗 , 𝑀𝑗 ) is a classification tree with the variables and split points represented by 𝑇𝑗
and the terminal nodes denoted by 𝑀𝑗 and computed with respect to the values 𝑥, 𝐷 that belong
to the individual whose response is 𝑌. Essentially, 𝑔 is a function that gives to each individual 𝑖
their expected value in the jth tree, 𝜇𝑖𝑗 ∈ 𝑀𝑗 . This part of the model, ∑𝑚
𝑗=1 𝑔(𝑋, 𝐷; 𝑇𝑗 , 𝑀𝑗 ), operates
in the same way as the usual linear predictor in an ordinary regression model, in fact if we
substitute the sum of trees with the usual linear predictor, we were dealing with the ordinary linear
regression model with the least square estimator. Of course the problem at hand needs a very
flexible model, which would be very difficult to be obtained with a linear predictor. Essentially,
viewing by 𝑇𝑗 and the terminal nodes denoted by 𝑀𝑗 as model parameters, we allow the data to
define the terms that enter into this kind of linear predictor instead being fixed beforehand by the
analyst. The final score estimated for the ith individual would correspond to the average of the m
scores over all trees in which each tree has been grown in order to capture a specific aspect of the
relation between the response is 𝑌 and the rest of predictors. It is well known that, in order to
minimize the forecast error, classification trees tend to grow disproportionally until generating
overfitting in the response and that in general an estimator obtained from many simple trees is
more efficient than another one obtained from a single complex tree. Examples of these types of
models are Boosting (Schapire and Singer, 1999) and Random Forest (Breiman, 2001).
In order to achieve this necessary tree simplification, we use a regularization prior on the size
of the tree 𝜋(𝑇, 𝑀) as specified in Chipman et al (2010). This regularization prior precludes the
tree from growing too much and makes sure that each of the 𝜇𝑖𝑗 contributes in a marginal way to
the estimation of the response function. As Chipman et al (2010) show, the hyper parameters of
all prior distributions are specified in relation to the observed sample. It produces priors that are
dependent on the sample. This procedure, which is not very orthodox from a Bayesian point of
view, is part of the approaches known as empirical Bayes that are very popular and have been
enhanced from a theoretical point of view by Petrone et al. (2013). As explained by Hill (2011),
the results of this type of analysis are robust with respect to prior modifications.
Using the priors specified above it is possible to simulate samples of the posterior distribution
with a non-excessive computational effort using Markov Chain Monte Carlo (MCMC), more
specifically using Metropolis Hastings. In particular, the proposal distribution used in the MCMC
to update the values of 𝑇𝑗 and 𝑀𝑗 consists of adding/dropping a terminal node and changing a split
variable or a split point with the probabilities specified in Chipman et al. (2010). Such
probabilities, which finally define the proposal in the MCMC scheme, are set according to the
observations in order to guarantee an optimal mixing of the chain and so increase the precision in

[Link] 9
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

the posterior estimation. Once the posterior distribution of 𝜃 = (𝑇1 , … , 𝑇𝑚 , 𝑀1 , … , 𝑀𝑚 ) has been
obtained, the predictive distribution for the probability of job exit is:

𝑚(𝑌𝑖 |𝑥𝑖 , 𝑧𝑖 ) = ∫𝜃∈Θ 𝐿(𝑌𝑖 ; 𝜃)𝑑𝜋(𝜃|𝜋(𝜃)) (8)

where 𝐿 is the likelihood function for 𝜃 ∈ Θ and 𝜋(𝜃) is the posterior. Integral (8) is practically
estimated by generating values of 𝑃𝑖 = 𝑃𝑖 (𝑌𝑖 = 1), using the normal distribution with the mean
and variance for each value 𝜃 in the chain MCMC and the regression tress computed using the
values of regressors for individual 𝑖 , that is 𝑥𝑖 and 𝑧𝑖 . In particular we use 𝑚=500 trees and 5000
MCMC steps after an initial burn-in of 1000 steps.
In this way, the distribution for each individual and the corresponding counterfactual response
can be estimated simply by estimating the response in Di = 1 if the worker is affected by NMW
and in Di = 0 otherwise. Once these predictive posterior distributions have been obtained, the
difference between the factual and counterfactual responses are considered to obtain the
distribution of the individual causal effect. Finally, π(ATE\) is estimated from the set of the
differences for all the individuals. Finally, the estimation of the conditional causal effect is
obtained simply by considering the difference for the individuals that fulfill the condition X = x.

5 Results
As a first step, and to establish a benchmark to compare our results against, we report the results
of a probit model as specified in Equations (3) and (4), where job entry and job exit are functions
of the dummy variable for the treatment along with a set of covariates (Table 2).8 In this
regression, the parameter 𝑐 defined in the previous section is set to be 0.1 to ensure that treatment
and control individuals are comparable in terms of wages but the results are qualitatively similar
when we consider 𝑐 = 0.3, 𝑐 = 0.5 and 𝑐 = 1. The last row of Table 2 indicates that the
probabilities of job entry and job exit are both positively correlated with being in the treatment
group.
It is interesting to compare these results with those obtained with a standard matching
procedure such as the propensity score. The estimation results are qualitatively, and even
quantitatively, similar to those obtained from a regression probit model. More specifically, the
estimated causal impact for job entry is 0.051 with standard deviation 0.013 while that for job
exit is 0.03 with standard deviation 0.011.9 The fact that the two sets of results are very similar is
not surprising as the matching estimation can be interpreted as being similar to a regression

_________________________
8 Besides standard socio-economic characteristics, we also include an indicator to account for the fact that the age limit
for the adult rate was lowered from 22 to 21 from October 2010.
9 These results are available upon request.

[Link] 10
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

Table 2. Probability of job entry and job exit as a function of treatment and change.
Probit regression
Job entry Job exit
LR Chi2 445.13 400.17
Prob > chi2 0.0000 0.0000
Number of obs 7792 6746
Treatment .0517764** .024584**
(.00961) (.0086)
Notes: Marginal effects evaluated at mean values. Significance: ** 1%, * 5%.

that puts more weight on the observations in the treatment and control groups that are very similar
to each other.
The Bayesian approach considered here is instead based on the estimation of the expected
value of the treatment and control groups using the same explanatory variables in both cases.
Figure 1 reports the estimated distributions of the total causal impact of increases in the minimum
wage rate on job exit and job entry using the BART model with all workers aged 18–40. The
results indicate that the treatment has positive effects on both job entry and job exit, in a manner
similar to the probit results reported above. More specifically, the NMW exerts a positive impact
on job entry, and the mean value of this causal impact is 5% with a 95% confidence interval of
[3.2%, 6.9%]. For job exit, the effect is positive with the mean value equal to 2% and with a 95%
confidence interval of [1%, 4%]. Three cautionary notes are required here. First, the
aforementioned effects are only measured for those workers actually affected by the NMW
increase. The minimum-wage increases only affect low-paid workers and

Figure 1. Posterior distribution of the causal effect of job entry and job exit

[Link] 11
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

need not apply throughout the distribution of wages. Second, although the estimated effect on job
entry is larger than that for job exit, the overall effect of NMW is ambiguous as there are many
more employed workers (who are candidates for job exit) than unemployed individuals
(candidates for job entry). Third, it is possible that NMW increases have spillover effects whereby
wages just above the new minimum wage also increase. The available literature suggests that such
spillovers are small or none in the UK (Dickens and Manning, 2004; Stewart, 2011). If present,
such spillovers would bias the estimated effects downwards.
As discussed above, one of the most important advantages of the BART approach is that it
allows for the simultaneous estimation of any kind of interaction between the treatment variable
and any of the covariates. This is possible either at model estimation or at description level of the
obtained results. Here we consider the result at the description level by inspecting the interaction
between covariates and the estimated causal effect. In particular, the interaction with categorical
variables is evaluated trough boxplots, which include 95% percentile bootstrap confidence
intervals for the median, while the interactions with continuous covariates by local polynomial
regression smoother (loess) along with their 95% confidence intervals (Cleveland et al., 1992,
Chp. 8) .
In Figure 2, we present the interaction between the NMW increase and the size of the increase.
While the effect of size is significant, no systematic pattern can be discerned: the estimated effects
oscillate around the mean values reported in Figure 1, neither increasing nor decreasing as the
size of the NMW change goes up. Next, we interact gender with the effect of NMW increases
(Figure 3). Again, the previous finding of a greater effect of NMW increases on job exit than on
job entry is reproduced. Although for job entry it is clear that the median values are significantly
different, the distributions of the two effects are very similar which suggests that gender plays
little role. In Figure 4, in turn, we consider the interaction with age (expressed in months rather
than years). Here, the pattern is different for job exit and entry. While the causal impact of NMW
is decreasing with age in both cases, that decline is much steeper for job entry. This is not
surprising, given that young workers are more vulnerable to NMW increases. Besides, the
interactive effect is clearly stronger for job entry. In Figure 5, we consider the interaction with the
highest attained qualification. Again, although it is possible to observe significantly different
mean values associated to the different qualifications, the whole distribution of the estimated
causal effect indicates that this variable is not a relevant factor to explain differences in the causal
impact of NMW either for job entry or job exit. Finally, Figure 6 presents the interaction with the
regional business cycle – measured using the unemployment rate. Interestingly, this interaction
effect is very different for the two labor-market flows: the minimum-wage effect on job exit is
relatively low and depends little on the regional unemployment rate, whereas that for job entry is
higher and positively related to regional unemployment. This implies that the effect of the
minimum wage on job entry differs considerably between recessions and booms, whereas the
business cycle has little bearing on how the minimum wage shapes job exits.
So far we have been considering the effects of NMW changes for two similar groups, those
affected by the change and those who are unaffected but are otherwise similar to the affected
individuals both in terms of their wage and in terms of the other covariates used in the analysis.
However, to test for the robustness of our results even further, we carry out a falsification

[Link] 12
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

Figure 2. NMW Increases, Job Exit/Entry and Size of the Increase

Figure 3. NMW Increases, Job Exit/Entry and Gender

[Link] 13
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

Figure 4. NMW Increases, Job Exit/Entry and Age

Note: Shadow area indicates the 95% confidence interval of the local polynomial regression estimator (loess).

Figure 5. NMW Increases, Job Exit/Entry and Qualifications

Notes: 1 Degree or equivalent, 2 Higher education, 3 GCE A Level or equivalent, 4 GCSE grades A-C or equivalent,
5 Other qualifications, 6 No qualification, 7 Don’t know

[Link] 14
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

Figure 6. NMW Increases, Job Exit/Entry and Business Cycle

Notes: The horizontal axis measures the regional unemployment rate. Shadow area indicates the 95% confidence
interval of the local polynomial regression estimator (loess).

experiment whereby we define the treatment and control groups as if the NMW were equal to the
actual NMW plus 2£. Our hypothesis is that neither job entry nor job exit should be affected by a
wrongly-defined increase in the NMW. The results of this experiment, shown in Figure 7, indicate
that the causal impact of the (false) NMW increase is significant at the 5% level for job entry but
not for job exit. We find similar conclusions for other artificial NMW rates (results are available
from the authors upon request). Importantly, the insignificant falsification test results for job exit
give strong support to the finding that the employment of workers earning the minimum wage is
adversely affected by NMW increases.
The fact that the falsification test is significant for job entry decisions could be due to potential
unobservable variables not included in the model. Another possibility is that it is driven by
spillover effects of the actual NMW increase: the NMW change can lead to ripple effects for wage
rates above the minimum wage.10 To account for this possibility, we consider an alternative
definition of the control group:

_________________________

10 As long as the direct effect of the NMW increase is larger than the indirect (spill-over) effect, we can obtain a
significant result. Otherwise, it would be impossible to define control and treatment groups. As discussed above, the
available evidence so far suggests that such spillovers are limited or zero (Dickens and Manning, 2004; Stewart, 2011).

[Link] 15
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

nmwt+1 ∗ (1 + a) ≤ wit < nmwt+1 ∗ (1 + c) ∗ (1 + a + 𝑐) (9)

where a = 0.1. The treatment group is defined as before (see equation 2). The new definition
ensures that the control and treatment groups are close enough but are not immediately adjacent
to each other in the distribution of wages. The results with the alternatively defined control group,
presented in Figure 8, confirm the previous results: the mean effect is the same for job exit and is
only slightly higher for job entry with the new control group. This suggests that the spillovers are
very limited, if any.

Figure 7. Falsification Test: NMW + £2

Notes: The falsification test simulates the NMW being £2 higher than the actual value.

Figure 8. Alternative Control Group

[Link] 16
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

6 Concluding remarks
We estimate the causal impact of the NMW on the probability of job entry and job exit in the UK,
applying a novel methodology to this context, the Bayesian Additive Regression Trees (BART).
An important advantage of this procedure is that it allows the identification of the most important
interactions between the treatment variable and other covariates in the model. We find that the
NMW exerts a significantly positive effect both on job entry and job exit, with the impact on job
entry being relatively stronger (given that there are fewer unemployed than employed workers,
the absolute size of the flows cannot be readily compared). The causal effect of NMW is found
to be higher for young workers and in periods of high unemployment; both of these interactions
are more prominent for job entry than for job exit. However, no significant interactions were
found with gender and worker qualification. Overall, the effect of the NMW on low-paid workers
is stronger for job entry than for job exit.
Most importantly, our paper opens new lines of research that can be explored in subsequent
work. For example, this fully flexible approach could be adapted to deal with some recent issues
in the literature about the importance of the econometric specification on estimating the effect of
minimum wage using panel data models in US states. Also, it could be used to estimate the
possible interactions between the federal minimum wage and the state minimum wages, as done
by Baskaya and Rubinstein (2012), without the necessity of estimating two different models.

Acknowledgement We received helpful comments and suggestions from Brendan McCabe, Andy Tremayne and
David Forrest, as well as seminar/conference participants at the University of Liverpool and “The German Minimum
Wage – First Evidence and Experiences from Other Countries” conference at the Institute for Employment Research
(IAB) in Nuremberg. Comments and suggestions from two anonymous referees are gratefully acknowledged.

[Link] 17
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

References
Abadie, A., and Imbens, G.W. (2006). Large Sample Properties of Matching Estimators for Average
Treatment Effects, Econometrica 74(1): 235–267.
[Link]

Allegretto, S., Dube, A., and Reich, M. (2011). Do Minimum Wages Really Reduce Teen Employment?
Accounting for Heterogeneity and Selectivity in State Panel Data. Industrial Relations 50:
205–240.
[Link]

Allegretto, S., Dube, A., Reich, M., and Zipperer, B. (2013). Credible Research Design for Minimum Wage
Studies. Working paper, Institute for Research on Labor and Employment, UC Berkeley.
[Link]

Baskaya, Y., and Rubinstein, Y. (2012). Using Federal Minimum Wages to Identify the Impact of Minimum
Wages on Employment and Earnings across the U.S. States. Unpublished paper, Central Bank of
Turkey. [Link]

Breiman, L. (2001). Random Forests. Machine Learning 45: 5–32.

[Link]

Card, D. and Krueger, A.B. (1995). Time-series Minimum Wage Studies: A Meta-Analysis. American
Economic Review 85: 238–243. [Link]

Caliendo, M. and Kopeining, S. (2008). Some Practical Guidance for the Implementation of Propensity
Score Matching. Journal of Economic Surveys 22: 31–72.
[Link]

Chipman H.A. , George E.I., and Mculloch R. (2010). BART: Bayesian Additive Regression Trees. Annals
of Applied Statistics 4(1): 266–298. [Link]

Cleveland W. S., Grosse E. and Shyu W.M. (1992). Local Regression Models. Chapter 8 of Statistical
Models in S eds J.M. Chambers and T.J. Hastie, Wadsworth & Brooks/Cole.

De Linde Leonard, M., Stanley., T.D., and Doucouliagos, H. (2014). Does the UK Minimum Wage Reduce
Employment? A Meta-Regression Analysis. British Journal of Industrial Relations 52(3): 499–520.
[Link]

Dickens, R., and Draca, M. (2005). The Employment Effects of the October 2003 Increase in the National
Minimum Wage. CEP Discussion Paper No 693, Centre for Economic Performance, London School
of Economics. [Link]

Dickens, R., and Manning, A. (2004). Spikes and Spillovers: The Impact of the National Minimum Wage
on the Wage Distribution in a Low-Wage Sector. Economic Journal 114: C95–C101.
[Link]

Doucouliagos, H., and Stanley, T.D. (2009). Publication Selection Bias in Minimum-wage Research? A
Meta-regression Analysis. British Journal of Industrial Relations 47: 406–428.
[Link]
Dube, A., Lester, T.W., and Reich, M. (2010). Minimum Wage across State Borders: Estimates Using
Contiguous Counties. Review of Economics and Statistics 92: 945–64.
[Link]

[Link] 18
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

Hill, L.H. (2011). Bayesian Nonparametric Modeling for Causal Inference. Journal of Computational and
Graphical Statistics 20(1): 217–240.
[Link]

Hirsch, B.T., Kaufmann, B.E., and Zelenska, T. (2015). Minimum Wage Channels of Adjustment.
Industrial Relations 54(2): 199–239.
[Link]

Lee, K.C., Pesaran, M.H., and Pierse, R.G. (1990). Testing for Aggregation Bias in Linear Models.
Economic Journal 100(400): 137–150.
[Link]

Leonti, M., Cabras, S., Weckerle, C.S., Solinas, M.N., and Casu, L. (2010). The Causal Dependence of
Present Plant Knowledge on Herbals – Contemporary Medicinal Plant Use in Campania (Italy)
Compared to matthioli (1568). Journal of Ethnopharmacology 130 (2): 379–391.
[Link]

Morgan, S.L., and Winship, C. (2007). Counterfactuals and Causal Inference: Methods and Principles for
Social Research. Cambridge University Press.
[Link]
interest/counterfactuals-and-causal-inference-methods-and-principles-social-research-2nd-
edition?format=PB&isbn=9781107694163#t7APtUeQG3J4Kddu.97

Neumark, D., and Wascher, W. (2007). Minimum Wages on Employment. Foundations and Trends in
Macroeconomics 3(1–2): 1–182.
[Link]

Neumark, D., Salas, I.M.I., and Wascher, W. (2014). Revisiting the Minimum Wage Employment Debate:
Throwing out the Baby with the Bathwater? Industrial and Labor Relations Review 67(3): 608–648.
[Link]

Petrone S., Rousseau J., and Scriocciolo, S. (2014). Bayes and Empirical Bayes: Do They Merge?
Biometrika, 101 (2): 285–302.
[Link]
merge

Sabia, J.J. (2014). Minimum Wages and the Business Cycle: Does a Wage Hike Hurt More in a Weak
Economy? Employment Policies Institute: Washington, DC.
[Link]
hurt-more-in-a-weak-economy/

Schapire, R.E., and Singer, Y. (1999). Improved Boosting Algorithms Using Confidence-rated Predictions.
Machine Learning 37: 297–336.
[Link]

Sparapani, R.A., Logan, B.R., McCulloch, R.E., and Laud, P.W. (2016). Nonparametric Survival Analysis
Using Bayesian Additive Regression Trees (BART). Statistics in Medicine 35(16):
2741–2753. [Link]

Stewart, M.B. (2004). The Employment Effects of the National Minimum Wage. Economic Journal
114(494): C110–C116. [Link]

Stewart, M.B. (2011). Wage Inequality, Minimum Wage Effects and Spillovers. Warwick Economic
Research Paper 965. [Link]

[Link] 19
 Economics: The Open-Access, Open-Assessment E-Journal 11 (2017–15)

Tan, Y.V., Flannagan, C.A.C., and Elliott, M.R. (2016). Predicting Human-driving Behavior to Help
Driverless Vehicles Drive: Random Intercept Bayesian Additive Regression Trees. Available
online: [Link]

[Link] 20
Please note:
You are most sincerely encouraged to participate in the open assessment of this article. You can do so by
either recommending the article or by posting your comments.
Please go to:
[Link]

The Editor

© Author(s) 2017. Licensed under the Creative Commons License - Attribution 4.0 International (CC BY 4.0)