6. RESULTADOS
6.4. EVALUACIÓN ECONÓMICA FINANCIERA
6.4.1. Inversiones
Simultaneous-quantile regression analysis is used here to capture the impact of various factors that may affect labour income differently, depending on the portion of the earnings distribution that is examined. This technique allows a focus on the determinants of low earnings, as opposed to the determinants of mean earnings. Table 15 shows the results of the estimations of hourly earnings for the 25th, 50th, and 75th earning percentiles.
The basic estimates of the gender wage gap, measured by the coefficient of the female dummy variable, are all statistically significant, indicating that women face a clear disadvantage in terms of pay. The size of the gender earnings gap in low-paid jobs is about 14 percent. As regards ethnicity, there seems to be little difference between people from Central Asian origin, while Russian speakers tend to get higher earnings. Looking at the effect of age within a quartile, there does not seem to be correlated with the level of pay, although age is likely to affect which quartile an individual is in.
The data also show some evidence those individuals who recently migrated from other parts of Kyrgyzstan face a significant disadvantage in terms of pay that is particularly marked in low-paid activities. For instance, compared with individuals who never migrated, the earnings of new migrants – those who migrated less than 2 years before the date of interview - were on average 35 percent less in low-paid activities and 23 percent less in middle-paid activities. But the negative earnings effect associated with internal migration seems to disappear over time as migrants settle-down. More surprisingly, individuals with no residence permits tend to get higher earnings.
With respect to health status, the data show that poor health conditions are an important factor affecting negatively the earnings of low-paid individuals. But the health variable is no longer significant among high-paid workers. This indicates a greater vulnerability of low-paid workers to health risks.
The same Table 15 shows that the type of education has a great influence on the level of earnings, and this is even truer among low-paid workers. Compared with individuals with higher general education, individuals with primary or no education earn 58 percent less in low-paid activities, 47 percent less in middle- paid activities, and 41 percent less in well-paid activities. A higher negative premium is also associated with technical secondary education (-40 percent in low-paid activities), than with general secondary (-32 percent).
The returns to job tenure are concave in low-paid activities, but job tenure does not play a significant role in high-paid activities. Among low-paid workers,
there is an increase in earnings up to an average of 20 years of tenure, and a decline thereafter.
With respect to job characteristics, the data indicate that in low-paid activities, individuals engaging in agriculture, services, and trade earn significantly less than those engaging in all other sectors. Trade is however no longer associated with lower earnings in better-paid activities. The latter likely reflects the heterogeneity of this sector, which may range from low-paid street trading and small-scale informal activities to well-paid activities linked with exports. What is also interesting is that agriculture is the sector with the lowest return in terms of pay at all level of the earnings distribution.
The nature of employment has also a substantial impact in terms of pay. Working as a self-employed and in the private sector is associated with higher earnings. Moreover, the returns to private sector employment, relative to public employment, are much higher in high-paid jobs than in low-paid activities. In low-paid activities, working in the private sector is associated with a wage premium of only 10 percent, compared with 19 and 27 percent in middle-paid and high-paid activities. In other words, private sector employment is providing relatively higher income opportunities than public sector employment, but this is truer in high-paid activities. What is also remarkable is that informal employment is associated with lower-earnings in both low and middle-paid activities, but not in high-paid activities. This indicates that for most workers in Kyrgyzstan, the informal sector resemble more as a survival mechanism associated with lower earnings. For a few better-off individuals, however, it is no longer associated with lower earnings.
With respect to local labour market conditions, the data show that individuals in rural areas face a substantial disadvantage in terms of earnings that is particularly marked in low-paid activities. The rural-urban earnings gap stood at about 20 percent in low-paid activities, compared with 16 percent in high-paid activities. Another interesting result is that regional unemployment exerts a strong moderating impact on earnings at the top of the earnings distribution. This suggests that growing wage flexibility is taking place in the Kyrgyz labour market in high-paid activities.
To sum up, the previous results have shown that the determinants of earnings and the size of their impact are not necessarily identical in all portion of the earnings distribution, reflecting a form of segmentation in the labour market. In low-paid activities, having health problems, having migrated recently, or holding an informal job resulted in lower earnings, while these events did not seem particularly significant in high-paid activities. Another difference is that the returns to private sector employment were more pronounced in high-paid
activities. Yet, in both low-paid and high-paid activities, lower earnings are observed for women, workers with little or inadequate skills, workers in agriculture and services, and individuals living in rural areas.
Table 15: Simultaneous-quantile regression estimates of log hourly earnings
Explanatory variables 25th percentile (coeff.) 50th percentile (coeff.) 75th percentile (coeff.) Individual characteristics Male (reference) Female -0.152*** -0.156*** -0.144*** Age 16-25 0.031 -0.041 -0.038 Age 26-45 0.060 0.023 0.031 Age 46+ (reference) - - - Job tenure 0.024** 0.013** 0.009
Job tenure square(×100) -0.061* -0.019 -0.010 Kyrgyz (control) - - -
Russian, Ukrainian, or Byelorussian 0.158*** 0.185*** 0.191*** Uzbek or Tadjik -0.100 -0.185 -0.121
Kazakh 0.138 0.137 0.244
Tatar -0.212 0.095 0.116
Other -0.653*** -0.287*** -0.167 Never migrated (reference) - - -
Migrated less than 2 yrs ago -0.439*** -0.264** -0.038 Migrated between 2-5 yrs ago 0.050 0.082 0.144 Migrated more than 5 yrs ago 0.046 0.043 0.045
No propiska 0.372*** 0.195* 0.375***
Disability/illness -0.105** -0.076* 0.015 Primary education or less -0.877*** -0.644*** -0.528***
Less than secondary -0.450*** -0.472*** -0.387*** General secondary -0.397*** -0.326*** -0.283*** Technical secondary -0.509*** -0.423*** -0.382*** High technical -0.249*** -0.237*** -0.238*** High general (reference) - - -
Source: KPMS 1998.
Note: ***, ** and * means statistically significant at 1 percent, 5 percent and 10 percent levels respectively. The percentage effects of a dummy variable X in semi-logarithmic equations of the form log(Y)=aX+b d is exp(a)-1, not a.
Attacking the problem of low-earnings in Kyrgyzstan would therefore require a mix of regional policies aiming at developing rural areas, sectoral policies that increase labour productivity in low-paid industries like agriculture, public sector reforms to allow an increase in public wages, in particular in services, and measure that ensure equity in pay by gender. There is also a need to work-out sustainable measures to enhance labour demand in the formal sector.
Table 16: Continued Explanatory variables 25th percentile
(coeff.) 50th percentile (coeff.) 75th percentile (coeff.) Job characteristics Transport (control) - - - Agriculture and forestry -0.871*** -0.778*** -0.694***
Mining 0.016 0.410** 0.304*
Manufacturing -0.045 -0.073 -0.145 Electricity, gas, water -0.117 -0.164** -0.048
Construction 0.012 -0.032 0.116 Trade -0.226* -0.048 -0.064 Financial services 0.021 0.078 0.069 Services -0.291*** -0.246*** -0.202** Self-employed 0.458*** 0.427*** 0.695*** Private 0.095** 0.171*** 0.242*** Informal -0.124*** -0.099*** -0.064
Local labour market characteristics
Rural -0.235*** -0.172*** -0.177*** Regional unemployment rate 0.053 -0.139 -0.420***
N 2784 2784 2784
Pseudo R2 0.215 0.164 0.121 Source: KPMS 1998.
Note: ***, ** and * means statistically significant at 1 percent, 5 percent and 10 percent levels respectively. The percentage effects of a dummy variable X in semi-logarithmic equations of the form log(Y)=aX+b d is exp(a)-1, not a.