Intergenerational Mobility of Earnings and Income Among Sons and Daughters in Vietnam

In this paper, I investigate intergenerational mobility of earnings and income among sons and daughters in Vietnam. In particular, my objective is to estimate intergenerational elasticity (IGE) of sons’ and daughters’ individual earnings and individual income with respective to their fathers’ individual earnings. The two-sample two-stage least squares (TS2SLS) estimation is employed to achieve the research objective using two primary samples of father-son pairs and father-daughter pairs from Vietnam Household Living Standards Survey (VHLSS) of 2012 and one secondary sample from Vietnam Living Standards Survey (VLSS) of 1997-98. My results show that the baseline IGE estimates of Vietnamese sons are 0.361 and 0.394 for individual earnings and individual income, respectively. For Vietnamese daughters, the baseline IGE estimates are 0.284 and 0.333 for individual earnings and individual income, respectively. These IGE estimates explicitly reveal that Vietnam has the intermediate degrees of individual earnings and individual income mobility across generations for both sons and daughters by the international comparison.


Introduction
Inequality has increasingly been viewed as a stylized problem facing a modern state in the twenty-first century (Piketty, 2014a(Piketty, , 2014b(Piketty, , 2015. As social scientists and policy-makers have paid considerable attention to inequality, they have placed prominence to equality of opportunity in addition to how socioeconomic outcomes are equally distributed among social classes (Corak, 2013;Krueger, 2012). The extent to which a child's socio-economic status in the current generation is determined by his or her parents' socio-economic outcome in the previous generation probably gives an in-depth understanding of the degree of opportunity equality (Corak, 2013). This has been a very important motivation for massive academic investigations of intergenerational mobility that has been witnessed over last three decades (Black and Devereux, 2011;Solon, 1999).
Intergenerational mobility provides an exploration of the relationship between the parents' socio-economic status and that of their children as adults. This research topic has been investigated by both sociologist and economists (Blanden, 2013;Torche, 2015). The main difference in the approach to intergenerational mobility between sociologists and economists is how they define a measure of socio-economic status or outcome.
From sociologists' perspective, a proxy for socio-economic status is usually related to social class or social status. Among them, occupation is predominantly chosen as a main indicator for socio-economic status in sociology (Hout, 1988;Mazumder and Acosta, 2015). 1 In a different manner, when economists explore economic mobility across generations, they place a lot of emphasis on earnings and income as key indicators of socio-economic outcome or socio-economic success (Black and Devereux, 2011;Solon, 1999). 2 In this study, from an economic perspective, I examine the persistence of economic outcome between fathers and offspring, including both sons and daughters as adults, in Vietnam. For the measurement !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 1 In addition to occupation, education can be used as another measure of socio-economic status in 2 Other measures of economic status used in the literature include wealth (Asadullah, 2012;Charles and Hurst, 2003), and consumption expenditure (Aughinbaugh, 2000;Charles et al., 2014;Waldkirch et al., 2004).! of fathers' economic outcome, individual earnings is chosen. There are two different measures of economic status for children as adults, including individual earnings and individual income.
Vietnam has been characterized by increasing inequality aligned along recent economic reforms and achievements, and expansions of education that are typically distinguished for a transition economy (Haughton, 2001). Extensive research on economic inequality has been carried out in Vietnam (Adger, 1999;Nguyen et al., 2007;van de Walle and Gunewardena, 2001). However, most studies of economic inequality primarily focus on how economic outcome is distributed among social classes or social groups at a specific year or a period within one generation. Such inequality measuring using cross-sectional data, therefore, cannot reveal the transmission of inequality from generation to generation as well as the degree of opportunity equality in Vietnam. Studies of intergenerational mobility can importantly overcome this shortcoming. Within this context, Vietnam becomes an important case to investigate intergenerational mobility.
Research on intergenerational mobility in Vietnam is almost nonexistent. The paper from Emran and Shilpi (2011) is currently so far the only original paper of intergenerational mobility in Vietnam. However, Emran and Shilpi (2011) concentrate on intergenerational mobility of occupation from sociologists' view rather than from economists' perspective. There has been no empirical evidence on intergenerational economic mobility in Vietnam. Hence, this study significantly fulfills the gap.
Moreover, findings from this study are compared to results from other countries, especially for developing countries and Asian countries, in order to reveal whether or not Vietnamese society is relatively mobile.
In studies of intergenerational mobility, researchers' main objectives are to estimate intergenerational elasticity (IGE) or correlation (IGC) of earnings or income. In this study, I focus on the former estimate, IGE. IGE is a reasonable statistic that accounts for the degree of the intergenerational association between parents' economic resources and economic status of their children. In principle, a high IGE estimate explicitly provides an implication of a low degree of mobility with a measurable magnitude of intergenerationally perpetuated inequality. In other words, a poor child is less likely to escape poverty and move upwardly while the likelihood for a child who was born in a wealthy family to remain at the top position from the social ladder of economic outcome is comparatively high. In such a society with high IGE, the degree of equality of opportunity is relatively low. In contrast, a modest IGE estimate indicates a high level of economic mobility across generations, and therefore a high degree of the equality of opportunity.
In order to obtain IGE estimates, researchers ideally demand a representative sample in which information on the approximate permanent economic outcome for both parents and children as adults is available. Unfortunately, such datasets are rarely available, especially in developing countries including Vietnam. As a result, I cannot apply this approach to this country. In this !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 3 For previous intensive surveys, see Björklund and Jäntti (2009), Black and Devereux (2011), Blanden (2013), Corak (2006), and Solon (2002). study, in order to surmount the problem of lack of data, I use the two-sample two-stage least squares (TS2SLS) estimator to estimate IGEs between fathers and sons and between fathers and daughters. TS2SLS is first developed by Björklund and Jäntti (1997)  The remainder of this paper is organized as follows. Section 2 provides the research context that is connected to the present study. In section 3, data sources and samples are presented while section 4 discusses research methods.
Section 5 shows empirical results, and section 6 presents robustness checks.
Finally, conclusion is pointed out in section 7.  (Anwar and Nguyen, 2010;Ngoc, 2008;Nguyen and Xing, 2006;Vu, 2008). By 1999, the FDI sector has contributed about 13% to the total GDP growth and approximately 25% to the entire tax revenue (Freeman, 2002). Meanwhile, Vietnam's exporting activities have grown considerably during the age of renovation.
Economic policy reforms have promoted Vietnam to become one of the most remarkable emerging economies with highest economic growth rates in Southeast Asia (Irvin, 1995). By illustration, on average, the annual GDP growth rate of Vietnam was approximately 8.6% between 1991 and 1998 (Nghiep and Quy, 2000), and reached the apex of about 9.5% in 1985 (WDI, 2014). Moreover, Vietnam has successively retained its high rate of economic growth since the 1997 Asian financial crisis. Specifically, from 2000 and prior to the 2008 global financial crisis, the average annual economic growth rate of The Sixth National Congress of the Vietnamese Communist Party in December 1986 had launched a new plan for changing the economy from a centrally-planned to a market-oriented system (Thayer, 1987). Vietnam was 7.5%, which was higher than corresponding figures of the world economy, ASEAN, Asia Pacific, and India with 3.9%, 5.4%, 5.9%, and 7.3%, respectively (IMF, 2009). Therefore, the economic reforms have provided the positive impacts on economic growth of Vietnam's economy.
Economic growth has led to the amelioration in GDP per capita for Vietnamese compared to the period before the 1986 Đổi Mới (UNDP, 2011). In 1985, Vietnam was still one of poorest countries in the world with low GDP per capita (Thayer, 1987

Expansions of Education
During the era of renovation of Vietnam's economy, expansions of education have been witnessed in both demand and supply sides. Changes in the structure of the economy from an agriculture-based to an industry-and-service-based economy have increased demand for skilled workers in Vietnam (Cai and Liu, 2014). Moreover, the demand for education has also gone up because of wealth effects stemming from substantial income growth among Vietnamese households (Glewwe and Jacoby, 2004).
From the supply side, the provision of education has increased because of enlarged investments from the government and the private sector. Historically, Vietnam's education system has predominately been funded, managed and controlled by the state. In the new era of educational reform, the public budget for education investments has increased. For example, the budget was over 13% of GDP in 2010(GSO, 2011. In addition to developments of state-funded educational institutions, the private sector has increasingly contributed to the human capital in Vietnam (Ngo et al., 2006;Mok, 2008). This growth of nonpublic education has advanced Vietnamese citizens' accessibility to education (Goyette, 2012). For example, the enrollment rates for lower secondary and upper secondary schools had gone up from 66% to 72%, and from 23% to 31% between 199372%, and from 23% to 31% between and 199872%, and from 23% to 31% between , respectively (GSO, 1999).
An important contribution of education expansions to Vietnam's economy is to provide more educated workers for labor markets. For example, rates of workers with primary education qualification or non-diploma had gone down from 49% in 1993 to 51% in 200251% in and 44% in 200651% in (GSO, 199451% in , 200351% in , 2007. In contrast, rates of those hold tertiary qualifications had increased from 1.8% in 1993 to 3.3% in and 4.2% in 2006(GSO, 1994, 2003. The rates of workers who hold the secondary and high school qualifications had increased with 26% and 14% in 1993and 14% in , 30% and 16% in 2002and 14% in , and 33% and 19% in 2006and 14% in (GSO, 1994and 14% in , 2003and 14% in , 2007. Having more skilled workers has allowed the shift from the physical-capital-accumulation-based growth to the productivity-based one in Vietnam's economy (Saich et al., 2008;Welle-Strand et al., 2013;World Bank, 2013).
Also, the returns to schooling has ameliorated in Vietnam. For instance, wage has gone up from 4.2% in 1993 to 4.8% in 1998 for female workers (Liu, 2006), and from 2.9% in 1993 to 5% in 1998 for whole labor force (Gallup, 2004). 5 Increases in returns to human capital can lead to wage differentials, and income inequality when educated workers have more opportunities to improve their income due to their higher levels of education, especially in the private and nonfarming sectors (World Bank, 2013). Sakellariou and Fang (2014) implicitly reveal that labor market reforms along with the Đổi Mới have led to real wage growth, increases in labor earnings, and income inequality. In this context, inequality in outcomes from labor markets has been given massive attention from researchers as well as policy makers (Imbert, 2013). 6 Liu (2006) provides an inverse result, which is a decrease in returns to education from 5.9% in 1993 to 3.5% in 1998 for male workers, respectively. 6 Inequality in labor market outcomes can be accepted in some extent because it helps create and retain innovation and hard working motivations (Acemoglu and Robinson, 2013). However, if inequality comes from other factors such as institutional weaknesses, or parental positions rather workers' human capital or

Inequality
During the period of reforms and renovation, inequality in Vietnam has constantly risen although its magnitude is not at the apex in Asia (World Bank, 2014). Demonstratively, Gini indices increase from 33, 35.4 to 40.7 in the years of 1993of , 1998of , respectively (Fritzen, 2002. In 2012, Vietnam's Gini index equals 0.39, which is lower than that of China, Thailand, and Indonesia, and larger than that of India, and Cambodia (World Bank, 2014).
Inequality is a public concern because it can undermine harmonious growth, an important target of Vietnam's development (World Bank, 2014). Moreover, whether poor citizens are marginalized from the economic growth is of considerable concern to economists and social researchers (Fritzen, 2002).
There have been numerous studies into inequality in socio-economic outcome in Vietnam over last decade. From the sociological perspective, there are studies of inequality in education (Glewwe, 2004), and inequality in health (Granlund et al., 2010;Huong et al., 2006;Khe et al., 2004;Minh et al., 2003Minh et al., , 2006Wagstaff et al., 2003). Furthermore, inequality studies in the economic literature focus on economic outcome such as income (Liu, 2008;Milanovic, 1998), and consumption expenditures (Fesselmeyer and Le, 2010;Le and Booth, 2014).
However, most studies in inequality in Vietnam focus on inequality at one time point within a generation rather than the transmission of inequality from one generation to the next that shows the inequality of opportunity. Also, the inequality of opportunity is increasingly paid more attention in Vietnam (World Bank, 2014). However, there is currently no solid evidence on the inequality of opportunity in economic outcome in Vietnam. This study is the first study on intergenerational economic mobility that informatively provides the understanding of the inequality of opportunity in Vietnam.

Data Sources and Samples
efforts, it is probably disincentive. The inequality of opportunity is therefore a massive problem for a permanent innovative economy.

Data Sources
The two main sources of data used in this study are VLSS and VHLSS. The first source is from VLSS that was implemented between 1993 and 1998 by the General Statistical Office (GSO) of Vietnam as the main census of Vietnamese households before the year of 2000. 7 In the survey, households' socio-economic information, including education, employment, health, activities of agricultural production, activities of non-agricultural production, housing, migration, fertility, and savings and credit in each household is elicited (Haughton and Nguyen, 2010;World Bank, 2001). The secondary sample of "potential" fathers used in this study is extracted from VHLSS of -1998(GSO, 1999. The VLSS of 1997-98 has a representative sample with 6,000 households from representative communes 8 cross the country surveyed (World Bank, 2001).
The second source of data is from the VHLSS, which is the most important data source for basic socio-economic information of Vietnamese households since

Samples
One sample of male workers from the VLSS of 1997-98 and two samples of father-son pairs and father-daughter pairs from the VHLSS of 2012 are used to [Insert Table 1  For the secondary sample of "potential fathers," male workers whose ages vary from 31 to 51 are included. The size of this sample is 1041 observations.
In all three samples, individuals' information on socio-economic characteristics including education, occupation of employment, industry of employment, and geographic region are uniformly coded. In the case of education, there are five dummy variables, including (1) non-diploma or primary, (2) secondary, (3) vocational, (4) high school, and (5) (5) skilled manual workers, (6) semi-and un-skilled manual workers, and (7) farmers and farm workers in agricultural production.
Regarding the classification of industry of employment, there are ten categories, including (1) Table A1, Table A2, and Table   A3 of Appendices.
In empirical studies of intergenerational economic mobility, economists emphasize the source of measurement errors that result in lifecycle bias and attenuation bias. Referring to lifecycle bias, Haider and Solon (2006) show that when a child's economic outcome is not measured in long-run, a consequence is the measurement error which potentially generates lifecycle bias in IGE estimates. Specifically, if economic outcome is measured in the early or late ages of a child's working life, IGE results tend to be underestimated. They also suggest that when constructing a sample, including only children who are aged around the age of 40 is an appropriate choice because at around this age, a measure of economic outcome is the most suitable proxy for permanent outcome. As a consequence, the potential lifecycle bias is minimized.
However, due to the availability of data, I use a wider range of ages for both sons and daughters in primary samples in order to estimate baseline IGEs in this study rather than the age range as seen in Haider and Solon (2006). In particular, the primary sample of sons includes those aged between 25 and 54 while the age interval for the primary sample of daughters is from 25 to 47.
Moreover, most children in each sample are under the age of 30. Illustratively, there are 73.36% of sons aged from 25 to 30 while the corresponding figure for daughters is 77.85%. Therefore, the primary samples used in this study are relatively young.
Although Vietnam has been seen as a country with high rates of young labor force participation, a common census statistic shows that the proportion of young workers who are under the age of 30 was around 30% from 2007 to 2011 (GSO, 2012). In this study, with very large proportions of sons and daughters aged under the age of 30 compared to the common census statistic, it is important to concede that samples are probably not well representative of the population. Such samples with a large proportion of young workers used in this study can be explained by a fact that the available data source is limited to children who are co-residential with their fathers. The distribution of children's ages in the primary samples of sons and daughters are respectively demonstrated in Figure A1 and Figure A2 of Appendices. Hence, according to Haider and Solon (2006) I expect that with the available data, the baseline IGE estimates for full samples of sons and daughters in this study are downwardly biased in some extent.
If I use the age interval proposed by Haider and Solon (2006)  For attenuation bias, using a current or short-run measure of economic outcome of "potential" fathers in the secondary sample may result in a substantial underestimation of the true IGE estimates because the temporary economic outcome is potentially a noisy proxy for long-run one (Solon, 1992;Zimmerman 1992). This measurement error leads to attenuation bias that results in a downwardly biased IGE estimate (Solon 1992;Zimmerman 1992).
In this study, I use the TS2SLS estimator as a useful approach to measurement errors stemming from using a one-year measure of individual earnings as the proxy for fathers' economic outcome. This is because when the transitory shocks are not correlated with the predictors of fathers' individual earnings, estimates from the TS2SLS estimator are consistent (Inoue and Solon, 2010).
When comparing the distributions of fathers' socio-economic characteristics between the primary and secondary samples as shown in Table 1

Two-Sample Two-Stage Least Squares Estimation
Basically in many studies of intergenerational mobility, IGE is typically estimated from the following regression: where is the log of the i th children's permanent economic outcome, correspondingly denotes the log of their father's long-run economic outcome, and is an error term. In this study, economic outcome for children is measured by two different variables including individual earnings and individual income. For fathers, the proxy of economic outcome is their personal earnings from the labor market.
The coefficient 1 in equation (1) is the parameter of interest. The coefficient 1 is the measure of IGE, and then (1-1 ) measures intergenerational economic mobility. If information on lifetime economic outcome for both children and fathers is available, ordinary least squares (OLS) estimator can be applied to consistently estimate 1 . However, even a proxy for lifetime economic outcome such as multiple years of averages is rarely available, and it is especially true for datasets from developing countries like Vietnam. In many available datasets, only information on children's economic outcome ( ) is reported while information on parents' economic outcome ( ) such as earnings or income is commonly not recorded. Fortunately, information on parents' socio-economic characteristics such as education, occupation, and industry which is used to predict fathers' economic outcome is available.
In this paper, the problem of unavailable data is overcome by applying twosample two-stage least squares (TS2SLS) estimation for Vietnamese data. The TS2SLS estimator is based on the idea of the two-sample instrumental variable (TSIV) estimator invented by Angrist and Krueger (1992). Inoue and Solon (2010) show that in the two-sample environment, TS2SLS is asymptotically more efficient than TSIV.
The first application of the TS2SLS estimation is based on Swedish and American data in Björklund and Jäntti's (1997) paper. Since then numerous studies have used TS2SLS to investigate intergenerational mobility such as Fortin and Lefebvre (1998) for Canada, Lefranc and Trannoy (2005) for France, Dunn (2007) for Brazil, Gong et al. (2012) for urban China, Piraino (2015) for South Africa, Lefranc et al. (2014) for Japan, or Cervini-Plá (2014) for Spain.
In this study, TS2SLS is used to estimate 1 in (1). In doing so, two different samples and two stages are used to obtain 1 . The first is the primary sample.
In this sample, information on children's long-run economic status is available.
However, information on fathers' permanent economic outcome is not available. Male workers in this sample are the potential fathers for individuals in the primary sample, and variables of these male workers are employed to predict the missing economic outcome of fathers in the primary sample.
Regarding the two stages to achieve the IGE, the first is to predict the missing values of fathers' permanent economic status in the primary sample. To do this, firstly it is necessary to estimate the relationship, , between fathers' socioeconomic status and fathers' current economic outcome ( ) using the secondary sample.
In the primary sample, missing values of the logs of fathers' permanent earnings or income are calculated by the following equation: where represents fathers' predicted economic outcome, and is the corresponding coefficients of estimated in the first stage.
The second stage is to estimate the IGE for fathers and children using the primary sample. In other words, children's economic outcome is regressed on the imputed economic outcome of fathers. In summary, 1 which is estimated by the TS2SLS estimator is the IGE of children's economic status with respect to their fathers' economic success in this study.

Transition Mobility Matrix Approach
The transition matrix approach is a complementary method to the least squares regression approach, and it is also useful to examine the pattern of There are two benchmark cases for mobility including "perfectly mobile" and "zero mobile." Perfect mobility is the case in which the father's economic success does not affect the child's economic outcome at all. In this case, therefore, there is a 25% possibility for a child to be in any quartile regardless of his or her father's position from the distribution of economic outcome. In contrast, there is no chance for a child to change the position given his or her father's position from the distribution of economic outcome in the benchmark of zero mobility. In this case, the transition matrix becomes an identity matrix with all elements of 1 in the main diagonal and 0 elsewhere. Although it is rare to have such extreme cases in reality, these two cases can be used as two bounds in order to recognize the degree of mobility of earnings or income in an empirical study.
This approach is applied in some previous studies. For example, Dearden et al. (1997)

First-Stage Results
The secondary sample of potential fathers from VLSS 1997-98 that consists of 1041 male workers aged from 31 to 54 and is used to predict missing information on true fathers' individual earnings in the primary samples.
Potential fathers' average age in this secondary sample is 39.97. The rationale for the choice of this age range is based on an age-range-to-minimize-lifecyclebias suggestion from Haider and Solon (2006).
In the first stage, the log of potential fathers' individual earnings is regressed on age, age squared, and dummy variables for education, occupation, industry, and geographic region. The analysis focuses on the estimates for these socioeconomic characteristics because these are parameters of interest in the firststage model. The results from the preferred specification are presented in Table   2. Accordingly, the model has a R 2 of 0.186, which suggests that nearly 20% of the variation in the log of individual earnings of potential fathers can be explained by these socio-economic characteristics in this model. In Table 2, it can be seen that education and geographic region have larger variations on male workers' individual earnings rather than occupation and industry. This can be explained by increases in earnings differentials along with increased returns to education (Imbert, 2013;Liu, 2006), and increased gaps of earnings among different geographic regions (van de Walle and Gunewardena, 2001;World Bank, 2014) in Vietnam over two last decades.
Note that, age and age-squared are included in independent variables in the first-stage model. However, estimated coefficients for age and age-squared individual earnings. In all cases, the independent variables include children's age, children's age-squared, and the log of father's individual earnings imputed from the first stage.

Baseline Intergenerational Elasticity for Sons
The baseline IGE estimates for sons are reported for two different cases of using sons' different economic outcomes as dependent variables and presented in Table 3. The sample size for these estimates is 1344 individuals.
[Insert Table 3 here] In Table 3, it can be seen that the baseline IGE estimates are all statistically significant at the level of 1%. In Column 1 of The results also indicate that the baseline IGE estimate for individual income is higher than that for individual earnings. This is because a son's individual These IGE results are lower than those in other countries such as South Africa with an estimate of 0.62 (Piraino, 2015), Brazil with an estimate of 0.60 (Ferreira and Veloso, 2006), urban China with an estimate of 0.63 (Gong et al., 2012), Chile with an estimate of 0.57 (Núñez and Miranda, 2010), and Italy with an estimate of 0.50 (Mocetti, 2007;Piraino, 2007).
Of course, many other countries are more mobile relatively compared to Vietnamese society when IGE the estimates for sons are considered. For example, Björklund and Jäntti (1997) find an estimate of 0.28 for Sweden.

Transition Mobility Matrix for Sons
Next, I analyze the mobility patterns across generations from economic outcome distributions for sons. There are two cases of quartile transition matrix for two measures of economic outcome. In all these cases, the proxy for fathers' economic outcome is personal earnings which is predicted in the first stage using fathers' education, occupation, industry, and geographic region.  [Insert Table 4 here] The pattern is the same for individual income. The result is presented in Table   A4 of Appendices.

Baseline Intergenerational Elasticity for Daughters
Similar to sons' results, there are two cases for estimating the baseline IGEs for daughters which are corresponding to two measures for daughters' economic outcome, including individual earnings and individual income. Note that for these two cases, the unique measure for father's economic outcome is personal earnings.

Transition Mobility Matrix for Daughters
Regarding the transition mobility matrix for daughters, Table 6 presents the changing mobility patterns of daughters' position on individual earnings compared to their fathers' individual earnings. In general, the transition matrix for individual earnings mobility for daughters is relatively symmetric, and it is comparatively similar to that of sons. In addition to the IGE results, this result of the transition matrix also provides evidence on the finding of the modest difference of degree of mobility across generation between sons and daughters.
[Insert Table 6  The result of the transition mobility for individual income is the same as that for individual earnings. The result is presented in Table A6 and Table A7 in Appendices.

Robustness Checks
Having presented the main IGE results of individual earnings and individual income of sons and daughters with respect to their fathers' individual earnings in section 5, I now analyze the robustness of these IGE estimates along two dimensions. Firstly, the sensitivity of the IGE estimates of full samples of sons and daughters to the various first-stage model specifications is examined.
Secondly, the sensitivity of the IGE estimates to the different age ranges of children in primary samples is specifically checked.

Model Specifications
As

Analysis for Sons
The full sample which consists of 1344 sons aged from 25 to 54 is used to estimate the IGEs. (occupation and region) and is 9.14% higher than the baseline estimate. The gap between these two estimates is about 0.115.
When using an individual predictor in the first stage model as shown in case 1, 2, 3, and 4, it can be seen that the estimator with education (case 1) produces the largest IGE with an estimate of 0.400 while that with geographic region (case 4) creates the smallest IGE estimate with a degree of 0.315. However, the gap between these two extreme IGE estimates is relatively small with a degree of 0.085.

Analysis for Daughters
For daughters, the full sample comprises 632 individuals aged from 25 to 47. In  Table 8.
Firstly, Column 1 of Hence, these IGE estimates are higher or smaller than the baseline one with a maximum proportion of 43.24% or 18.02%, respectively. The absolute gap between these upper-and lower-bounds is 0.204. This gap is larger than that for individual earnings with a difference of 0.169. Also, it is larger than the corresponding figure for sons with a gap of 0.115.
In terms of using individual predictor, on the IGE estimate in Column 2 of Table 8, the estimator with occupation (case 2) produces the largest IGE with a degree of 0.433 while the estimator with education (case 1) yields the smallest IGE estimate of 0.273. This reveals an opposing result for sons' individual income where the estimator with education (case 1) produces the largest IGE and the estimate with geographic region (case 4) is the smallest one.

Robustness Checks of IGE Estimates to Different Age Ranges
From the existing literature, changes in children's age ranges in the primary sample may lead to the variation of the IGE estimates (Grawe, 2006;Haider and Solon, 2006). In this section, the sensitivity of the IGE estimates to different sub-samples of various age intervals is analyzed for both sons and daughters. Also, the results are compared to the baseline IGE estimates in section 5. [Insert Table 9 here]

Analysis for Sons
The results explicitly indicate that there are considerable variations of IGE estimates across sub-samples with different age intervals of sons. In Column 1 of   [Insert Table 10  intergenerational mobility as shown in Black andDevereux (2011), andBlanden (2013). Comparatively, these results show that Vietnam has the same mobile position as Japan (Lefranc et al., 2014), Taiwan (Kan et al., 2015), and South

Analysis for Daughters
Korea (Kim, 2013) in Asia. Furthermore, the results indicate that Vietnam is more mobile than other developing countries such as Brazil (Dunn, 2007), or South Africa (Hertz, 2001;Piraino, 2015).
Yet, it is necessarily wary in interpreting the IGE results because of some limitations from data facing this study. One of the shortcomings is the use of the small samples at one point in time. Hence, the IGE estimates do not demonstrate the long-run trends of intergenerational mobility in Vietnam.
Understanding the long-run trends of intergenerational mobility can potentially result in an in-depth comprehension of the fundamental mechanisms of the transmission of economic outcome from generation to generation (Aaronson and Mazumder, 2008;Lee and Solon, 2006). Moreover, small samples potentially provide less reliable estimates, especially for estimating IGEs in specific regions or age groups.
In addition, the samples in this study include children and fathers who live together within families. Consequentially, the downwardly bias estimated results of IGE potentially suffer from this shortcoming of data. Furthermore, Haider and Solon (2006) show that IGE estimates tend to be downwardly biased if offspring's economic outcome is measured at young or old ages. The South Korea, Lefranc et al. (2014) for Japan, and Piraino (2015) for South Africa.
In addition, in this study I find that there are variations of the IGE estimates across sub-samples of children with different age ranges. Specifically, individuals from older groups tend in general to have larger IGEs than younger groups.
These empirical findings in Vietnam again consolidate the conventional pattern of age effects on the IGE estimates in the existing literature.
In Vietnam, along with positive achievements of economic growth for a typical transition economy, the rise of inequality is viewed as a massive problem facing this emerging economy. The literature on inequality in Vietnam has been abundantly conducted over the last decades. However, almost all previous studies on inequality are conducted within a generation. Measures of inequality from cross-sectional data traditionally give "snap-shots" at a moment in the timeline, and thus do not provide information on the transmission of inequality overtime. Instead, measures of intergenerational elasticity of economic outcomes provide a "dynamic picture" of inequality from one generation to the next which is investigated in this thesis.  Note: * significant at 10%, ** significant at 5%, *** significant at 1%. Omitted variables: (1) non-diploma or primary in the education group; (7) farmers, and farm workers in the occupation group; (9) mining in the industry group; and (4) Central Highlands (CH) in the geographic region group.   Note: * significant at 10%, ** significant at 5%, *** significant at 1%. Bootstrapping standard errors (with 1000 replications) are in parentheses. Father's individual earnings is predicted by education, occupation, industry, and geographic region.    Note: * significant at 10%, ** significant at 5%, *** significant at 1%. Bootstrapping standard errors (with 1000 replications) are in parentheses. Father's individual earnings is predicted by education, occupation, industry, and geographic region. Note: * significant at 10%, ** significant at 5%, *** significant at 1%. Bootstrapping standard errors (with 1000 replications) are in parentheses. Fathers' individual earnings is predicted by education, occupation, industry, and geographic region. Figure     • Cleaners and domestic helps • Low-skilled workers in agriculture, forestry and fisheries

APPENDICES
• Workers in mining, construction, industry, and transport • Assistants in food preparation • Street-based and sales-related workers • Waste collectors and other low-skilled workers (1) agriculture • Agriculture and related services (crop production, husbandry, and agricultural services) • Forestry and related services • Aquaculture production and exploitation (2) manufacturing • Foodstuff production and processing • Beverages production • Production of cigarette products • Textiles • Costume production • Production of leather and related products • Wood-processing and making of wood and bamboo products (except beds, wardrobes, desks, chairs); making products from straw and plaiting materials • Producing paper and paper-based products • Printing and reproduction of recorded media • Production of coke coal and refined oil products • Production of chemicals and chemical products • Production of medicines, pharmaceutical chemicals and materials • Manufacturing of rubber and plastic products • Manufacturing of products from other non-metallic minerals • Production of metals • Manufacturing of products from cast metal (except machines and equipment) • Manufacturing of electronic products, PCs and optical products