The Size of the Employment Effect

3.1 Existing Estimates

How badly has the employment of migrants been affected by contractionary macroeconomic policies, the new Labor Law, and the global financial crisis? Three recent surveys reveal a very gloomy picture. According to a January 2009 survey by the Ministry of Agriculture in 150 counties and 15 provinces known as major migrant-originating areas, around 15.3% of the 130 million migrant workers—20 million—have lost their jobs because of the global financial crisis.²

Another official survey by the National Bureau of Statistics among 68,000 rural households in 31 provinces, 857 counties, and 7,100 villages between the end of 2008 and early 2009 shows that, of the 140 million or so rural migrants who worked in cities in 2008, 50% had returned to their home villages by the end of the year. Eighty percent of those who returned home went back to the cities after the Chinese New Year, while 20% (10% of the total migrant labor force) chose to stay on in their rural hometowns.

The third source of information is a series of small surveys conducted by the Xilu Migrants Survey Group between November 2008 and January 2009. Of the 809 individuals surveyed, 89.5% had returned home for the Chinese New year by 14 January, 12 days before the holiday. Forty-six percent of the returning migrants said that they had returned earlier than usual, mostly (in 76% of cases, or 35% of the total migrant labor force) because their workplaces had shut down, downsized, or forced them to take extended holidays. Of those who had returned home for the Chinese New Year, 69% indicated that they would definitely be going back to the cities, 7.2% suggested that they would not do so, while the remaining 24% had not decided either way at the time of the survey.

All of the three studies seem to indicate a similar size of adverse impact from the economic downturn. But these studies are based primarily on surveys in the migrant-originating areas and on the stated intentions of the migrants to return to the cities. In addition, the timing of these studies around the Chinese New Year may have biased the results because migrant workers usually quit their jobs to go home and then find new jobs after coming back. What really happened is not revealed in the literature. Did the workers really go back to the cities as they said they would, and if so could they find jobs? In this paper, we use more objective information gathered by sampling, interviewing, tracking, and re-interviewing a group of migrants in the RUMiCI survey to gauge the size of the adverse impact of the economic downturn on migrant employment.

3.2 RUMiCI Survey and Tracking Method

The RUMiCI project is a research collaboration initiated by the Australian National University and Beijing Normal University. The project surveys three groups of households: 5,007 rural–urban migrants who worked in 15 designated cities³ in 2008, 5,000 urban households in the same cities, and 8,000 rural households from 10 provinces or metropolitan areas where the 15 cities are located.⁴ While the urban and rural household surveys use household survey samples from the National Bureau of Statistics, there is no available sampling frame for rural–urban migrants. Previous migration surveys normally used a household-based sampling methodology, whereby interviewers randomly selected migrants in selected urban neighborhood communities. The main limitation of this sampling method is that only a small proportion of migrants in PRC cities live in urban neighborhood communities. A large proportion of them live in factory dormitories, at the back of the restaurants or construction sites where they work, or in rural suburbs in the vicinity. A sample drawn from city residential communities may be quite unrepresentative of the migrant population. The RUMiCI research team employs a unique and innovative sampling strategy to address this concern.

Essentially, the survey uses a sampling frame based on information collected in a census of migrant workers at their workplaces. More specifically, during the census stage, we first defined the boundaries for the 15 cities so that they included at least part of the areas where there were manufacturing firms with large migrant worker concentrations. These were often at the junction of the city and its surrounding rural areas and often excluded from the definition of the city. Our city areas were then divided into 500 x 500 meter blocks and a number of blocks (equal to around 12% of the sample size for each city) were randomly selected. Within each block, all the workplaces were covered by the census. The questions included the industry type, the total number of workers, and the total number of migrant workers. The data obtained from this census of workplaces were then used as the sampling frame and simple random sampling was used to select our sample individuals. Once we had located the individual migrants, the their households were also interviewed.⁵

Relative to household-based sampling, the advantage of using workplace-based listing information is that it includes all the migrants who are working regardless of where they live. The sampling method excludes migrant households whose members are all unemployed at the time of the sampling. Fortunately, as most migrants have no access unemployment benefits or any other safety net, those who cannot find a job generally do not stay in the city. As a result, the unemployment rate for migrants has been low (see, for example, Du, Gregory, and Meng [2006]).

In October–December 2007 the project team conducted a block census in the 15 cities. The sampling and survey took place between March and May 2008, after the migrants had gone back to the cities from their annual home visit during the Chinese New year. In the 15 cities 5,007 migrant households were interviewed. During the sampling in March–April 2008 we observed that a sizable number of workplaces, mainly in Dongguan, Shenzhen, and Guangzhou, had already shut down. The survey company indicated to us that this was the result of the implementation of contractionary macroeconomic policies and the introduction of the new Labor Law.

Because this was to be a 5-year project, a tracking strategy was developed. An important feature of the migrant population is its high degree of geographic mobility, making tracking very difficult. At the time of the survey we recorded the individual migrants' work and home addresses and other contact details in the cities as well as in their home villages. We also recorded the phone numbers of three close relatives or friends of each interviewee so that we could track them even if they and their households moved. These are normal tracking strategies adopted by any panel survey. We realized, however, that our sample population was different and that, because of its high mobility, we might have to enhance the simple tracking design. We therefore thought of further tracking incentives. First, we decided to conduct three lottery activities each year for all our respondents. The prizes in the first year ranged from CNY50 to CNY2,000, and the prize amounts are set to increase each year. The lottery is designed to cover only 1.5% of the respondents, but if an individual participates in the survey for 5 years the probability of receiving a prize goes up to 20%. Only when an individual is successfully tracked is his or her name entered in the lottery. The respondents were told all this at the time of the survey interview. We hoped that this would encourage them to keep us informed of their whereabouts. Another incentive strategy we devised was to send a present to each respondent's rural home each year before the Chinese New Year, in the hope of strengthen the link between the respondents and the project team.

Five months after the first survey (October 2008) we contacted all the respondents to confirm their contact details and send each one a small gift. At the same time, we revealed the first lottery results. Soon after, in December and again in February 2009, we conducted two more rounds of tracking.⁶ However, despite all these efforts, the 2008 survey had a very high attrition rate, closely associated with the economic downturn. To compensate for the high attrition rate, we carried out a re-sampling based on 2007 census data before the 2009 survey. But first we had to validate the 2007 census framework, given the economic downturn. We revisited 13% of the census blocks (64 blocks) from the original total of 489 blocks surveyed in 2007 to find out whether the workplaces recorded at the time had changed their operating status.⁷ This small-scale re-census provided us with another set of information for determining whether the original workplace had shut down, had changed to a new entity, or had not changed. It did not, however, give us information about the degree of downsizing within existing workplaces. Hence, it gave us a lower-bound estimation of the impact of the economic downturn.

3.3 Estimates of the Adverse Employment Impact Using RUMiCI Data

Table 1 [ PDF 25.4KB | 2 page ] presents the number of workplaces and rural migrant workers, by city and industry, according to our 2007 census data. The table shows that, among our census migrants, 67% were employed in services and wholesale–retail trade, while 18% were in manufacturing. Manufacturing was heavily concentrated in the coastal regions. In addition, more than 70% of the migrants were employed in eight coastal cities (Guangzhou, Dongguan, Shenzhen, Shanghai, Nanjing, Wuxi, Hangzhou, and Ningbo). These data gave us a starting point for describing the employment status of rural–urban migrants in our sample cities before the economic downturn, and for comparing the data with employment data after the downturn.

We based our assessment of the employment situation since the economic downturn on two data sources: the block re-census data and the sample tracking data.

We classified our block re-census workplaces into those that had been shut down and those that had not. We then used this information together with the 2007 census data to draw implications regarding the proportion of workplaces in each industry within our 2007 census blocks that had shut down and, using probability distribution as weight, to generate an implied shutdown ratio for the city as a whole. In addition, with the 2007 census data on the number of migrants employed in each workplace, we were able to calculate the implied employment impact of the shutdowns.⁸ Table 2 [ PDF 22.9KB | 2 page ] shows the proportion of workplaces that have shut down since the economic downturn, by city and industry. In all 15 cities, around 9% of the workplaces have shut down since November–December 2007. Among these cities, Wuhan and Dongguan have had the highest proportion of shutdowns and almost all industries in both cities have been adversely affected. In the 15 cities as a whole, shutdowns have been highest in construction, manufacturing, and various types of agencies.

Matching the closed workplaces with the number of migrant workers they hired in November–December 2007 allowed us to examine the employment effect of the shutdown. Table 3 [ PDF 83.5KB | 1 page ] presents the industry and city distribution of the impact of the shutdown on migrant employment for both calculations. On average, around 13% (1.4 million) of the migrant employment in the 15 surveyed cities has been affected by the shutdowns in the economic downturn. Among the 15 cities, Dongguan, where around 34% of the migrant employment has been affected, is the worst hit. The other cities that have been badly affected are Wuxi and Ningbo, where around 20% of the migrant employment has been affected by shutdowns. If we believe that the global financial downturn has mainly affected export-oriented cities, we may rank our 15 cities according to their export concentration. Appendix A [ PDF 19.2KB | 1 page ] gives the 2006 the share of export value in city gross domestic product for each of our 15 cities and ranks them accordingly. Shenzhen and Shanghai rank first and third, but surprisingly their migrant employment has been among the least affected by shutdowns. This suggests that perhaps factors other than the global downturn are at work here.

Further, the impact on migrant employment has been felt in both the traded goods industry (manufacturing) and the non–tradable goods industry (wholesale and retail trade. In Wuhan, 28% of migrant employment in the wholesale–retail trade sector has been affected. From this point of view, it seems that both domestic policies and the global financial crisis have had sizable impact.⁹

Note that the shutdown effect is a lower-bound estimate of the extent of the effect of the total economic downturn on employment. From our re-census data, we could not gauge the size of the downsizing in the existing workplaces. Our tracking data for the 5,007 migrant households surveyed in 2008 may shed some light in this regard. However, using tracking data to infer the effect of the economic downturn effect on employment can be quite tricky, as in a normal environment one expects a certain attrition rate. The literature does not provide a benchmark attrition rate for an average population in a normal economic environment. The Household, Income and Labour Dynamics in Australia (HILDA) survey and the British Household Panel Survey (BHPS) may provide such benchmarks. Between the first wave and the second wave, HILDA had an attrition rate of around 13.2%, while the rate for BHPS was 12.4% (Watson and Wooden 2004). We could therefore regard the 12%–13% attrition rate for HILDA and BHPS as a benchmark for an average population in a normal economic environment.

Our sample population, however, is much more mobile, and hence the attrition rate for the average population represented in HILDA and BHPS may not be a suitable benchmark for our purpose. In developed countries, the highly mobile population is the youth, in particular young men (see, for example, Olsen [2005]). The group aged 20–24 in the HILDA survey had an attrition rate of 23.4% between wave 1 and wave 2 (Watson and Wooden 2004).

The migrant population represented in our sample, although older, with an average age of 28 years, faces a very different institutional environment from that of the youth in developed countries. Because of the restrictions on access to jobs in the formal sector and to social services and social safety net in the cities, migrants' jobs are very insecure and there is no safety net to rely on in case of job loss (Meng 2000; Meng et al., forthcoming). Hence, migrants tend to move around much more than the average population or even ordinary youth.

Using the data on the length of time since the first migration and the number of cities migrants have worked in since the first migration, we can calculate the average length of stay of migrants in one city. In our 2008 sample of 5,007 migrant household heads, the median length of stay in one city for work purposes is 3 years (see Figure 1 [ PDF 13.3KB | 1 page ]). Thus, the average annual mobility should be around 30% without the economic downturn. With this information as well as information about a “normal” mobile population in the HILDA survey, we can expect a normal attrition rate of around 30% for our sample between the first and second wave in a normal situation without the economic downturn.

As discussed earlier, we tracked our 5,007 migrant households three times, in October 2008, December 2009, and February 2009. Each time, our targeted population was the 5,007 original households surveyed. The attrition rate for each of these tracking dates is reported in Table 4 [ PDF 13.6KB | 1 page ]. The attrition rate for the first tracking in October 2008 was 34.2%, already higher than what we expected under a normal economic environment. The second tracking 2 months later resulted in a 39.4% attrition rate for the original 5,007 sample households, or an additional 5.2 percentage point loss. Relative to the 3,296 households tracked in October 2008, the second attrition rate was 16.7%. Finally, just before we embarked on the second survey, the third tracking recorded an attrition rate of 48.8% with respect to the original sample (14.6 percentage points below the first tracking), or 24.9% with respect to the sample of 3,035 households in the second tracking. This third and final tracking gave us an attrition rate 18.8 percentage points higher than our expected attrition rate for a normal economic environment. This may serve as an upper-bound estimation of the employment effect of the economic downturn.

The industry and city distributions for the total lost sample from the third tracking (February 2009) are presented in Table 5 [ PDF 16.8KB | 1 page ]. The table shows that all the cities had lost more than 30% of the original sample by the third tracking. The cities that had lost more than 50% of the original sample were Guangzhou, Dongguan, Nanjing, Hangzhou, Ningbo, and Wuhan. Although most of these cities, except Wuhan, are from the coastal region, where the export industries are concentrated, the cities with the highest and third-highest export concentration, Shenzhen and Shanghai, are not part of this list. It is also interesting to see that Nanjing and Ningbo had extremely high rates of attrition, more than 10 percentage points higher than that of Dongguan, whose export concentration exceeds that of the two other cities by a large margin.

Among the six industry groups, construction, manufacturing, and services had lost more than 50% of the original sample. The impact on manufacturing can be traced directly to the global financial crisis because of export intensity, but the impact on construction and services is not so easily explained. The generally high attrition rate across cities and the fact that the non–tradable goods sectors (construction and services) have been hit equally hard suggests that the impact on migrant employment is not only from the global financial crisis but a combined effect of domestic macroeconomic policies and external shocks.

Download this Paper [ PDF 387.9KB| 28 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.