Suresh Narayanan Comment on Wage Differentials in Malaysia: Public Employment, Gender,  and Ethnicity

Abstract

Suresh Narayanan, Universiti Sains Malaysia: This is an interesting paper on an under-researched topic. However, the gaps in the paper diminish somewhat its value.The paper is motivated primarily by empirical observations elsewhere indicating that public sector earnings are, on average, higher than in the private sector. There is no theoretical framework guiding this. The authors have therefore overlooked several theories of public sector wage determination. The oldest, best known, and perhaps most pertinent to this paper is the work by Fogel and Lewin (1974), who argued that discretion in decision making and the political process in wage setting provide upward-biased wage rates for most government jobs. Thus, the Fogel-Lewin framework predicts that public sector wages will be higher than private sector wages for most job descriptions.The authors give the impression their work on public sector wage differentials is the first because they do not refer to any other study in their literature review. This suggests that they were unaware of an important earlier study on public–private sector employment decisions and wage differentials in Peninsular Malaysia by Seshan (2013), who used individual-level data from 1995 and 2007.The justification for the study can be strengthened. Even if earnings differences exist between the two sectors, as they argued, can that alone justify why it is important to study these differences? The authors can enrich the paper by expanding on the role of the public sector and its importance in the Malaysian context as the basis for studying public–private sector differences in earnings.The authors examined three questions. Are similarly productive individuals rewarded with higher earnings if they are employed by government agencies, as opposed to private sector establishments?Is the relative size of gender wage differences smaller in the public sector as opposed to the private sector?Is the earnings gap between ethnic Malays and ethnic Chinese lower in the public sector?The first objective follows from empirical findings elsewhere that public sector employees are usually better paid than their private sector counterparts. The second objective follows from the empirical findings in Malaysia (and elsewhere) that suggest that men, on average, are better paid than women. However, the third question appears to proceed largely based on assumption as no previous empirical studies on this issue was discussed in the literature review section. Also, this objective is missing in the statement on objectives of the paper. In any case, to accommodate these three objectives logically, the motivation for the study and the problem statement should have been reworked. As it stands, the third objective appears as an afterthought.It is interesting that the authors note early on that wage differentials could “arise due to government regulations, coverage by federal laws such as minimum wage laws, and union strength.” But the authors failed to consider possible discrimination against women or certain ethnic groups, which is often cited as a key factor in empirical work on differences in earnings.The paper makes a strong case for looking at earnings differences in the public sector but fails to define what constitutes public sector employment in Malaysia. This is important because there is no universally accepted definition. For example, in the UK, public servants are defined (narrowly) as those working in government departments that report to ministers. This leaves out the employees of the UK's National Health Service, which employed 1.3 million people (in 2014). In contrast, Malaysia includes all doctors, nurses, and support staff in the public health system as part of the civil service. Similarly, in Australia, teachers, doctors, soldiers, and police are not included as part of their Australian Public Services, but in Malaysia they are.An advantage the authors have is their access to data from the Salaries and Wages Surveys, which are usually not available to researchers. Malaysian agencies pride themselves on collecting data and keeping these away from academics and researchers.However, it is important to note that the survey data excluded several groups, including unemployed workers. The latter omission results in a familiar selection bias problem that the authors fail to recognize or acknowledge. Excluding unemployed workers would not matter if their missing earnings data were missing completely at random. However, the decision to work or not to work is made by individuals and they constitute a self-selected sample, not a random sample. Some individuals may choose not to work because the remuneration is deemed inadequate. Ignoring these individuals will bias the estimated earnings. There are solutions to this problem, but if the authors have reasons to believe that making these corrections are only likely to yield marginal gains, they should say so. In any case, the problem should have been acknowledged.In studies of this nature, education, age, and experience are key explanatory variables. The data set lacks information on experience, a key variable. The authors, therefore, use age as a proxy, without alerting readers to the bias this will introduce. Age is a poor proxy for experience and will overstate the experience of young workers, and women who move in and out of the workforce due to marriage, childbearing, or family commitments. An alternative (but still an imperfect one) is to estimate experience by subtracting years of schooling plus six years from the age. This will still lead to an overestimate of experience, particularly for a relatively youthful workforce—but less so. Once again, there is no indication that this issue was recognized.The presentation of the empirical results also appears incomplete. In my view, it would have been clearer if the omitted variable is identified in each case during the discussion rather than leaving the reader to guess from Table 6. Moreover, tables that present regression results should contain basic information such as R2 values, F-tests, a note on whether the figures in parentheses refer to standard errors, and clearly defined variables. These are all missing from Tables 6. There were also no reports on tests of robustness such as tests for multicollinearity (VIF statistics) and heteroscedasticity. Being cross-sectional data, I am fairly certain heteroscedasticity will be present. So, reporting heteroscedasticity-corrected standard errors (robust standard errors) is a good practice (Stock and Watson 2018).I was also puzzled as to why Section 5.2 was titled “Age earnings profiles.” It looked more like an augmented Mincerian function to me. Besides, was not age introduced as a crude proxy for experience?The other point is that when the left-hand side is a variable expressed in log form, the coefficient of a dummy cannot be interpreted directly, as has been done in the paper. Although it is convenient to do so, and I have seen this done in other papers as well, the correct method is to take the antilog of the coefficient and subtract one from it. Multiplying the result by 100 will give the percentage change in earnings when the dummy shifts in value from zero to one [that is, 100 (eβ)-1]. The difference can be quite substantial (Wooldridge 2016). For example, in Table 6, the coefficient of the dummy for public employees of 0.575 was interpreted to indicate that public employees had about 57.5 percent higher earnings than those in the private sector in 2011. The correct figure would be 77.7 percent higher.Also, the sources for Tables 1–5 are more accurately the Salaries and Wages Surveys of 2011 and 2016, respectively. The authors’ calculations are based on these data. I think this should be made clear by stating Authors’ calculations based on SWS of 2011 and 2016.The empirical specification added rather arbitrarily dummies for life in an urban area, state of residence, occupation, industry, and field of study. Yet Table 6 leaves out state, occupation, industry, and field of study to “make the discussion easier.” This begs the question as to why they were included in the first place. Since their inclusion affects the estimates we should be told at least if they were statistically significant or not.I also have a question about the interaction terms. The interaction terms have a public sector–private sector dummy, with the dummy taking a value of 1 for the public sector and zero otherwise. This being the case, are the interpretations drawn from them correct? Additionally, since age was a proxy for experience, why was an interaction term not introduced for age?Finally, the authors claim to have made three contributions: First, they showed that the gender and ethnic earnings gap remain but have declined in recent years. Second, the gender earnings gap is much smaller in the public sector. Third, and in their view, “most important,” one must consider differences in public and private employment when examining earnings in Malaysia.The third, is to my mind, a non-starter: The public sector was ignored not because it was deemed unimportant but because the data needed were often unavailable or because researchers were denied access to them. Even so, as noted previously, there was an early study on the subject (Seshan 2013).It was disappointing that the authors offered no possible explanations for the other two findings. Although the finding that the ethnic and gender earnings gap have declined is no surprise, the authors should discuss, even if briefly, the developments that might have contributed to this decline in the earnings gap.The third finding was the most intriguing, but it was not taken up by the authors. They found that the gender earnings gap in the public sector was much smaller than in the private sector. The surprise here is not that the gender gap was smaller but that it exists at all. The smaller gap is expected, given that the government is committed to eliminating gender-based discrimination and has passed several legislations in support of this commitment. The real question to explore is why wage differentials favoring men continue to persist in the public sector? The authors are silent on this. Seshan (2013), contrary to the findings of the present study and conforming to a priori expectations, found little evidence of a gender wage gap in the public sector; he noted that a gender wage gap was more evident in the private sector. Are we then to believe that somehow gender wage differences have cropped up in the public sector over the years? And if so, what might explain this?