The data in our study are from the Manpower Utilization Survey (MUS) 2000-2010, which represents repeated cross-sections with different individuals in each survey.
The MUS is conducted by Taiwan’s Directorate -General of Budget, Accounting, and Statistics in May every year since 1978. All persons aged 15 and over in the household were interviewed. The MUS collects information on employment status, demographic characteristics, occupation, industry, firm size, monthly wages, tenure, and work history.
We restrict our sample to non-agricultural full-time (hours worked more than 35) workers aged between 26 and 60. The age restriction is to eliminate those just entering the labor market or retiring. People served in the armed forces are also excluded. We use data from 2000-2004 and 2006-2010 representing the pre-policy and post-policy periods, respectively, and drop the transition year 2005. The final
(5)
13
sample consists of 168,859 respondents.
Table 2 shows the definitions and descriptive statistics of the variables by sector.
Nearly ten percent of the sample changed job last year. The mean tenure for all individuals is 7.17 years. The mean tenure for public-sector workers is 11.42 years.
The mean tenure for private-sector workers is 6.98 years. 83.6% of the individuals work in the private sector. The mean age of all respondents is 38.95 years old. The average age of public-sector workers is three years older than private-sector workers.
The average year of schooling is 12.35 in the full sample. Fifty-eight of respondents are male. 64.5% of individuals are married. Comparing with public-sector workers, the percentage of mover is higher for private-sector workers. 10.4% of private-sector workers changed job last year. The mean tenure of private-sector workers is 4.43 years (11.42-6.99) shorter than public-sector workers. The percentage of young worker (age<50) in the public and private sectors are 77.2% and 87.1%, respectively.
The average age for public-sector workers is 41.93 years old and 38.36 years old for private-sector workers. The mean year of schooling for public-sector workers is one year more than private-sector workers. The mean real hourly wage is NT $252.01 (US
$8.44) which is 1.47 times more than private-sector workers (NT $171.23 or US
$5.73). The percentage of white collar workers is 72.9% and 46.2% in public and private sector, respectively. In full sample, 46.1% of the respondents work in northern Taiwan. 77% of private-sector workers are in the small firms whereas only 8.3% of them work in the large firms.
For the estimation of Mover equation, we also include those who are voluntarily unemployed for less than 17 months and had worked prior to being unemployed as movers. In addition, we drop those who did not work at least three months prior to current job because their work history was not collected and they are likely to be just
14
Table 2. Definitions of Variables, sample means, and standard deviation: 2000-2010
variables definition Sample Mean (Standard Deviation)
Full sample public sector private sector mover Equals 1 if tenure is less than 12 months and 0 otherwise 0.099 (0.299) 0.073 (0.261) 0.104 (0.306)
tenure number of years at current job 7.717 (7.073) 11.423 (8.775) 6.988 (6.442)
priv Equals 1 if working in the private sector and 0 otherwise 0.836 (0.370) 0.000 (0.000) 1.000 (0.000) young Equals 1 if age is less than 50 (45, or 55) and 0 otherwise 0.855 (0.352) 0.772 (0.419) 0.871 (0.335)
age Age of respondent 38.951 (8.804) 41.937 (8.880) 38.364 (8.669)
schooling years of schooling 12.350 (3.265) 13.968 (2.986) 12.032 (3.223)
male Equals 1 if male and 0 otherwise 0.579 (0.494) 0.567 (0.495) 0.582 (0.493)
married Equals 1 if married and 0 otherwise 0.654 (0.476) 0.763 (0.425) 0.633 (0.482)
real_wage respondent’s real hourly wage 184.492 (96.779) 252.007 (100.116) 171.225 (90.362)
w_collar Equals 1 if white collar and 0 otherwise 0.506 (0.500) 0.729 (0.444) 0.462 (0.499)
northern Equals 1 if working in the northern region and 0 otherwise 0.461 (0.498) 0.394 (0.489) 0.474 (0.499) central Equals 1 if working in the central region and 0 otherwise 0.227 (0.419) 0.226 (0.418) 0.228 (0.419) southern Equals 1 if working in the southern region and 0 otherwise 0.285 (0.452) 0.317 (0.465) 0.279 (0.449)
eastern Equals 1 if working in the eastern region and 0 otherwise 0.026 (0.160) 0.064 (0.244) 0.019 (0.137)
priv×small Equals 1 if firm has less than 100 employees at current job 0.644 (0.479) 0.000 (0.000) 0.770 (0.421) priv×large Equals 1 if firm has greater than 500 employees at current job 0.069 (0.254) 0.000 (0.000) 0.083 (0.276) priv×shortage A indictor of private sector shortage rate 2.251 (1.158) 0.000 (0.000) 2.693 (0.643)
urate county unemployment rate (as a percentage) 4.523 (0.826) 4.541 (0.818) 4.520 (0.827)
Number of Observations 168,859 27260 138,311
Notes: Standard deviations are in parentheses.
15
entering the job market. After the restrictions, the sample consists of 168,064 respondents.
We first calculate the simple DD estimate of treatment effect by using the following equation:
∆yi = [(yi|private, post2005) − (yi|private, pre2005)]
− [(yi|public, post2005) − (yi|public, pre2005)]
Table 3 shows that the difference in the changes of the percentage of mover and tenure between private and public sector is -0.01 and 1.32, respectively. The results suggest that the 2005 pension reform may not reduce the job-lock effect.
Table 3. The result of simple DD estimate of treatment effect
treatment group:
We also calculate the simple DDD estimate of treatment effect by using the following equation:
∆yi = {[(young, post2005) − (young, pre2005)|private]
−[(old, post2005) − (old, pre2005)|private]}
−{[(young, post2005) − (young, pre2005)|public]
−[(old, post2005) − (old, pre2005)|public]
We use young workers whose age is less than 50 as an example to calculate the simple DDD estimate. In Table 4, the difference in the changes of the percentage of mover is -0.009 and in the changes of tenure is -2.69 between young private-sector
16
workers and their counterparts after 2005. The simple DDD estimate shows that the effect of the 2005 reform on mobility is mixed. To confirm that the 2005 reform has effect on mobility, we need to control other variables that might affect mobility.
Table 4. The result of simple DDD estimate of treatment effect
particularly sensitive to change in all samples. The coefficient of post2005×priv is negative and statistically significant in all samples. The mobility probability for private-sector workers decreases by 0.021 percentage point after 2005 in full sample and decreases by 0.02, 0.019, 0.022, and 0.029 percentage points in male, female, married and single samples, respectively. The coefficient of priv indicates that the mobility probability of private-sector workers are 0.145, 0.146, 0.148, 0.134 and 0.185 percentage points greater than public-sector workers in full, male, female, married and single samples, respectively. After 2005, the mobility probability decreases by 0.015 percentage point in full sample. The negative sign of