國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
25
6. Results
Table 5 shows the results of the difference in differences model (DID). Column
Ⅰfrom Column Ⅵ in Table 5 represents different windows of observations used in the estimation. Column Ⅶ presents the results by dropping some observations which might affect the accuracy of the analysis.
First, Column Ⅶ excludes the period from 2003 to 2004 due to the abnormally high layoff in the reference group (Figure 2). Second, according to the older pension system in LSA, employees who satisfy some specific conditions tend to choose the older pension system. If this is the case, they would not be affected by the severance pay change.
Such kinds of specific conditions are the requirements for the employees to apply for retirement pensions in the older pension system, which are listed as follows:
“1. When the worker attains the age of fifty-five and has worked for fifteen years.
2. When the worker has worked for more than twenty-five years.
3. When the worker attains the age of sixty and has worked for ten years.”
Employees who satisfy the requirements above are excluded from the sample to assure the remaining employees in the treatment group are indeed covered by LPA.
Third, employees hired in the period after the policy change must be the ones to whom LPA is applicable. These employees might find it easier to meet the expectations of the employers and have higher productivity, resulting in lower layoff rates. Therefore, Column Ⅶ also eliminates such kinds of employees in order to get pure estimates of the severance pay effect.
Generally, most of employees to whom LSA is not applicable belong to the public sector and have relatively higher job security than other employees. As a result, one would expect that the layoff rate of employees in non-public sector businesses
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
(treatment group) will be higher. That is, the corresponding coefficient of LSA should be positive. Except for the results in Column Ⅱ and Column Ⅲ, Table 5 shows that employees who are covered by the Labor Standards Act do have higher layoff rates.
In Column Ⅶ, one can see that layoff rate increases by 0.68% on average for the treatment group, relative to the reference group.
As Figure Ⅱ shows, laid off rates for both the treatment group and the reference group are declining before implementation of the Labor Pension Act.
However, there are no obvious trends in layoffs in the two groups after implementation of LPA. Hence, the coefficient of IMP (time trend effect) is ambiguous and hard to predict. Actually, all results in each column show that time trend does not have significant effect on the layoff rate.
The interaction term LSA*IMP reflects the effect of policy change, and one would expect it to be negative due to the reduction of severance pay. However, the results from columnⅠto columnⅤsuggest that the severance pay seems to have no significant effects on the layoff rate.
Although the DID estimate in Column Ⅵ suggests that the severance pay has significant negative impact on the layoff rate, it is inconsistent with the expectation of a positive impact on the layoff rate. In fact, one should notice that the negative result might have been caused by employees hired after the policy change, as explained above. After dropping such observations, Column Ⅶ suggests that severance pay does not have any significant impact on the layoff rate.
The remaining effects of the control variables are described as follows. Log of income is expected to have negative effect on the layoff rate. One of the reasons is that employees who have higher income are usually important to the firm and have some characteristics which are hard to be substituted by other employees. Therefore, employees with higher income are not easily laid off even in business contractions or
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
27
other similar situations. Results from Column Ⅱ to Column Ⅶ suggest that log of income does have significant effect on the laid off rate. In Column Ⅶ, one can see that a one percent increase in income results in a decrease of 0.39% in layoff rate, on average.
The effect of gender on the laid off rate is opposite to the expectation. Usually, one will expect that females are more easily laid off than males due to gender discrimination. However, all results from Column Ⅰ to Column Ⅶ show that males face higher layoff rates than females. In Column Ⅶ, one can see that the layoff rate increases by 0.42% on average for males relative to females. This result might be explained by the fact that most females are white collar workers or work in the service industry, and such kinds of jobs generally prefer females than males. Also, many males are menial or blue collar workers, and such kinds of jobs are relatively easy to suffer layoffs, thus causing the higher layoff rate for males.
Generally, skilled workers or workers in important positions have higher education and these workers are relatively hard to be laid off. As a result, education is expected to have negative effect on the layoff rate. The results from Column Ⅳ to Column Ⅶ suggest that education does have a significant effect on layoff rate. In Column Ⅶ, one can see that a one year increase in education will decrease the layoff rate by 0.05% on average.
Age seems to have positive effect on the layoff rate, as the results suggest, but the effects are significant only in Columns Ⅰ, Ⅱ and Ⅲ. In Column Ⅲ, one can see that one year increase in age increases the layoff rate by 0.02% on average. The reasons why older workers face higher layoff rate might be the fewer opportunities for promotion relative to young workers. That is, older workers might have less creativity and productivity than young workers, hence resulting in the higher layoff rate.
Different from age, job seniority reflects relatively higher and abundant
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
28
experience and skills. Therefore, one would expect that higher job seniority will lead to a lower layoff rate. In fact, all results from Column Ⅰ to Column Ⅶ suggest that job seniority does have a significant effect on the layoff rate. In Column Ⅶ, one can see that a one year increase in job seniority results in decrease of the layoff rate by 0.010% on average.
Household size means the number of members over the age of 15 in the family.
Since a larger household size implies that employees have more responsibilities, one would expect such kinds of employees will work harder in order to support his/her family. Hence, employees with larger household size are supposed to have lower layoff rates. The results from Column Ⅱ to Column Ⅵ suggest that household size does have negative effect on the layoff rate.
Whether an employee is the head of the household might affect the responsibility he/she faces and hence the effect on the layoff rate should be negative. However, none of the results from Column Ⅰ to Column Ⅶ show that being the head of the household has a significant effect on the layoff rate.
Working areas control the differences caused by some specific characteristics of each area. According to the development plan issued by the Economic Bureau in 1979, Taiwan can be separated into four areas, northern, central, southern and eastern area.
Generally, one would expect that more developed areas such as the northern area will have higher layoff rate due to the competition. However, all results from Column Ⅰ to Column Ⅶ suggest that the four areas are not significantly different in terms of the layoff rate.
The subprime mortgage crisis began in September 2008 and resulted in the global recession and financial crisis. Not only the stock market but also export industries in Taiwan were affected by this event, thus resulting in business contraction and involuntary unemployment. Treating the year 2008 as a dummy variable, the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
29
results from Column Ⅳ to Column Ⅵ show that subprime mortgage did have a significant effect on the layoff rate. In Column Ⅶ, the estimate of subprime mortgage crisis effect is absent because the number of employees laid off in the year of 2008 happens to be zero, after excluding observations which might affect the accuracy of the analysis. Nevertheless, in Column Ⅵ, one can still see that layoff rate increases by 0.88% on average when the subprime crisis happens.
‧
Table 5: Impact of Severance Pay on Layoff Rate (Average Marginal Effects)
(Ⅰ) (Ⅱ) (Ⅲ) (Ⅳ) (Ⅴ) (Ⅵ) (Ⅶ)
Subprime Crisis 0.0100*** 0.0082*** 0.0087***
(0.000) (0.000) (0.000)
N 16,771 16,392 33,163 49,446 65,608 80,800 58,602
pseudo R2 0.0286 0.0959 0.0522 0.0656 0.0663 0.0654 0.1382
Note: ColumnsⅠ to Ⅳ do not cover the period of subprime mortgage, so the dummy of subprime crisis is not included. Column Ⅶ excludes the period from 2003 to 2004, and eliminates employees who chose LSA rather than LPA. Besides, employees who had never worked before or had changed jobs after the reform are dropped in Column Ⅶ as well. The dummy of subprime crisis is absent in Column Ⅶ since the number of employees laid off in the year of 2008 happens to be zero after dropping some specific observations. p-values in parentheses: * p<0.1, ** p<0.05, *** p<0.01
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
31