In this chapter, we examine the effect of exercise intensity and related variables on
mental health and depression. We begin our discussions of our results by presenting for
the case of the continuous and censored specifications of the mental health dependent
variables. The estimation results for our third binary dependent variable, whether each
individual is defined as depression or not, are exhibited in section 5.2.
5.1 The determinants on mental health
The estimation results of the fixed-effect OLS and fixed-effect Tobit model are
presented in Table 5. We report two sets of regression models: the first part is the results
of the fixed-effect OLS model, and the second part is the marginal effect of the
fixed-effect Tobit model. The estimated coefficients of the Tobit model are not reported
because they show no meanings of the magnitudes or quantitative effects of the
exogenous variables on mental health. In general, the estimated coefficients and implied
marginal effects are consistent for the two models, though quantitative differences do
observed.
The results, most noticeably, indicated that regular exercise dose significantly
improve elderly mental health. However, the intensity of exercise plays an important
33
role in determining mental health. In particular, evidence shows that the effect of light
exercise is more substantial than the effect of moderate exercise. The estimates indicate
that comparing to non-exercisers, light intensity exercisers lead to a lower CES-D score
by 0.62 point in the OLS model, and by 0.67 point in the Tobit model. On the other
hand, the effects are subtle for moderate intensity exercisers; they have a lower CES-D
score by about 0.3-0.4 point in these two models. This improvement is roughly half as
long as light intensity exercisers. One possible interpretation is that moderate intensity
exercise may be too much for maintaining the necessary health condition of the elderly.
Therefore, moderate intensity exercise is not as effective as light intensity exercise on
elderly health.
Among the other factors, family characteristics are also significantly associated
with mental health among the elderly. First, evidence points to a positive relationship
between marital status and mental health. Elderly with a spouse or partner have better
mental health by about 0.8 point in these two models. Second, living alone is negatively
associated with mental health. The estimated marginal effect is 0.66 in the Tobit model,
and a slightly higher effect is found in the OLS model (0.73). Health status also plays an
important role on mental health. Both OLS and Tobit results indicate that an additional
score of mobility impairment increases the CES-D score by 0.27 point.
Regarding to the two lifestyle indicators, current drinkers are more likely to have
34
better mental health, though the effect is only significant in the Tobit model. The
marginal effect of drinker implies that the CES-D scores of drinkers are 0.38 compared
to non-drinkers. As to the roles of social support, interestingly, we find a strong and
significant relationship between volunteer work and mental health. The estimated
coefficient in the OLS model shows that the CES-D scores of volunteers are reduced by
approximately 0.9 point. The quantitative size of this effect (1.2) is much larger in the
Tobit model. This likely reflects that volunteer involvement is beneficial to improve the
individual mental health.
Table 5 Estimation of continuous CESD score
Model
Fixed-Effect OLS Model Fixed-Effect Tobit Model
Variable Coefficient S. E. Marginal Effect S. E.
Note: S.E. = standard error. ***, ** and * indicated the significance at the 1%, 5%, and 10% levels.
a STATA reports the constant term as average value of the fixed effects.
35
5.2 The determinants on depression
Our third dependent variable, whether the individual has depressive symptoms or
not, is coded as a binary response. The estimates from the fixed-effect probit model are
shown in Table 6. In addition to the estimated coefficients, we also report the marginal
effects to quantify the effects of the exogenous variables on the probability of being
depressed.
Table 6 Estimation of the depression equation
Variable Coefficient S.E Marginal Effect S.E
Light Exercise -0.292 *** 0.111 -0.115 *** 0.044
Note: S.E. = standard error. ***, ** and * indicated the significance at the 1%, 5%, and 10% levels.
The results present a notable difference between light and moderate intensity
exercise on the likelihood of being depressed, which is different from the exercise
intensity effects on mental health. In the fixed-effect probit model, moderate intensity
36
exercise is not significant to determine the likelihood of depression, while light intensity
exercise leads to a lower probability of being depressed: 11.5% at the 1% level of
significance. This is a reasonable result, as those moderate intensity exercisers might
receive less benefits from exercise, which results in insufficient efforts to fighting
against depression. The empirical literature is inconclusive on the relationship between
exercise intensity and depression of the elderly. Our result is similar to Lindwall et al.
(2007), who found that light intensity exercise had a strong effect on elderly women, but
differs to Singh et al. (2005), who found high intensity PRT was more effective than low
intensity PRT.
Living alone appears to exert an influence on depression. The estimated marginal
effect indicates that elderly who live alone have a higher probability of being depressed
(23.6%). In addition, working status is associated with depression. Elderly who have a
current job are less likely to suffer from depression. On average, they have 12.4% lower
in probability of being depressed, holding other things constant.
As in mental health equation, health status is also significant in determining
psychological disorders. An additional score of mobility impairment increases the
likelihood of depression by 3.6%. Volunteer work can reduce depression as well. The
estimated marginal effect suggests that volunteers for social activities have a 16.2%
lower in probability of being depressed.
37