3.1. Contexto investigativo
3.3.2. Conclusiones entrevista
The legislation change on 27 November 2012, which is used to identify the effect on real
wages, raised the minimum wage for Zone B and left unaffected zones A and C. So, these
untreated zones, A and C, constitute the natural control group for the DiD regressions.
But, given that the intervention was not performed in a randomized way, a common
concern in this kind of observational studies is precisely the validity of the control group.
By construction, minimum wages zones in Mexico are different. Their classification
was not carried out by regional differences, but by the level of economic development
across municipalities. Under this circumstances, in which control and treatment were
not randomized and they are not identical in their observable characteristics, the key
identifying assumption is that, in absence of the intervention, control and treatment
groups follow the same trend in their outcome variables before the policy change. To
verify this assumption, we implement a graphical inspection of the trends on real wages
by each zone.
Figure 3.1 shows the outcome trend variable (real hourly wage) for the period 2011Q1-
2014Q4. Before the intervention — the dotted vertical line denotes the period in which
the legislation change took place — the trends followed by all three zones are and parallel.
That is, we can argue that the three minimum wage zones exhibit the same trend on real
wages as the distance between the mean wage in Zone B with respect zones A and C
remains basically constant. This implies that in absence of treatment, real wages in all
three zones respond in the same way to exogenous shocks, or in other words, there is no
evidence of differences on real earnings behaviour among wages zones. As a consequence,
2The model on the estimation of the probability of being active in the labour market variable, in Chapter 4 (see Table 4.2) is not corrected for sample selection bias. It includes the variable Head of household directly as a regressor.
Figure 2.1
Real hourly wage trends by wage zone (Mexican pesos of 1F December 2010)
treatment and control groups are valid for implementing DiD regressions.
With respect to the post-treatment period, although all three zones exhibit decreasing
trends in real wages, it is possible to notice that the gap between Zone B and the other
two zones increases after 2012Q4. Visually, the difference between Zone B and Zone A
is more compelling. This, because Zone A suffered an important reduction in the real
earnings level.3 But, it is also possible to appreciate that the distance between zones
B and C was constant before the intervention, and after that, the gap increases. This
means that Zone B’s relative wages increased after the policy change so that the minimum
wage effect should be positive. However, of course it is necessary to implement a formal
causality analysis to talk about treatment effects.
A legitimate concern may be that the trend for the level of prices can be different
among zones. As specified in Appendix 2.A, real wages are calculated deflating nominal
wages by using the National Consumers Price Index (INPC). So, a single price index is
used for the three zones. The reason is that there are no price indices at the municipality
3Literature on Mexican labour market has no explored the causes of this decrease in wages in Zone A. Figure 2.2 shows that the decline is also observed in nominal wages. Appendix 2.C implements some descriptive analysis to explore some different explanations that could have lead to this reduction. But, it is important to highlight that in our econometric specifications we introduce a DiD parameter on Zone A — equation (2.1b) —to check this decline is consequence of the intervention. In all cases, the corresponding parameter is not statistically significant.
level. In fact, to construct the national price index, INEGI collects data only from 46
representative urban cities. As minimum wage zones are classified at the municipality
level, it is not possible to construct a precise price index by wage zones.
Therefore, to confirm that the pretreatment trends are actually the same, and this is
not a consequence of the use of a single price index for the three zones, we implement
the following analysis. First, Figure 2.2 shows trends by nominal wages by zones. The
trends are positive because the lack of consideration of the inflationary process, but the
conclusion remains: before the intervention there exists a parallel trend of the nominal
wages by zones, but after the policy change in 2012Q4, the gap increased.
Figure 2.2
Nominal wage trends by wage zone (Current Mexican pesos)
Figure 2.3
Real wage trends by wage zone, separate CPI (Mexican pesos of 1F December 2010)
Second, although there is no information on price indices at the municipality level, we
construct an approximation calculating a simple average of the available cities by zones
(see Appendix 2.B for a listing of the cities included in the INEGI’ INPC calculation by
minimum wage zone). Of course, there can be an important bias on these exercise because
there may be an underrepresentation of municipalities by zones. But, the purpose is only
to show that there are no fundamental differences generated by the dynamics of prices
in the wages zones. We use these new zones’ Consumer Price Indices (CPI) to deflate
shows that zones A, B, and C exhibit parallel trends before the intervention even if we
use their respective price index to deflate nominal earnings.
In summary, this subsection presents descriptive evidence to show that average real
wages in zones A and C follow parallel trends with respect to Zone B. This, independently
of the level of prices at the national and at the wage zone level. Therefore, zones A and
C constitute a valid control group to Zone C.