Introduction

Since 1986 Vietnam has been experiencing a fundamental restructuring of its economic system toward a market economy. In this new phase, one of the many challenges faced by policy makers is to formulate adequate labour market policies. In an attempt to provide some insight into the functioning of the Vietnamese labour market, this thesis focuses on understanding public ? private wage differentials in general and moonlighting in particular.

An analysis of public ? private wage differentials is particularly important and timely for Vietnam because fiscal deficits (4.15 per cent for 1995 - 2000) and external debt have placed public sector employment and compensation under increased scrutiny. Since the government wage bill forms a high proportion of current budget expenditures (21 per cent for 1996 - 2000), decreasing it is often viewed as an attractive way to reduce fiscal deficits. Furthermore, government pay scales in Vietnam often serve as the prevailing model, if not as a lever for wage earners in the private sector. When public sector employment is the dominant component of the wage sector (in our sample, public sector comprises 39 per cent), the pay structures and work conditions have strong influences on the private sector. To the extent that the wage productivity relationship is weak in the public sector, allocate inefficiencies are generated and modification of public wage scales may be in order.

A natural question that arises is ?Do workers with the same productivity characteristics receive equal total remuneration in the public and private sectors?? In other words, ?Are government workers underpaid in comparison with their private sector counterparts?? The answer to this question not only has implications for the size of the budget but also for the entire economy. If public wages are too high, they exert upward pressures on wages in the private sector, with obvious employment and efficiency implications. If they are too low, they will lead to an unmotivated public workforce. In addition, distortion effects on the public sector also occur, through, for instance, moonlighting activities.

The thesis deals with a main question: Are government workers underpaid in comparison with their private sector counterparts?

In addressing the research question, qualitative and quantitative methods are used, including statistical and descriptive analysis, review of historical trends, and comparative methods. Besides, quantitative method is extensively used.

Data used in the analysis extracts from the VLSS97/98 that was conducted by General Statistics Office with the technical assistance of World Bank.

Excluding the Introduction and Conclusion parts, the thesis is structured into four chapters. Chapter I introduces a general framework of public private wage differentials. Chapter II is devoted to measuring the wage differentials between public and private sectors in Vietnam in 1997 ? 98. Chapter III decomposes these differentials into various factors to make clear whether public workers were underpaid. Chapter IV will examine whether these wage differentials have impact on public moonlighting activities or not.

Chapter I: Theoretical review and methodological framework

I. Theoretical review

The causes of variations in wages among people are complex and controversial. However, they can be divided into two groups. With the assumption of perfect labour movement, Orthodox economists believe that, under competitive conditions, the same wage must be paid for a given labour grade of worker no matter where it is. In other words, wage differentials between workers result from the differences in worker capital. However, the Orthodox interpretation of the labour market and wage determination has long been a subject of controversy and debate. An alternative to the Orthodox theory is the Institutional view of the labour market.

The Institutional economists agree with the Orthodox view that there is a perfect movement of labour, and then wage differentials reflect the characteristic differences between the two groups of workers. However, they argue that many barriers such as trade unions, the role of political, ideological and workplace relation as well as the state lie outside the standard Orthodox treatment of the labour market.

II. Methodological framework

1. Wage equations

A Micerian statistical wages function is introduced.

Ln W = a + bX + U                                                                               (1)

Where: ln W is the natural log of hourly wage; a is intercept; X is a vector of individual characteristics such as education attainment, experience, and diplomas; b is a vector of coefficients or return to characteristics; U is a random disturbance term reflecting unobserved ability characteristics and the inherent randomness of earning statistics. It is usually assumed that U is normally distributed with mean zero and constant variance.

The above Mincerian wage equation may face the problem of selectivity bias. This possible non-randomness of the sample implies that OLS may not be consistent. This problem will be solved with a two ? step procedure proposed by Heckman (1979).

Heckman (1979) introduce a two ? step procedure to overcome the selectivity problem. The first step is to estimate a multinominal probit model, which determines the probability of entering the public sector.  From this equation, it is possible to calculate the correction term called lambda. This correction term is then included in the wages equation as an additional regressor in the second stage. The presence of the correction term corrects for possible sample selection bias.

2. Decomposition of public private wage gap

Because discrimination may exist between private and public sectors, a dummy variable is included to reflect this discrimination. As a result, the equation (1) will be modified as follows:

Ln W = a + a₁D + bX  + b₁DX + U                                     (2)          

Where: D=1 if public and D = 0 if private. If there is no discrimination between the two sectors, then both a₁ and b₁ equal zero.

An important feature of the least square estimator is that the fitted regression line passes through the points of sample means.  This implies that:

Public: _Pu= a + a₁ + (b  + b₁) _Pu                                       (3a)

Private: _Pr= a  + b _Pr                                                          (3b)

Then the differentials between the two sectors (public minus private) are

_Pu -  _Pr = a₁ + b_1Pr  + (b+b₁)(_Pu - _Pr)         (4a)

or

_Pu -  _Pr = a₁ + b_1Pu  + b(_Pu - _Pr) (4b)

Equations (4a) and (4b) are very important, since as stressed by the Blinder and Oaxaca (Berndt, 1991), they state that the mean difference in log wages between public and private sectors can be decomposed into three components:

The first component is the environment difference (a₁₎, which means that the economic environment in one sector is more favorable than that in other sector. The second component is the return difference (b_1Pr) in (4a) or (b_1Pu) in (4b) which means that one sector may have higher returns to the characteristics than the other does. The third component is the characteristics difference [(b + b₁)(_Pu - _Pr)] in (4a) or b(_Pu - _Pr) in (4b) which means that one sector (say, public sector) may have richer endowments (characteristics) than the other may.

The first and second components together reflect the effects of discrimination after controlling for worker characteristics.

The Oaxaca decomposition is subjected to index number problem. The decomposition can be quite sensitive to which wage structure is used, but neither is preferable to the other in a priori. Neumark (1988) proposes a general model to overcome this problem. Neumark shows that the nondiscriminatory wage structure can be estimated from wage function that is estimated over the pooled sample (that is both public and private workers). With this nondiscriminatory or ?pool? wage structure, g, the Neumark decomposition is thus:

_Pu -_Pr =a₁ +g(_Pu  -_Pr) +_Pu(g  - (b + b₁)) +_Pr (g - b) (5)

The first component is the environment difference (a₁₎, which means that the economic environment in one sector is more favorable than that in other sector. As a result, workers in the former can be better off than workers in the latter (after controlling for worker characteristics). The second term on the right side is that part of the wage gap explained by differences in charateristics, given nondiscriminatary returns. The third and fourth terms show the contribution of differences between actual and pooled returns to public and private workers, respectively. The sum of environment difference, deviation of public returns, and deviation of private returns reflects the effects of discrimination after controlling for worker characteristics.

The Neumark decomposition is attractive, although, it should be interpreted with care. The first note is that whether the pooled coefficients will, in fact, be good estimators of the nondiscriminatory wage structure is not clear. Further more, conventional wage functions are likely to be mis-specified, omitting a number of important variables that affect productivity. As a result, we refer to the coefficients, g, as giving the pooled wage structure instead of automatically attributing public ? private differences in returns to disrimination.

III. Literature review

A lot of works have been done to compare wages between public and private sectors in developing countries. To my knowledge, works have been done for Peru (Stellner, 1989), C?te D?Ivoire (Van Der Gaag, 1989), Ethiopia (Mengistae, 1998), Haiti (Terrell, 1993), Latin America (Ugo, 2000), Bolivia (Trine M., 2000), Indonesia (Filmer D. and L. Lindauer, 2000), Turkey (Tansel A., 2000), and Vietnam (Bales S. and Rama M., 2001).

From the above works, the status of public and private workers is different from one country to another. In some countries, public workers are better paid in comparison with private ones. On the contrary, in some countries, private workers are better treated. The methodology used is also different from each paper. Some papers use switching regression function to correct for selectivity bias. Some use Heckman two-step selection model. However, they all take into account of and deal with the problem of possible selection bias.

Chapter II: Public - private sector wage comparisons

I. Data set and general picture of wage earners

The data used in this study are drawn from VLSS 1997 ? 98, an unusually comprehensive and ?clean? micro data set developed jointly by the World Bank (WB) and Vietnam General Statistics Office (GSO). The VLSS 1997-98 provides detailed socioeconomic information over 6000 households (GSO, 1999).

The analysis is confined to wage and salary earners who were in labour force and who reported positive remuneration and positive hours worked in their main job during the week prior to the survey.

The total sample consists of 2981 workers. Of which, the public sector workers comprise civilians employed by the governmental administration, police, military, Communist party and social organizations, and state enterprises. Private workers are individuals employed by cooperatives, private enterprises, small household enterprises, joint companies (stock or limited liability companies), 100 per cent foreign enterprises, and joint ventures. In short, public workers are those whose wages are from the government budget and hence controlled directly by the state (Resolution 06/CP dated 21 Jan. 1997). In our sample, government workers comprise 39 per cent (1154 in absolute value).

The dependent variable is the natural log of hourly wage rate. Cash and others in kind benefits are included in the wage rate.

II. Wage comparison between public and private sectors

This part has shown multi-aspects of wage differentials between public and private sectors. Although private wages sometimes are higher than public wages as we take education levels, gender, economic activities, professions, regions, and urban-rural into account, on average the former are lower than the latter. These non-econometric analyses of wages across many fields are informative, however, they do not enjoy the advantages of a regression approach.

Chapter III: Econometric decomposition of public - private wage differentials

I. Description of variables, optimum regression and the results

1. Description of variables

The dependent variable is the natural log of hourly wage rate, corrected for cross-region price index.

The set of regressors includes the human capital variables as follows

?          Education

This is the previous education of the worker. This category is classified into 5 levels: no education, primary education, lower ? secondary education, upper ? secondary education, and university or higher. No education is treated as the benchmark. Four dummy variables are used for the remaining levels. It is expected that higher levels of education will have bigger returns.

?          General experience

General experience is potential years of working experience of the worker. It is defined as (current age ? age when respondent stopped studying and began working). As stated by the human capital theory, workers with higher general experience are likely to receive higher wage rates. Furthermore, human capital theory also suggests that wages should generally not be constant after leaving school but should follow a parabolic shape, peaking somewhere in midlife so that we need to add experience in quadratic to the regression function.

?          Specific experience

It is experience obtained in the current main occupation. This category is measured in months. Specific experience squared is also included in the regression function for the same reason as the general experience squared.

?          Vocational training

Vocational training is recorded one if the worker attended or are attending a vocational training course and zero if otherwise. No vocational training is used as the base. Workers with appropriate vocational training are likely to work more effectively so that they would be offered larger wages.

?          Gender

It is the sex of the worker. Because wages may be different between male and female workers as well as between sectors, this variable is taken into consideration to examine whether there are gender wage discriminations between the two sectors. A dummy variable is defined for gender that takes the value one if the worker is male and zero otherwise (female).

?          Parental education

Father and mother education is included because there would be intergenerational benefits. The children of better-educated parents grow up in a more desirable home environment and receive better care, guidance, and information preschool education. In addition, parental education also affects the worker?s possibility to have a well ? paid job. Parental education is measured as schooling years of the worker?s parents. It is likely that parents? education has positive impact on the worker?s wage rates

In addition, some variables reflecting job characteristics such as professions are taken into the regression function. The inclusion of these variables is based on the argument that works are different and hence are compensations. 4 dummy variables are used for 5 professions as managerial/clerical category is treated as the benchmark. Regional variables such as regions and rural-urban residence are also used to reflect geographical difference in wages between the two sectors.

2. Explanation for the optimum regression function

The equation (2) introduced in the first chapter is used to decompose the wage differentials between the two sectors. The optimum regression functions with and without selectivity correction have been achieved by utilizing a ?top ? down procedure? to drop out insignificant variables (Gujaratti, 1995, chapter 8). Having these optimum regression functions at hand, we have no worry about insignificant coefficients as they have been dropped out at 10 per cent level of significance.

3. Explanation of the results

As having explained in chapter one, the thesis uses both Oaxaca and Neumark decompositions to analyze the wage differentials between public and private sectors. In addition, results of regression and decomposition with and without selectivity correction are also reported for each type of decomposition.

For Oaxaca decomposition, both functions (4a) and (4b) in chapter one are used to explain the results. The thesis also uses Neumark decomposition to explain the public private wage differentials. However, before achieving the results of regression and decomposition, a pooled regression function is estimated to get estimates of all relevant variables. The top ? down procedure to estimate the optimum pooled regression function is used.

A methodological issue relates to the appropriate wage gap to be used in the docompositions. In the model without selectivity correction, we use actual wages received by individuals in our samples. The difference in actual wages is a measure of the difference in the accepted wages by public and private workers. But in the model with selectivity correction, the above method is inappropriate. According to Appleton (1999), we should take the wages offered to public and private workers into our analysis. This is net of the impact of selectivity correction, that is: _Pu - _Pr ? (t_Pul_Pu - t_Prl_Pr), where l are the mean of selectivity correction terms and t their parameter estimates using the extended wage function. The offered wage gap specified above has already been corrected for unobserved characteristics that are correlated with wages.

II. Estimation results on public private wage differentials

1. Estimates of selection model

Taking into account of selection bias, we see that there is a large effect of education and training on being a public worker (see appendix - table 1). The probability of being a public worker increases with levels of education. The picture is the same with training. Age of workers also affect the probability of being employed by the public sector. One additional year of age contributes one per cent to probability of being working for the public sector. It is worth noting that a female is more likely to be found in the public sector than a male. Other variables such as marital status, parental education and rural/urban residence do not have any effect on the selection choice because their coefficients are insignificant at 10 % level. In the next section, we will see effects of selectivity function on the public private wage gap.

2. Wage functions of public and private sectors

2.1. Private workers' wage equation

Results in appendix - table 2 only show significant variables affecting private wages. For the regression without selectivity correction, the estimates of returns to education, specific experience and training are positive. There is gender advantage for male in the private sector because return to a private male worker is positive. Returns to private worker who live in or near the South (called southern regions) are positive. However, return to father education and rural/urban residence is not significant at 10 percent. In the regression with selectivity correction, returns to education, specific experience and training are positive but smaller than they were in the regression without selectivity correction. Returns to gender, southern regions are little higher but the return to father education is negative. The estimate of constant is higher.

2.2. Public workers' wage equation

For the regression without selectivity correction, returns to public worker education, experience, training, gender, and urban are positive. Unlike negative return to father education in the private sector, return to father education of a public worker is positive. Returns to worker's location are different between public and private sectors. Public workers in such the regions as South Central Coast and Central highlands can earn 12 and 9 percent, respectively, lower than those in the four northern regions. On the contrary, public workers in South East and Mekong Delta are likely to receive 32 and 6 percent, respectively, higher than the benchmark group. In the regression with selectivity correction, returns to all factors excluding training and gender are reduced (see appendix - table 3).

2.3. Differences in returns to characteristics between public and private sectors

In the regression without selectivity correction, returns to public worker's characteristics are lower than that to private worker's. Returns to education, training, experience, southern regions, and male are higher in the private sector than in public sector. On the contrary, returns to father education and urban are higher in the public sector than in private sector. There is no environment gap.

Unlike the regression without selection correction, in the regression with selection correction, there is no difference in returns to classical human capital such as education, training, and specific experience. However, public return to worker's father education is larger than private return. Returns to southern public workers are lower than that to private workers. Further more, public sector also favors urban public workers because they can earn 16 percent higher  than their private counterparts or rural public workers. Notably, although there is no difference in returns to classical human capital, there is large difference in estimates of constants between public and private sectors.

III. Decomposition of wage differentials.

This section will show directly what factors and how much these factors cause wage differentials between public and private sectors basing on the above results. As stated, two types of decomposition (Oaxaca and Neumark) are applied to have a close picture of public private wage differentials. Besides, each type of decomposition is divided into two parts: regression without and with selectivity correction. Especially, Oaxaca decomposition is introduced with two different return structures: public and private wage structures. By so doing, we are able to take advantages of and get rid of disadvantages of each type of decomposition.

1. Oaxaca decomposition

1.1.  Oaxaca decomposition without selectivity correction

The Oaxaca decomposition without selectivity correction brings about the same total gap regardless of which wage structure used. The total gap is 0.0364 in favor of public sector. In Oaxaca decomposition with public returns, the characteristics gap is positive (0.2631). The return gap is negative and stands at minus 0.2267 (623 percent of the overall gap). The negative return and zero environment gaps make the discrimination present. It means that, having adjusted for differences in workers? characteristic, public workers receive 25 percent less than private workers. The discrimination against public workers still exists when we decompose the regression using private returns. Public workers earn 18 percent lower than their private counterparts. Therefore, from Oaxaca decomposition, it can be concluded that public workers are underpaid despite which wage structure used.

Table 1: Oaxaca decomposition of public - private wage differentials without selectivity correction

Source: Author?s estimates from VLSS98

1.2. Oaxaca decomposition with selectivity correction

In the public sector, there is a positive correlation between the unobservable characteristics of the public workers that affect both their choice to work in the public sector and their wages. On the other hand, in the private sector, there is a negative correlation (minus 0.1529) between the two. As stated, we should take both public positive and private negative correlation into consideration by subtracting them from the actual wage gap. Having taken into account of possible selectivity bias, the offered wage gap is negative (minus 0.1290) and will be used for analysis.

Negative offered wage gap means that public wages are lower than that of private sector. The negative offered wage gap is contributed from three components: negative return gap (minus 0.1873), positive characteristic gap (0.2346), and negative environment gap (minus 0.1799). The sum of return gap and environment gap is minus 0.3636. Therefore, regardless of other things, workers are worse off when working for the public sector because they have to receive 44 percent lower wages (due to discrimination) than those in the private sector. The status of public workers does not change when the private wage structure is used for decomposition.

Table 2: Oaxaca decomposition of public - private wage with selectivity correction

Source: Author?s estimates from VLSS98

2. Neumark decomposition of public private wage differentials

2.1. Neumark decomposition without selectivity correction

The overall wage gap between public and private sector is 0.0364. It is decomposed differently into four components: positive characteristic gap (0.1464), positive deviation of public returns 0.0238), negative deviation of private returns (-0.1339), and zero environment gap. Public workers still earn more but it is not the case when the discrimination term is taken into consideration. The sum of deviation in the public returns, deviation in the private returns, and environment gap is minus 0.1152. Therefore, there is a wage discrimination against public workers. Having adjusted for characteristic differences, public workers now get 12 percent lower than private ones do.

2.2. Neumark decomposition with selectivity correction

The total wage gap between public and private sectors in decomposition with selectivity correction is negative. Inferior situation of public workers is explained by four parts: positive characteristic gap (0.1464), negative deviation of public returns (minus 0.0549), negative deviation of private returns (minus 0.0406), and negative environment gap (minus 0.1799). Having extracted the characteristic gap from the total gap, public workers now earn 32 percent less than private workers do (due to discrimination).

In summary, the results from both decompositions (Oaxaca and Neumark) make clear the normal thinking that wages are very low in the public sector in comparison with private sector. Now it is clear that, in general, public wages were higher than private wages (14 per cent). But, having taken some worker and work characteristics into account by using econometric regressions and decompositions, public workers are under-paid in comparison with private ones. Although, the degree of public workers? disadvantages in term of payment vary according to which type of regression and decomposition used, all findings support the inferiority of public workers.

In Oaxaca decomposition without selectivity correction, the wage discrimination against public workers lower their wages by 25 and 18 percent in the case of decomposition with public and private returns, correspondingly. If selectivity correction terms are included, the discrimination causes public wages 44 and 19 percent lower than private wages (taking public and private wage structure, respectively).

In Neumark decomposition without selectivity correction, having adjusted for characteristic differences, public workers earn 12 percent lower than private ones do. If selectivity correction is included in the Neumark decomposition, having extracted the characteristic gap from the total gap, public workers now earn 32 percent less than private workers do.

Table 3: Neumark decomposition of public - private wage differentials

Source: Author?s estimates from VLSS98

It is interesting to contrast two types of decomposition: with and without selectivity correction. The wage gap between public and private sectors is lower if we decompose the wage differentials without selectivity correction. However, as stated in the theoretical part, the regression and hence the decomposition without selectivity correction may face the problem of biased estimates. The regression results also detected the presence of correlation between unobservable characteristics and wages. Therefore, it is preferable to use the findings from decomposition with selectivity correction.

Chapter IV: Moonlighting in Public sector

In this chapter we will test the hypothesis that the public - private wage gap indeed contributes to the phenomenon of moonlighting by government workers. If it does, we have additional evidence that public wages are indeed too low. Besides, some policy implications should be drawn to strengthen the public sector because low wages in public sector may result in moonlighting and hence reduce public service quality.

For the purpose of this study, moonlighting is defined as having a second income activity, i.e. in additional to the primary job. The second jobs are taken in forms of secondary paid work (the work which the respondent devotes the most time after his/her primary job), self - employment (agricultural and non - agricultural). Moonlighting is much more prevalent among civil servants than among wage earners in the private sector. 27 per cent of public sector employees have secondary job while 14 per cent of employees in the private sector do.

I. The moonlighting model

An important assumption is that individuals make their labour supply decisions sequentially. First, they try to obtain a public sector job, then, given the income earned in this sector, they decide on whether or not to take a second job. This is a very reasonable assumption, although there are numerous sequences depending upon relative wages in the main and second job, unobserved ?tastes? for the two jobs, and most important, perhaps, constraints on the choice of hours in the two jobs.

The simplest model to test for the effect of wages on the public and private sectors (Van Der Gaag et al., 1989) is given in the appendix - table 4. The table shows the estimation results of the probit equations in which the dependent variable equals one if a person has a second job, and zero otherwise. We expect, in a priori, that higher public wage will reduce the probability of moonlighting by public workers, while a higher private wage offer will make moonlighting activities more attractive.

Note that public workers? wages now are the predicted accepted wages received by the civil servants (basing on equation (3a)). The private wages offered to public employees are predicted by using equation (3b). The private wage offered is used as a proxy for wages potential in the private sector. We also add a number of variables such as household size, age, years of schooling, sex, and marital status to see whether these characteristics have additional direct effects on the probability of moonlighting.

The estimation results almost confirm our expectations. Firstly, if public wages (predicted accepted wages) rise, the probability of moonlighting decreases. An increase of public wages by one thousand VND is likely to reduce the probability of having a second job by 12 percent. Secondly, private wage offer does affect the probability of moonlighting since its estimated coefficient is significant statistically at 10 percent. An increase of private wage offered by one thousand VND raises the probability of having a second job by a public worker by five percent.

Education has no effect on moonlighting probability because their estimated coefficients are insignificant statistically at 10 percent. Household size has negative effect on probability of moonlighting.  One more member in the family could reduce the possibility of moonlighting by two percent. Public workers seem to have more secondary jobs when they get older. Married public workers tend to have more moonlighting activities. This can be easily understood, as married one is likely to have more pressures on earning money to support his/her family. Furthermore, male public workers are less likely to hold a second job (by three percent) than female ones.

Thus, the analysis of moonlighting in the public sector confirms the hypothesis that lower wages in the public sector is partly responsible for the moonlighting activities of the government workers. Since government workers are much more likely to have a secondary job than wage earners in the private sector, the result is consistent with our finding in the chapter 3: wages in the private sector exceed those in the public sector. Besides, potential private wage offered increases the attractiveness of having a second job.

II. Discussion

The above comparisons are based on monetary remuneration only. Although this wage measure includes the monetary value of such benefits as travel allowances, job training and food received at work, the value of other fringe benefits, such as paid holidays or pensions, is not included. As is well known, such benefits are usually more prevalent in the public than in the private sector (see appendix table 5).

If the public wages are indeed ?too low?, why did civil servants not quit their government job? There are several reasons. One is that full- time jobs in private sector are not available. A second explanation is that some people are unwilling to forgo direct monetary rewards for job security, other intangible for job characteristics and fringe benefits such as paid holidays, sick paid, and pension. This view is consistent with our data on non - wage benefits in public and private sector jobs.

Perhaps the most plausible explanation is that government workers can have double-benefits. That is, they can enjoy the security and other benefits of having a government job and at the same time supplement their income by having a second job. As have been shown, the probability of finding a civil servant who has a second job depends significantly on the wage in the public sector. Further erosion of public sector wages can be expected to result in more ?double jobbing? by civil servants.

The consequences of having underpaid government workers for internal efficiency in the government and the concomitant effects on the economy as a whole are particularly serious. The Industrialization and Modernization that Vietnam are pursuing call for better educated and highly motivated civil servants to promote productivity and to provide advice to policy makers in the design and implementation of policies. One cannot reasonably expect to find these characteristics in a work force that is badly underpaid.

Before turning to the conclusion, the thesis introduces recent Vietnam labour market (especially, minimum wage, income, and social insurance) policies are introduced as an explanation for the government?s role in creating the wage differentials.

Conclusion

I. Summary of main findings

The estimation results yield two strong conclusions. The first is the answer to the main question of interest: public workers in general are underpaid in comparison with private workers despite of which type of regression and decomposition used.

Second, there was a negative relationship between public wages and moonlighting activities. Offered wages in the private sector increases the probability of having a second job by public workers.

II. Summary of policy implications

Basing on the analyses of the public - private wage differentials and moonlighting, various policy recommendations have been offered to make the labour market work more efficiently. Main policy implications are as follows:

- Wage policies play an important role in motivating worker in labour market, therefore, government attentio