The effect of pension portability on job mobility (so called “job-lock” effect) has been widely examined in the literature (Allen, Clark and McDermed,1988; Gustman and Steinmeier, 1933; Andreietti and Hildebrand, 2001). Job-lock effect is important because workers are discouraged from switching to jobs where they are more efficient producers, and subsequently leads to inefficiency in the labor market. Concerns with the job-lock effect and the lack of financial security after retirement have caught the public’s attention. As a result, the private-sector pension administered by the Bureau of Labor Insurance in Taiwan switched from the DB (defined benefit) plan to the DC (defined contribution) plan in July 2005.
The DB plan calculates pension using job tenure and the average wages six-months prior to retirement. It accrues value disproportionately in the later years of the employment. The ‘backloading’ feature of the DB plan implies that workers who leave the company prior to retirement give up a large share of their pension benefits. More importantly, only workers who work for the same company more than 25 years, or over 15 years and order than 55 are eligible to claim their pension. However, most companies in Taiwan are small- and medium-sized enterprises (SMEs). Their average lifetime is about 13 years, which means that the majority of the private-sector workers will not receive their pension. Unlike the DB plan, the DC plan eliminates the job-lock feature of the DB plan by assigning the ownership of the pension to the workers.
In addition, the theory of equalizing differences suggests that pension and wage are perfect substitutes (Woodbury, 1983). Workers receiving higher pension benefits may accept a lower wage. Prior to the reform, employers are required, but not strongly enforced by the government, to contribute 2% to 15% of worker’s monthly wage to
2
retirement reserve fund. In contrast, the DC plan requires employer to contribute 6%
of worker’s monthly wages to their personal pension account. If employer treats the 6% pension contribution as a deferred compensation, we will expect that employers may shift part of the pension contribution to workers by reducing their monthly wage. Luoh and Yang (2009) find that the switch from the DB plan to the DC plan in 2005 in Taiwan decreases the wages of private-sector workers by about 5.92% which is equal to the contribution rate mandated by the government. This large shit implies that labor supply is perfectly inelastic. Based on their study, we use different approach to estimate the effect of 2005 pension reform on wages.
Although the job-lock effect has been widely examined, no consensus emerged from these studies. Mitchell (1982) shows that employer-provided pension reduces the probability of job changed and the effect is more significant for males than for females. Using data from UK, Lluberas (2008) compares the effect of occupational pension plans on job mobility among workers with the DB and DC plans. The results show that workers with the DB plan are less likely to change jobs than those with the DC plan. On the other hand, Gustman and Steinmerier (1993) find that the lack of pension portability is not the main determinant for the low job mobility in the United States. They show that pension covered jobs often offer better compensations and benefits. It is these benefits that account for lower turnover rates among pension covered workers rather than the non-portability. Andrietti and Hildebrand (2001) find no evidence that the losses in pension portability deter job mobility among workers covered by employer provided pension. More importantly, they show that the DC plans have a significant negative effect on mobility.
Hermaes, Piggott, Vestad and Zhang (2011) and Rabe (2004) use the “potential portability gain (loss)” (PPG or PPL) as a proxy for the pension costs of changing
3
jobs in Norway and Germany, respectively, and employ probit models to estimate the job change propensity equations. The PPG (PPL) is defined as the gain (loss) from moving to a new job with identical wage compared to staying. Their results indicate that there is no significant job-lock effect of the employer-provided pension plan (under the DB plan) in Norway and Germany.
The employer-provided health insurance (EPHI) has also played an important role in job-lock effect. Madrian (1994) finds that the mobility reduces by 25% for workers with EPHI by employing the DD (difference-in-differences) model. Monheit and Coooper (1994) predict the likelihood of compensation and insurance status using their predicted values as independent variables in a job-lock probit equation. They find that individuals who are covered with health insurance are 3 to 6 percentage less likely to change jobs. Using statewide variation in Consolidate Omnibus Budget Reconciliation Act (COBRA), Gruber and Madrian (1997) show that health insurance continuation mandates do increase job mobility.
Dey and Flinn (2005) indicates that the employer-provided health insurance tends to reduce mobility. Similarly, Sane de Galdeano (2006) finds that the 1996 Health Insurance Portability and Accountability Act (HIPAA) has no negative effect on mobility using DDD (difference-in-difference-in-differences) model. Okunade and Wunnava (2002) estimate the effect of employer-provided health insurance on job-lock by using tenure as dependent variable. Their results suggest that women with medical insurance coverage have almost 66 more weeks of tenure than women without this insurance.
The theory of equalizing differences suggests that workers receiving higher fringe benefits are paid a lower wages. The theory has been tested on various fringe benefits, such as health insurance, maternity benefits, and pension benefits, and the results are
4
mixed. Gruber and Hanratty (1993) show that both employment and wage increase after the implementation of Canadian Nation Health Insurance (NHI). Kan and Lin (2009) use DD approach to show the impact of Taiwan’s NHI on hours worked and wages. Their results indicated that hours worked among private sector workers declined relative to their public sector workers. On the other hand, the wages of private-sector workers were not affected by the NHI. Sheu and Lee (2012) make some improvement upon Luoh and Yang’s (2009) and Kan and Lin’s (2009) studies. They extend the time period from five years to ten years for analysis the effect of NHI and the 2005 pension reform on wage. They find that the new pension scheme (the DC plan) has a negative effect on wages which reduce by 1.92% and the implementation of NHI reduces wage by 4.4%.
Olson (2002) uses husband’s firm size, union status and health insurance coverage as instruments of wife’s employer-provided health insurance to estimate the trade-off between wife’s wage and health insurance. The results suggest that wives with employer-provided health insurance accept 20% lower wage than without employer-provided health insurance. Jensen and Morrisey (2002) view employer-sponsored fringe benefit as endogenous in the wage equation because workers have heterogeneous preferences of fringe benefits. They find an insignificant tradeoff between fringe benefits and wages by employing two stage least squares (2SLS). Vestad (2011) estimates the effect of the mandatory occupational pensions in Norway, which is implemented in 2006, on wage. This creates a natural experiment for analysis. He finds that less than 50% of the cost of a minimum pension contribution is shifted from the company to workers.
In this paper, we use the data from Manpower Utilization Survey (MUS) 2000-2010 to examine the effect of 2005 pension reform on mobility using DD and
5
DDD approaches. In Taiwan, the private-sector pension is covered by the Labor Insurance while the public-sector pension is covered by Government Employees’ and School Staffs’ Insurance (GESSI). The policy change creates a natural experiment, which allows us to examine the impact of 2005 reform on job mobility using DD approach.
During the transition period, workers currently working have to decide whether to switch to the DC plan within five years, but workers who just enter in labor marker or change jobs after implementation date are mandatory switched to the DC plan.
However, enrolling in the DC plan may not an optimal choice for workers whose tenure is over 15 years (Council of Labor Affairs), because they have already been eligible to claim the pension under the DB plans. As a result, old workers can be served as additional comparison group. This allows us employing DDD estimator to reduce the effect of unobserved heterogeneity between private- and public-sector workers on our estimates of mobility. We also make improvement upon Luoh and Yang’s study by employing DDD approach to examine the effect of the 2005 reform on wage.
The organization of this paper is as follows. Section II contains a brief review of the pension schemes in Taiwan. Section III lays out the empirical framework. We describe our data and the definitions of the variables, and empirical results, in Section IV and V, respectively. Finally, we provide a conclusion in Section VI.