Lucas Ibarra

The results are obtained from a dataset with 22,086 observations of households in Kenya. The goal is to determine the effect of work experience on monthly earnings in USD of households in Kenya. In order to get a more well rounded outlook, I asses some aspects of internal validity of the study while observing a few factors that may influence work experience, and, therefore, monthly earnings.

tinytable_3hypxwwh771bxi85hdek
Table 1. Descriptives of earnings in Kenya
Mean Median SD
educ_yrs 9.81 10.00 4.06
age 33.00 32.00 9.15
month_interview 4.24 4.00 1.38
wealth_group 3.26 3.00 1.36
female 0.56 1.00 0.50
rural 0.56 1.00 0.50
earnings_usd 333.03 162.79 814.29
years_lived_elsewhere 28.22 27.00 10.40
christian_main 0.58 1.00 0.49
christian_evang 0.29 0.00 0.45
muslim 0.08 0.00 0.27
kikuyu 0.17 0.00 0.38

Table 1 is a summary of the descriptives from the dataset.

Figure 1 shows the relationship between monthly earnings in USD and work experience for an individual. Due to the extreme outliers that appear in the dataset, I have limited the monthly earnings to less than $750, so the effect of experience is slightly magnified. Nonetheless, there does appear to be a slight increase in earnings as years of experience approaches 10, but a gradual decline in earnings as work experience surpasses 10 years; suggesting any work experience past 10 years is detrimental to monthly earnings.

Figure 2 is a histogram density chart of years of work experience for those living in a rural area compared to those living in an urban area. There appears to be more individuals living in urban areas with around 10 years of experience than there are in rural areas. Conversely, there are more people in rural areas with 20+ years of experience. Additionally, there is a notably large number of individuals in both urban and rural areas with 13 years of experience.

Figure 3 tells us that more non-Muslim females have around 10 to 25 years of work experience than Muslim women do. However, more Muslim women have about 25 or more years of work experience than their non-Muslim counterparts. Furthermore, there is not much of a difference in work experience between Muslim males and non-Muslim males.

Figure 4 suggests a nonlinear relationship between years lived elsewhere and work experience such that there is a slight increase in experience as years lived where approached about 15, but then slightly decreases after 15 years until years lived where reaches 20, after which experience increases dramatically.

Figure 5 depicts two box plots. One compares the dispersion of education across wealth groups while the other compares the dispersion of work experience across wealth groups. The box plots show that the wealthier groups spend more time in school while the poorer classes spend more time working. These observations possibly suggest that wealth in Kenya is inherited, and, as a result, wealthier people have the opportunity and means to spend more time in school while lower class individuals must start working early either because they cannot afford further education or because they need to in order to support their families. These findings make logical sense as if one is in school then he/she cannot be spending as much time working.

Table 2. Regression table showing the effect of experience on monthly earnings in USD
(1) (2) (3) (4)
(Intercept) -175.872*** -175.872*** -583.960*** -284.760***
(15.447) (15.447) (33.071) (27.218)
educ_yrs 51.895*** 51.895*** 55.277*** 55.229***
(1.992) (1.992) (2.094) (2.095)
age 11.362***
(0.598)
years_lived_elsewhere 6.772***
(0.518)
kikuyu 28.157+
(15.962)
christian_main -123.470***
(15.073)
christian_evang -169.686***
(15.550)
Num.Obs. 22086 22086 22086 21974
R2 0.067 0.067 0.083 0.078
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001

Table 2 shows the regression results for a regression run on monthly earnings in USD with a base specification of work experience. Experience seems to have a statistically significant (p<.001) impact on monthly earnings regardless of control variables. However, the relationship between experience and monthly earnings may suffer from Omitted variable bias (OVB). When controlling for years of education, the coefficient on experience dramatically increases; more experience decreases earnings, so less experience would, naturally, increase earnings. Education implies less experience, so, therefore, more education implies an increases in monthly earnings.

At the same time, age is another OVB for the relationship between earnings and experience, which works in the opposite direction but with the same exact magnitude of effect. Each additional year of age corresponds to another year of experience, thereby contributing to the fall in monthly earnings.

Despite OVB we still see that experience has a statistically significant effect (p<.001) on earnings across all regressions with about an increase of about $4 per month per extra unit of experience gained when controlling for years of education, among other factors. At the same time, we see years of education has a statistically significant (p<.001) impact on earnings such that each additional year of education increases monthly earnings by about $30 across all regressions for which it is included, regardless of other control variables. However, we see earnings decrease by $25.84 when age is controlled for. Simultaneously, earnings increase by $30.35 for each additional year alive. Furthermore, years lived away from Kenya (p<.01) and being christian (p<.001) –– mainstream or evangelical –– both have a statistically significant impact on monthly earnings.