Employment Structure and Returns to Skill in Vietnam: Estimates Using the Labor Force Survey

This paper uses Labor Force Survey data to assess key aspects of the labor market in Vietnam over 2007-14. The analysis finds large growth in wage employment in the foreign-owned and domestic private sectors. However, the state sector remains a major employer, particularly for workers with higher education, employing 70 percent of wage workers with a university degree. Low-skilled occupations dominate the stock of existing jobs, but the top growing occupations overwhelmingly belong to high-skilled categories. The paper notes that the high unemployment rates of recent university graduates, which have raised concern about a mismatch between skills and employer needs, reflect the transition to the job market and diminish sharply as graduates age. The returns to education in the private sector are highest for university graduates. Finally, women and ethnic minorities are less likely to work in wage jobs, and those that do earn lower wages, although the wage gap for women has declined over time.

This paper uses Labor Force Survey data to assess key aspects of the labor market in Vietnam over 2007-14. The analysis finds large growth in wage employment in the foreign-owned and domestic private sectors. However, the state sector remains a major employer, particularly for workers with higher education, employing 70 percent of wage workers with a university degree. Low-skilled occupations dominate the stock of existing jobs, but the top growing occupations overwhelmingly belong to high-skilled categories. The paper notes that the high unemployment rates of recent university graduates, which have raised concern about a mismatch between skills and employer needs, reflect the transition to the job market and diminish sharply as graduates age. The returns to education in the private sector are highest for university graduates. Finally, women and ethnic minorities are less likely to work in wage jobs, and those that do earn lower wages, although the wage gap for women has declined over time.

Introduction
Vietnam is in the midst of an ongoing structural transformation that continues to propel economic growth and lift citizens out of poverty. While growth during the 1990s in Vietnam was driven by productivity growth in agriculture, following the decollectivization of farms and the creation of tradable land use rights, since 2000 a principal factor has been the shift of workers out of agriculture into manufacturing and other higher productivity sectors, which has involved an expansion of the portion of the population working in wage jobs.
This transformation has taken place simultaneously with a large increase in educational attainment. One measure of the value of education is the private returns observed for workers in wage jobs. The trajectory of the returns to education over time during structural transformation is uncertain because while the demand for skills is almost certainly increasing, the supply of educated workers is increasing at the same time. In addition to the supply of educated workers and the demand for skills in the private sector, the observed returns to education are determined by the "quality" of the education system-in terms of its effectiveness in equipping students with skills-and the premium paid to educated workers in the public sector, which employs a large share of such workers in Vietnam.
The principal contribution of this paper is a new time series of the returns to education over the period 2007-2015, using Vietnam's Labor Force Survey (LFS), which provides for a more finegrained analysis than previous work on the topic. By way of background, we also sketch a profile of the labor market in 2015 and changes in its key features over this period. We also examine the differences in wages and employment between men and women and between ethnic minorities and members of the ethic majority.
The Vietnamese LFS was first introduced in 2007 and has evolved substantially over time with modifications in the sample, the survey instrument, and the variables collected. By carefully reviewing the documentation available for each round of the LFS, we construct a time series of comparable labor market variables. The resulting analysis is the first using LFS data over such a long time series. A secondary contribution of this paper is documentation of the changes in the LFS.
The paper is structured as follows. Section 2 reviews previous work on the returns to education and gender and ethnic wage differentials in Vietnam. Section 3 describes the Labor Force Surveys and outlines the steps that were taken to standardize the data to make it comparable for analysis over time. Section 4 provides a description of the profile of the labor market and its evolution over the period 2007-2014. Section 5 presents an analysis of the returns to education in wage employment as well as the differences in wages by gender and ethnic minority status. Section 6 concludes. The analytical findings from the paper are presented via graphical summaries and key tables in the main text. More complete results can be found in a series of tables in the annex. 3

Returns to Education
Previous research on the returns to education in Vietnam has examined major trends in rates of return during the country's transition from state ownership to greater market orientation, policies influencing those trends, and inequality between the public and private sectors. While there are differences in the methodologies and data employed, studies consistently have found that wages increased rapidly during Vietnam's transition and that returns to education in Vietnam are low but have increased over time.
Using Vietnam Living Standards (VLSS) data, Gallup (2002) finds a very low rate of return to a year of schooling in the 1990s, with an increase from 2 percent in 1993 to 4-5 percent in 1998. This study shows that the lowest rates of return to schooling were in the regions that have the lowest education levels, while the highest rates of return are found in the regions that have the highest levels of schooling. Liu (2006), also using 1993 and 1998 VLSS data, analyzes the changes in the wages of men and women with different education levels during Vietnam's transition. The author concludes that earning differentials between workers of different education groups can be explained by the increase in the relative demand for better-educated individuals.
Doan, Tuyen, and Quan (2016) use eight rounds of the Vietnam Household Living Standards Surveys and find an increase in returns from 1998 to 2008 and a significant decline since 2008. The authors suggest that the lowering of returns to education could be explained by an expansion of higher education and lower economic growth. They question if the higher-educated labor force is oversupplied or if there is a large distortion in the labor market and/or mismatching in the labor market and outdated skills in training. 2 Several studies have mentioned the significant wage gap between the state and non-state sectors. Imbert (2010) notes the rise in the relative average earnings of state sector workers from 1993 to 2006 and finds that the rise in the state sector wage premium cannot be explained by a change in worker selection into the sector. Phan and Coxhead (2011) suggest that the greater increase in skill premium for state workers is related to the privileged position that the state sector has with respect to trade, access to capital and regulatory treatment. Phan and Coxhead (2013) find that capital and labor market segmentation creates a two-track market for skills, in which state sector workers earn higher salaries than non-state workers. Tien (2014) argues that the most educated part of the workforce is attracted to the public sector while the majority of workers with lower levels of education end up taking low-end jobs in the private sector.

Labor Market Outcomes for Women and Ethnic Minorities
Several studies of the returns to education in Vietnam also examine gender wage differentials. Gallup (2002) finds lower rates of return for women compared to men, with a shrinking of the difference between 1993 and 1998. Liu (2005) also finds lower returns for women than for men in 1993 and 1998, with a larger gender wage gap in the private sector than the government sector. Pham and Reilly (2007) investigate gender wage disparities for wages in Vietnam over the period 1993 to 2002 using mean and quantile regression analysis and find that the gender wage gap fell by half over that period. Tien (2014) finds a gender wage gap in most economic sectors and notes that women are mainly involved in low-value manufacturing while men are more likely to work in medium-value manufacturing. Young women are often hired for assembly-line work in the footwear and garment industries in Vietnam. In general, these jobs offer lower wage rates, involve longer working hours, and insecure job tenure (Baulch, Dat and Thang, 2012).
Several studies examine the gap in living standards between the ethnic minority and other households in Vietnam (e.g., Baulch et al., 2007;Baulch et al. 2010;Turner, 2011;Cuong, 2012). For example, Cuong (2012) documents differences in farm income and non-farm income per working hour among those living in the Northern Mountains area.

Data
The Labor Force Survey is conducted by the General Statistics Office (GSO) of Vietnam. In 2014 the LFS was conducted on a monthly basis and covered a sample of 16,880 households per month. The LFS collects information about demographic characteristics and main activities related to the labor market for individuals aged 15 and above. The LFS is statistically representative at the regional level by quarter and at the level of provinces by year (GSO, 2015).
Over the period 2007-2014, several changes were introduced in the LFS questionnaire, together with updates in concepts and definitions used. As a result of this, a careful standardization process was needed in order to compare labor market outcomes over the 2007-2014 period. The standardization work included both 1) combination of information collected via differently worded questions in different years in order to create comparable labor market indicators, and 2) combination of different response options within a question in order to create comparable classifications over time. Annex 1 describes in detail the steps taken to standardize the data.

Structure of the Labor Market
This section highlights critical features of Vietnam's labor market, beginning with the overall distribution of employment, followed by various breakdowns. 5 The share of the population that is of working age, the labor force participation rate, and the employment rate are all quite high in Vietnam. Figure 1 presents an overall portrait of the employment picture in Vietnam as of 2014. 3 Due to a rapid decline in fertility in the 1980s and 1990s, the share of the population that is of working age is now at a peak. Of a population of 91 million, 62 million (68 percent of the population) are between 15 and 64 years of age. The labor force participation rate is substantial: 8 of 10 of those of working age (52 million) reported having worked in the 30 days before the time of the survey. 4 The unemployment rate-calculated as per the International Labor Organization definition-is extremely low. Just 1 million people (2 percent of those in the labor force) are classified as unemployed.
Following two decades of rapid structural transformation, the profile of employment is complex. A simple breakdown into three categories provides a birds-eye view. Among the employed population of working age, more than 4 out of 10 workers (21 million total) are in farming, which overwhelmingly consists of smallholder agriculture. An additional 21% work in non-farm selfemployment, while 37% hold wage jobs. Breaking down wage work further, various forms of classification can be used to determine whether a wage job is "formal." One simple measure is whether the worker holds some form of employment contract. 5 Wage workers with a contract make up 23% of the workforce while 15% hold wage jobs without contracts. These four major categories-farming, non-farm self-employment, wage with contract, and wage without contract-are used in the following discussion to provide a profile of Vietnamese workers. 6 6

Figure 1: Distribution of Population in Vietnam
Source: Analysis of 2014 Labor Force Survey Note: The total number of employed individuals (50,6 million) is larger than the sum of the people employed in the underlying subsectors (50.4 million). This is due to the fact that for 0.13 million individuals, there was not enough information to classify the sector. Similarly, the total number of wage earners (18,9 million) is higher than the sum of the underlying categories (18,7 million) because of missing information. Contrarily, the number of wage earners with contract (11.3 million) is larger than the sum in the underlying categories due to rounding. The definition of the main labor market aggregates, i.e. out of the labor force, employed, unemployed, follows the ILO definition. A detailed discussion on the concepts used in the LFS can be found in the "Report on Labour Force Survey 2014", published by GSO. A description of the concepts introduced in this paper, i.e. farming, non-farm self-employed, and wage workers can be found in Annex 2. As in other fast-growing countries in East Asia, concerns have been raised about apparently high rates of unemployment among fresh graduates in Vietnam. The employment rate for university degree holders aged 20-23 was 27% in 2014. However, Figure 2 shows that by tracking age cohorts over time, we see that these high rates of unemployment do not persist. Unemployment rates were also high (20%) for university degree holders between 20 and 23 years of age in 2010. This group corresponds to those born between 1987 and 1990. The unemployment rate for this same cohort of university degree holders in 2014-when they were age 24-27-dropped to 9%. Likewise, for the cohort of university degree holders who were 24-27 in 2010, the unemployment rate dropped from 6% to 2% between 2010 and 2014. 7 This finding suggests that rather than facing major enduring unemployment, many graduates experience a search period during the transition from school to work. A breakdown using the four major categories by gender, urban/rural location, ethnic group, and major region provides a coarse profile of employment (Table 1). Women are slightly more likely to work in farming, but also slightly more likely to hold wage jobs with a contract, reflecting the substantial number of women who are employed in public sector jobs and manufacturing. The breakdown also shows the diversity of the rural economy, with large numbers of workers in wage jobs and non-farm work. Farming is overwhelmingly dominant among ethnic minorities as well as in the two regions-the Northern Mountains and the Central Highlands-where they are concentrated. Finally, rates of wage employment with a contract are high in the Southeast region 7 Note that as with any quasi-panel analysis, the cohort groups are not identical in 2010 and 2014 due to entrance and exit. In particular, over this period, new members entered the group by obtaining university degrees or migrating from abroad. Others departed the group due to out-migration or death.  1987-1990Born 1983-1986Born 1979-1982Born 1975-1978 In 2010 In 2014 8 (including Ho Chi Minh City and nearby industrial areas) and the Red River Delta, which encompasses the industrial areas around the cities of Hanoi and Hai Phong. There is a marked pattern of employment by age, reflecting both life-cycle patterns and shifts by cohort ( Figure 3a). Most workers under age 20 and over age 50 are in agriculture. The predominance of farming among older workers may reflect shifting patterns by cohort. Older workers who entered employment before the economy's structural transformation was initiated have worked in farming all their lives and are still principally employed in agriculture. Wage jobs with a contract are most common among Vietnamese in their 20s. Non-farm self-employment is uncommon among fresh entrants in the labor market, but accounts for a larger share of the workforce after age 30. This trend could reflect workers' choice to exit the wage sector once they have accumulated enough human, social, and financial capital to start their own business. Work in the wage sector without a contract is most common for young workers. Workers with higher levels of education are much less likely to work in farming and much more likely to hold wage jobs (Figure 3b). Almost 90% of people with a university degree or a higher title hold a wage job with a contract, while just 1 in 4 Vietnamese workers with completed upper secondary education are in the wage sector with a contract. Non-farm self-employment is more common among those with intermediate levels of education, and the likelihood of having a wage without a contract declines with education.
The quality of employment varies across employment sectors. Individuals employed in farming work fewer hours per week (35 vs. 48 in the other three sectors) and almost never have written contracts (0.4%), health insurance (0.1%), or social insurance (0.1%). Only 3.6% of the individuals employed in farming report to be working less than 40 hours per week and willing to work more ( Table A6 in the annex). Many (35%) workers with a primary job in agriculture hold secondary jobs ( Figure 4). In the large majority of cases (87%), their secondary jobs are also in agriculture. Unlike in other sectors, the majority of the workforce in farming has been employed in the same job for over 10 years. 8 The educational categories "Trade Vocational school" and "Vocational School" are not necessarily presented in a sequential order, as these professional certifications can be obtained at different stages of academic education.  On the opposite side of the job quality spectrum, the wage sector with a contract is associated with benefits such as health and social insurance (86% and 84%, respectively) and virtually no underemployment (0.2%). While secondary jobs are not common in non-farm self-employment and in the wage sector, most of the people who do have a second job in these sectors have secondary employment in agriculture. While offering in general better job opportunities, there is some heterogeneity in the remuneration received by different population subgroups in the wage sector with a contract. Not surprisingly, hourly wages tend to be higher for more experienced workers, with the exception of the 60-64 age cohort, for which wages seem to be lower. The evidence presented so far would suggest that this could be the result of a high concentration of elderly workers in low-skilled occupations. Men tend to have higher salaries than women, while a first look would suggest that the same is not true for ethnic minorities as opposed to ethnic Kinh. Jobs in the public sector tend to offer high salaries, similar to jobs in high-skilled services. On the other hand, in manufacturing, low-skilled services and trade and hospitality, wages tend to be the lowest. The differences in wages across geography are likely to reflect the diverse sectoral structure in different areas of the country.

Evolution of Labor Market Structure
This section highlights changes in employment patterns over the 2007-2014 period. Overall, this was a period of remarkable job growth. Concurrently with an increase by almost 5 million individuals in the total working age population, the share of individuals in employment increased from 78.4% in 2007 to 81.7% in 2014. This trend was accompanied by a decrease of both the share of the unemployed (from 2.0 % to 1.6%) and of those out of the labor force (from 19.6% to 16.7%). At the same time, rates of employment and labor force participation increased, the share of wage jobs grew dramatically, from 32% of total employment in 2007 to 38% in 2014 ( Figure 7). This increase happened in parallel to a decline in both farming (from 45% to 41%) and in non-farm self-employment (from 24% to 21%). A closer look at the evolution of wage employment shows that the growth in wage jobs was principally in wage jobs with a contract, with both private domestic and foreign direct investment (FDI) firms ( Figure 8). The share in public sector wage jobs remained flat at 11 percent (public sector jobs cover three categories: "government" civil service, "public service units" in public health and educational facilities, and state-owned enterprises).  Next, we consider the shifts in major sectors within the large employment categories. Among wage jobs with a contract, the share working in manufacturing expanded. Manufacturing accounted for 31% of wage jobs with a contract in 2007, increasing to 36% in 2014. As a result of this trend, manufacturing became the largest employer of wage workers with a contract, followed by public services, whose share of wage employment with a contract declined from 46% to 34% between    Services workers (11%) and Plant and Machine operators/assemblers (10%) were other occupations that experienced employment growth significantly above the average. On the other hand, skilled agricultural, forestry and fishing workers as well as technicians and associate professionals experienced negative employment growth, with rates equal to negative 10 percent and 4 percent, respectively. Elementary occupations and craft and related trade workers were two occupational categories that grew at a slightly slower pace than the overall employment growth (2 percent and 3 percent, respectively). The employment growth of the leader/manager category was slightly above the aggregate employment growth, i.e., 5 percent versus an overall employment growth of 4 percent.

This subsection focuses on what jobs look like in
The same message is confirmed when we focus on the top growing and top shrinking occupations at the 3-digit level. 9 In fact, the top growing occupations overwhelmingly belong to high-skilled categories, especially professionals. Among professionals, different health specialists appear in the top 10%. Paramedical practitioners, nursing/midwifery, and traditional/complementary medicine professionals doubled over the period 2007-2014. Services related to tourism (65 percent increase), entertainment (53 percent increase), and sports/fitness (72 percent) also emerged among the 15 fastest growing occupations. However, high growth was still observed among tasks that experts Among the other shrinking occupations, three categories of teaching associate professionals experienced significant negative employment growth. This is in line with international standards, as in fact, these professions are specific to the Vietnamese labor market and are not defined at the associate professional level in the International Standard Classification of Occupations. This trend may suggest a move towards more specialization in the teaching profession at the primary level and for special needs students. No professionals appear to be in the list of the top shrinking occupations, which, except for the job categories already discussed, is dominated by low-skilled occupations.

Education Levels of the Workforce
This section describes the education levels of the Vietnamese workforce and examines various aspects of the returns to education. We consider two aspects of how education is associated with changing job outcomes. First, we analyze how the probability of holding a wage job increases with education level. Second, we examine the returns to education, in terms of higher wage job earnings, among those who have wage jobs.
While the flow of younger workers into the labor force has much higher levels of education than past generations, only modest changes in attainment can be observed over 2007-2014 for the stock of all workers. Over that period, the fraction of the workforce with university education increased from 5 to 8 percent, while the percentage with no more than lower secondary fell from 75 to 71 percent ( Figure 12).

Returns to Education in Terms of the Probability of Holding a Wage Job and Earnings
Next, we turn to analyzing how education level is associated with the probability of holding a wage job. We consider a simple analysis of the returns to education by level, estimated separately for each year during 2007-2014. Using a probit model, we regress a binary variable indicating employment in the wage sector on indicator variables for successive education levels as well as cubic in age, urban/rural, region, gender, and ethnic minority status. Summary results from this analysis are shown graphically in Figure 14. The omitted category is completed lower secondary education, and results are not shown in the figure for education levels below lower secondary. The probability of holding a wage job increases with successive levels of education. The wage probability "boost" associated with upper secondary is surprisingly small: a worker who completes the three years upper secondary was only 5% more likely to hold a wage job than someone who had completed lower secondary. This suggests that the labor market benefit of upper secondary is principally the opportunity to pursue a college or university education, which greatly increases the probability of holding a wage job, by 36% and 50%, respectively, in 2014. The wage probability "boost" associated with college and university education has declined over time. Next, we consider the returns to education (in terms of wage earning) among those with wage jobs. The returns were estimated using OLS regressions of log wages for year on the same specification used for the wage job probits (indicator variables for successive education levels as well as cubic in age, urban/rural, region, gender, and ethnic minority status). The time series starts with 2011 because comparable wage data were not available for previous years. 11 Figure 15 shows a graphical representation of the estimates for each education level. The patterns by education level are similar to those for the wage job probits. The returns to upper secondary alone are only 10%, which reinforces the conclusion that the main benefit of upper secondary education is as a gateway to the higher returns from college (43%) and university (66%) education. Over the period 2011-2014 there are no clear trends in education returns. 11 Please see Table A1 in the Annex for more information on the evolution of key variables in the LFS between 2007 and 2014. Separate estimates of returns to education in the private and public sectors are shown in Figure 16. The returns to education are generally lower in the private sector and over the last years the differences with the public sector at higher levels of education have increased. While in the private sector returns have declined at almost all education levels, in the public sector returns have increased, especially at university level.

Figure 17: Returns to Education in the Private and Public (Disaggregated) Sectors, 2014
Source: Analysis of 2014 Labor Force Survey Note: The reported coefficients are calculated from (hourly) log wage (including bonuses) regressions restricted to all wage workers which control for a cubic in age, urban/rural, region, gender, and ethnic minority status as well as an interaction term between education levels and type of employer.

Women and Ethnic Minorities in the Wage Sector
We consider the differences in the likelihood of holding a wage job and in wages among those holding wage jobs by gender and ethnic group. Figure 18 shows marginal effects for the female indicator variable and an ethnic minority indicator variable from the wage job probits presented in the previous section. Controlling for other characteristics, including age, education level, and region, women were 8.8% less likely to hold wage jobs in 2014, and this figure has been fairly constant over time. Ethnic minorities are also less likely than Kinh to be in the wage sector and this difference appears to have increased slightly between 2007 and 2014.

Figure 18: Probability of Holding a Wage Job Relative to Men (for Women) and to Kinh (for Ethnic Minorities)
Source: Analysis of 2007-2014 Labor Force Surveys Note: The reported coefficients are the marginal effectes of probit regressions that focus on the probability of holding a wage job (conditional on being employed) and control for a cubic in age, urban/rural, region, education level, gender, and ethnic minority status. Figure 19 shows results from similar specifications where the dependent variable is holding a wage job with a contract. (These results are shown only for 2013 and 2014, because data on whether employees have a contract were only collected in that year.) Although women are less likely to hold wage jobs generally, they are slightly more likely to hold wage jobs with a contract. Compared to men with similar characteristics, women are 2% more likely to have a contract than men. Ethnic minorities are 6% point less likely than Kinh to be employed in the wage sector and have a contract.

Figure 19: Probability of Holding a Wage Job with Contract Relative to Men (for Women) and to Kinh (for Ethnic Minorities)
Source: Analysis of 2013-2014 Labor Force Surveys Note: The reported coefficients are the marginal effectes of probit regressions that focus on the probability of holding a wage job with contract (conditional on being employed) and control for a cubic in age, urban/rural, region, education level, gender, and ethnic minority status.
Finally, we briefly examine wage premia/penalties for women and ethnic minorities, among those holding wage jobs. Among those employed in the wage sector, women and ethnic minorities earn lower wages than comparable male and Kinh workers. The gender wage gap has shrunk (from 15.4% to 12.6%) between 2007 and 2014. The ethnic minority wage gap, however, has been fairly stable.

Figure 20: Wage Premia by Gender and Ethnic Minority Status
Source: Analysis of 2011-2014 Labor Force Surveys Note: The reported coefficients are calculated from (hourly) log wage (including bonues) regressions restricted to all wage workers which control for a cubic in age, urban/rural, region, gender, and ethnic minority status as well as an interaction term between education levels and type of establishment.

Conclusions
Vietnam's economic trajectory since liberalization in the 1990s and its continued development successes make its labor market a continuing topic of interest to researchers. This paper traces out some basic characteristics of jobs in Vietnam using largely unexploited data from the annual labor market surveys. The analysis produces a number of salient findings.
First, although the country's recent economic development story has in large part focused on growth in jobs and incomes in foreign firms, employees in such firms remain a tiny fraction of overall employees-just 2 million out of 50.6 million employed workers in 2014. The labor force remains quite diverse, with 20.9 million still working in agriculture (largely family farming), 10.7 million self-employed in non-farm work, and substantial employment in state owned enterprises (1.4 million).
Second, high unemployment rates among recent university graduates appear to reflect the shortterm transition from school to work rather than limited long-term prospects for new graduates. Following cohorts of young university graduates over time, we see that their unemployment rates drop rapidly as they age.
Third, unsurprisingly, more education is associated with both a higher probability of holding a wage job and higher wages among those who hold wage jobs. For most young Vietnamese, the critical marginal education choice is whether to pursue additional education-most typically in the form of an upper secondary degree-after completing lower secondary. We find that the returns to an upper secondary degree alone are modest, both in terms of the probability of finding a wage job and in wages for those with such jobs. The key economic benefit from attending upper secondary is due to the opportunity it creates to attend college or university.
Fourth, controlling for other characteristics, women and ethnic minorities are less likely to hold wage jobs. Among those who do hold wage jobs, they also earn less than comparable men and members of the Kinh ethnic majority. In addition to changes in the number and type of questions asked, in some cases the same types of information were collected using different questions and/or different categories to record the respondent's answers. As a result of this, a careful standardization process was needed in order to be able to compare labor market outcomes during the 2007-2014 period. The standardization work included different types of data elaborations, including: i) combination of information collected via more than one question in certain years but via a single question in other years; ii) within the same question, combination of different categories of answers to create a new category that could be compared over time. While the standardization was possible for most of the variables of interest, for specific variables in certain years this was not possible. Examples include: 1) The inability to create separate "Government", "Public Service Units" and "SOE" categories for the variable "Establishment type" before 2011.
2) The inability to create a variable that consistently presents information about the type of work contract held by a worker because of the lack of this information in 2008 and 2010 and the fact that this question was only asked in certain months in 2011 and 2012.
The table below shows the changes in the classification used for some key variables between 2007 and 2014.