Intercepts-and-slopes-as-outcomes

The intercepts-and-slopes-as-outcomes model estimation results are shown in Table 5.

For the impact of GENDER on income level, the estimated coefficient value is 0.388, which is not significant. The findings of many previous studies suggest that the working incomes of female workers are lower than those of male workers (Glower and Hendershott, 1988; Crellin et al., 1988; Sirmans and Swicegood, 1997; Jud and Winkler, 1998; and Benjamin, Jud and Sirmans, 2000). However, Abelson et al. (1990) argued that the working incomes of female workers are greater than those of males.

This study concludes that the probability of high income among male workers is higher than that among female workers, but that the difference does not reach the level of significance.

The estimation coefficient of educational level above university is 0.303, which does not reach the level of significance. The coefficient of estimation of educational level of college is 0.427, which also does not reach a 10% significance level. This suggests that the agents with an educational level of college or an educational level above university do not perform better in business than agents with high school or vocational school educations.

The estimation coefficient of AGE is 0.067 (Odds Ratio = 1.070), reaching the 5%

significance level. This result is not consistent with the findings of Sirmans and Swicegood (1997; 2000), who found that older agents had lower incomes. This suggests that older agents have a higher probability of high incomes. The estimation coefficient of MAR is -0.041(Odds Ratio = 0.960), which also does not reach a 10%

significance level. This result is not consistent with the findings of Abelson et al.

(1990). They used the labor capital model to test the impact of marital status on performance, finding that married workers exhibit better performance. The estimation

coefficient of “having children aged under six” is 0.401 (Odds Ratio=1.493). Goldin (2006) argued that children were the most important factor related to out-of-work spells for women and a clear non-linearity exists in the impact of successive numbers of children. One child increased total time not at work by just 0.36 years on average, two children by 1.41 years, and three (or more) by 2.84 years. This result suggests that agents with children under the age of six have a greater probability of high income.

The empirical results are not consistent with the theoretical expectations, without reaching the 10% significance level.

Regarding the weekly average working hours, the estimation coefficient of the median average working hours (HOURM) is 0.969 (Odds Ratio=2.639), having reached the 5% significance level. The estimation coefficient of the high average working hours (HOURH) is 1.846 (Odds Ratio=4.81), having reached the 5%

significance level. This suggests that the individual incomes of agents with median and high working hours are significantly higher than those of agents with low average working hours. This indicates that working in the housing agency industry imposes a time cost. In addition to the familiarity with specific laws and regulations, it requires more efforts in development, marketing sources of clients and case sources to get higher income. The results are consist with the findings of Glower and Hendershott (1988), Crellin, Frew, and Jud (1988), Abelson, Kacmar, and Jackofsky (1990), and Sirmans and Swicegood (1997) regarding working hours and income.

The coefficient estimate of EXP,₈₀ , is 0.121 (Odds Ratio=1.129), which reaches the 10% significance level. The coefficient estimate of EXPS, ₉₀ , is -0.004, which does not reach the 10% significance level. This suggests that increased work experience can help improve the income of agents; however, the effects of that influence will not progressively decrease with the accumulation of work experience.

This result is partly consistent with the conclusions of Glower and Hendershott (1988) and Sirmans and Swicegood (1997) to the effect that experience increases the performance of brokers or salespeople but, beyond a certain point, additional experience is of lesser value.

According to the empirical results, the coefficients of GENDER, UNI, COLLEGE, MAR, CHILD, and EXPS do not reach the level of significance, indicating that these

variables have no significant impact on individual income. As observed previously in the literature, a variety of influencing factors can, due to different sampling techniques, lead to inconsistent results among different studies. In particular, conflicting results often occur with regard to gender, race, franchise affiliation, and the agent’s age (Sirmans and Swicegood, 2000).

Concerning the cross-level direct impact, the branch-level coefficient estimate of TYPE ,j ₀₁ , is 4.121, and does not reach the 10% significance level. According to empirical studies, the individual incomes of agents working for a regular chain are not significantly higher than those of agents working for franchises. This suggests that there is no significant difference between the regular chain system and the franchise system in terms of their respective impacts on the probability of high agent income. In recent years, the developmental trends of the brokerage industry in Taiwan suggest that operational systems have considerably changed from the types based on the regular chain system, in which agents receive a high base salary, to types based on the franchise system, in which agents do not receive a base salary. This is evidenced by examples including Yongqing Housing, Pacific Housing, and Dong Sen Housing. Such changing trends can also be confirmed by empirical results. The individual incomes of agents working under the regular chain system are not significantly higher than those of agents working under the franchise system.

The estimation coefficient of branch location is 1.100 (Odds Ratio = 3.003), which reaches the 5% significance level. This result suggests that the probability of better individual incomes among agents working in branches in the downtown area is higher than that among those working in branches in the suburbs, which confirms the findings of previous studies. For example, many authors (Follain et al., 1987); Glower and Hendershott, 1988; and Sirmans and Swicegood, 1997) have suggested that the incomes of agents working in the housing brokerage industry in metropolitan areas are higher than those of agents in outlying areas. Intermediary agency branches closer to downtown areas have more business activity and higher transaction prices, resulting in better income indicators. In addition, according to the estimation results of the intercepts-and-slopes-as-outcomes model, the u_0j variance did not reach a significant level, indicating that the intercepts do not have random components.

Regarding moderating effects, the results indicated that management type can significantly moderate the marginal effect of UNI on individual income. This indicates that the incomes of those with a university and above degree are higher than the incomes of those with a high school or below degree. Moreover, this gap is higher in the case of the regular chain system than the franchise store system. In addition, the management type can significantly moderate the marginal effect of age on individual income. When the age is older, the expected individual income is higher, and the marginal effect of the regular chain system is lower than that of the franchise system.

Management type can significantly and negatively moderate the marginal impact of working hours on individual income. This suggests that the individual incomes of those with median and high working hours are higher than those of agents with lower average working hours. However, the gap is smaller in the case of the regular chain system and the franchise system. The management type can also significantly and

positively moderate the marginal impact of working experience on individual income;

that is, the positive impact of working experience on individual income is more significant in the case of the regular chain system than the franchise system. Moreover, the marginal effect of working experience on individual income will decrease with increasing working experience. The speed at which this decrease occurs is faster in the case of the regular chain system as compared to the franchise system.

Finally, this study used the single level ordinal logit model to estimate the parameters. The estimation results suggest that the variable estimation coefficient and standard errors have no significant difference with intercepts-and-slopes-as-outcomes.

This implies the robustness of the research estimation results.

Table 4. Empirical results of the null model and the intercepts-as-outcomes model

Model Null model

Intercepts-as-outcomes model

Coefficient (se) Odds Ratio Coefficient (se) Odds Ratio

00 1.286

HOURM₇₀ 0.857

* p<0.10, ** p<0.05. In the part of fixed effects, ( ) is the robust error; in the random effects part, ( ) is the ² value, the null model’s degree of freedom is 53, and the intercepts-as-outcomes model’s degree of freedom is 51.

Table 5. Empirical results of the intercepts-and-slopes-as- outcomes

Model Intercepts and

slopes-as- outcomes

Ordinal logit model

Coefficient (se) Odds Ratio Coefficient (se) Odds Ratio

00 ^-3.094

AGE₄₀ 0.067

* p<0.10, ** p<0.05. In the part of fixed effects, ( ) is the standard error; in the random effects part, ( ) is the ² value and the model’s degree of freedom is 51.

6. Conclusion and suggestions