Religion as an investment: comparing the contributions and volunteer frequency among Christians, Buddhists, and folk religionists.

1. Introduction

The development of religious economics has taken place only very slowly, as pointed out by Iannaccone (1998). It took 200 years to produce the second religious economics publication (Azzi and Ehrenberg 1975) following the first by Adam Smith (1759). Azzi and Ehrenberg (1975) specified a mathematical model that makes allowance for afterlife consumption. Their model shows that the religious activities participated in by an individual increase with his non-labor income and age (if the salvation motive is more important than the consumption motive), and decrease with his market wage. The postulations of their theoretical model were verified by their empirical analysis. However, religious activities consist not only of allocating time, but also making monetary contributions. Sullivan (1985) simultaneously investigated the determination of church attendance and contributions. He verified the existence of a positive relationship between contributions and income, and a U-shaped relationship between income and church attendance.

Most religions promise a better afterlife. While the existing literature has verified the existence of a number of fundamental relationships between religionists' characteristics and their behavior in terms of attending church and contributing to church activities, these verifications have all been within the same religion, or at most, within the "different" religions with similar afterlife promises. Though different denominations were considered in Azzi and Ehrenberg (1975), those denominations were within the same religion, in particular in terms of the afterlife rewards they offered their believers. This then raises a question: Does a religion with a better afterlife reward induce its believers to be more fervent? In other words, ceteris paribus, will a religion with a higher afterlife return be more competitive in the religious market? The behavior of adherents of different religions with a distinct afterlife reward has not yet been investigated within the context of economic incentives. Indeed, if a religion that promises more rewards in the afterlife induces its adherents to devote more time and to contribute more money to their religion, the religion will be more sustainable and more able to attract adherents. If this is true, it will be argued that religious contributions and voluntary service may be regarded as an extended investment of "human capital" in the next life, rather than merely some psychological behavior. If this is so, in terms of attempting to explain religious activities, economics might be no less important than psychology or sociology. A positive answer to this question enables economics, in addition to psychology and sociology, to become an essential element in analyzing religious activities.

Furthermore, religionists with the same income or of similar age but with different degrees of belief in the existence of a promised afterlife are not likely to devote the same amount of time or contribute the same amount in terms of donations to their religions. In other words, the extent of belief in the existence of an afterlife is an important factor in investigating how the afterlife's rewards affect the religionists' behavior within the same religion. Unfortunately, this fact has been ignored in most existing studies. As argued in the last paragraph, the confirmation of different kinds of behavior in terms of contributions and volunteer frequency across religions demonstrates that religious activities are at least in part a reflection of an afterlife investment. This is like cross-sectional (across religions) evidence that economics is a useful tool as one analyzes religious activities. Similarly, if, within the same religion, religionists' contributions and volunteer frequency are strongly related to the extent of their beliefs, we will have "longitudinal" (within religion) evidence. If religionists do respond to economic incentives, this will imply that the fundamental microeconomics in which the present-life utility is maximized will have to be revised to take the afterlife into account, particularly in the case of those religionists. In regard to the authorities, when religious contributions and volunteer frequency are related to charitable purposes, for religionists to respond to an afterlife reward implies that an economy without religion or that is dominated by a religion with a low afterlife reward must spend more on social welfare programs.

Accordingly, the probability of believing in the existence of an afterlife and the diversity of religions, at least in terms of their afterlife rewards, are crucial when exploring the behavior of religionists. To sum up, unless we know the probabilities of religionists believing in the existence of an afterlife, or have data for religionists across different religions with distinct afterlife rewards, our studies are unable to completely verify whether the promised rewards of an afterlife affect the behavior of religionists. More importantly, we cannot ensure that economics is an essential element in explaining religious activities.

Moreover, rewards based on contributions and the time devoted to religion usually consist of the enormous but uncertain rewards in the afterlife and certain but not enormous rewards in the present life. The former, for example, promises an eternal life in the afterlife, while the latter provides the positive establishment of a social network and an improvement in family relationships. Are these rewards, being realized in the current life, also as important as the explanation of the religionists' behavior? Without taking these present-life rewards into account, the effects of afterlife rewards are likely to be overstated.

The purpose of this study is to investigate these questions. Protestants/Catholics, Buddhists, and Taiwanese folk religionists are chosen because each of these three kinds of religionists adhere to promised afterlife rewards that differ significantly from each other in terms of magnitude, and are also the most popular religions in the data-collecting area Taiwan. (1) According to John 3:16 in the Bible: "For God so loved the world that he gave his one and only Son, that whoever believes in him shall not perish but have eternal life." This passage clearly states that the afterlife reward for a Christian is eternal life after death. By contrast, the afterlife reward for a devotional Buddhist is only a better initial start to the next life. According to the doctrines of Buddhism, the death of a living being is not the end for this creature. The living being will take on the same or another life form, which is dependent on the performance of their present life. The rule is: "For a good cause, a good result; for an evil cause, an evil result." The Wheel of Life will repeat over and over again. (2)

In the case of Taiwanese folk religions, there are actually no substantial rewards in the afterlife resulting from religious behavior in the present life. (3) Folk religionists believe that the "soul" lives after a person's flesh dies. The soul becomes a good ghost that protects its descendants and offers good luck to them, provided that its living descendants worship it and provide offerings to it. (4) The soul turns into an evil ghost that attacks people if it does not have descendants who worship it and provide offerings to it. (5) The afterlife of a person does not depend on the person's religious behavior in the present life, but on whether it has filial descendants. However, in folk religions there are many different gods in different temples. These gods protect people and provide good luck to religionists who worship them. Hence, folk religionists worship their ancestors at home and worship the gods in temples. Therefore, folk religionists still have a motive to contribute to temples in return for a better life in the present. It is noteworthy that there is a lack of motivation on the part of folk religionists to improve their afterlife.

Obviously, the returns promised by these three religions are completely distinct, and, of course, eternal life in heaven (for Protestants/Catholics) has much greater significance than a better life form (for Buddhists) in the next life, and a better life form in the next life is better than nothing (for folk religionists). Therefore, it is expected that, given the same degree of devotionalism, Protestants/Catholics are more willing to devote more resources to their religion in their present lives, when compared with Buddhists and folk religionists in their present lives. Likewise, Buddhists are more willing to devote their resources to their religion than are folk religionists. This is the hypothesis that will be tested in this study. In addition, the extent of belief in the existence of an afterlife will also be tested.

The remainder of this paper is organized as follows. Section 2 formulates a theoretical model to explain how time and expenditure are allocated between the present life and investment in the afterlife. Section 3 describes the sources of data and the empirical model. Section 4 presents the empirical analysis. The time devoted and the contributions to the different religions and within the same religion will be explored. In addition, the effects of certain rewards in the present life and uncertain rewards in the afterlife will also be investigated. Since the religionists' behavior might, apart from being considered an investment in the afterlife, be a result of their doctrines and worship process, section 5 further verifies whether the behavior of the religionists is due to the religions' promise of an afterlife. Section 6 concludes.

2. The Theoretical Model

The utility function consists of two parts. The first part is the utility of the present life, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], which is determined by the resources allocated to activities other than religion as well as by those resources allocated to religious activities. The second part is the discounted expected utility of the afterlife. A rational religionist maximizes the sum of the present utility and discounted afterlife utility. That is, a rational religionist j who believes in religion i attempts to solve the following problem:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (1)

In Equation 1, subscript j denotes individual j, and [M.sub.j], [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] are total income, the income allocated to the present life, and the contribution to religion i, respectively. Likewise, T, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] are the total time available, the time allocated to activities other than religious activities, and the time allocated to religious activities, respectively. [L.sub.j] represents the expected years left for the present life, and r is the common discount rate. The [x.sub.j] are the characteristics of individual j. The [y.sub.j] denote the endowed resources in the afterlife for individual j, which are determined by his contribution and investment of time in the present life. This utility function reflects the belief on the part of religionists that input in terms of religious activities in the present life will enhance the utility of the afterlife.

R is the measure for the improvement in the religionists' current life because of religious activities. It is a function of the contributions and time devoted to these religious activities by these religionists. Many studies (see the survey by Iannaccone 1998) have concluded that religious activities improve people's current lives. The afterlife utility is an expected discounted utility that is dependent upon the extent of subjective belief in the existence of an afterlife, B. The second term in the objective function in Equation 1 can be rewritten as [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], where P([B.sub.j]) is the subjective probability of believing in the existence of an afterlife, and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is the discounted afterlife utility of individual j. The first-order conditions can then be easily derived as

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2a)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (2b)

The implications of Equations 2a and 2b are straightforward. The marginal utilities of resources (both income and time) in the present life that are allocated to religious activities (the left-hand side) are equal to the marginal utilities of resources allocated to nonreligious activities (the right-hand side). A large [partial derivative] [U.sub.p]/[partial derivative]R implies that religion greatly improves the individual's current life. All things being equal, because of decreasing marginal utility, it follows that religionists will contribute more or devote more time to having a small [partial derivative]R/[partial derivative][m.sub.f] in Equation 2a and a small [partial derivative]R/[partial derivative][t.sub.f] in Equation 2b.

Likewise, religionists who strongly believe in the existence of an afterlife (a higher P(B)) will be associated with a small [partial derivative][U.sup.*.sub.f]/[partial derivative][m.sub.f] in Equation 2a and a small [partial derivative][U.sup.*.sub.f]/[partial derivative][t.sub.f] in Equation 2b. In other words, they will tend to contribute more and devote more time to their religion. The rewards in the afterlife for Protestants/Catholics are much greater than those for Buddhists and folk religionists, given the same subjective probability P([B.sub.j]) and the same level of inputs of current resources ([m.sub.f] and [t.sub.f]). That is, for the same level of inputs of current resources for each of these religions, it is expected that [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], where c, b, and f denote Protestantism/Catholicism, Buddhism, and folk religion, respectively. To meet these first-order conditions, it is natural that, ceteris paribus, Protestants/Catholics will make more contributions and will devote more time to their religion than will Buddhists and folk religionists to their own religions.

3. Sources of Data and the Empirical Model

Sources of Data

The data used are obtained from the" 1999 Taiwan Social Change Survey (Year 5 of Cycle 3)" sponsored by the National Science Council of Taiwan, and compiled by the Institute for Social Sciences and the Office of Survey Research, Academia Sinica. The data were collected by means of face-to-face interviews and include two data sets--Culture and Value Judgment, and Religion. In this study, we focus on Part II of the data set (Religion). The ages of the sampled individuals range from 20 to 70 years old. There were 1925 observations in the original data set, but after deleting the missing data, there were 1617 observations. Since atheists and those who subscribe to minor religions are not used in this study, only 1278 observations remain after ruling out these other observations. The 1278 individuals comprise folk religionists/Taoists, Buddhists, and Protestants/Catholics, and make up the sample used in this study. The definitions of the variables and summary statistics are provided in Table 1.

All of these 1278 individuals will be used to analyze their religious behavior in regard to contributions, but folk religionists will not be used to analyze religious volunteer behavior (for reasons to be explained later). Therefore, a full sample of 1278 observations and also a subsample that excludes folk religionists are needed. Table 1 reveals the characteristics of both the full sample and the sub-sample.

[D.sub.F], [D.sub.B], and [D.sub.C] are the dummy variables for folk religionists, Buddhists, and Protestants/ Catholics, respectively. DEV is a dummy variable that denotes a self-rated devotionalism. It takes on a value of 1 if the respondent answered that he or she was a devotional religionist, and a value of 0 if he or she answered that he or she was not a devotional religionist. Individual characteristics (i.e., the [x.sub.j]) comprise the wage (WAGE), sex (male is 1), age, marital status (MAR, single is 0), years of education (EDU), and family income (FINC). (6) To investigate the kinds of behavior affected by rewards from religious activities, the independent variables in the empirical model include indexes of gains in the religionists' present life from religious activities and the extent to which they believe in the existence of an afterlife as promised in their religious creed. As to the former, questions 44, 45, and 46 in the questionnaire ask if the religion is helpful to the respondent's business/study, social network, and family, respectively. The answers "a lot of help" and "some help" are given the value 1, while the answers "not much help" and "no help" are denoted by 0. Three dummy variables, DHS, DHN, and DHF, which correspond to business/study, social network, and family, respectively, are specified. To explore whether the uncertain reward of the afterlife is important to religionists, a proxy for subjective belief ([B.sub.i], i = B or C) that denotes the existence of the afterlife is also incorporated into the empirical model. In the questionnaire, question 31 states: "Do you believe that there is a heaven and hell?" A devotional Christian must answer such a question affirmatively. Likewise, question 33 states: "Do you believe that people will reincarnate to the next life?" A devotional Buddhist must answer such a question with a resounding "yes." The answers given are classified according to the extent to which the belief is "strong belief," "some belief," "not much belief," or "little belief." They are, in that order, indexed with the numbers 3, 2, 1, and 0, respectively. The indexes imply that we assumed that the probabilities of believing in the existence of an afterlife in that same order were 75%, 50%, 25%, and 0%, respectively. [B.sub.B] and [B.sub.C] are proxies for the subjective beliefs of Buddhists and Protestants/Catholics, respectively. Note that there is no proxy for folk religionists, because heaven and reincarnation are not consistent with their creed.

Of the full sample, about 60%, 32%, and 9% are folk religionists, Buddhists, and Protestants/Catholics, respectively. It should be noted that the observation that the average of [B.sub.B] (the extent of Buddhists' beliefs in the afterlife) is larger than the average of [B.sub.C] (the extent of Protestants/Catholics' beliefs in the afterlife) might be misleading, because they are both calculated by the sum of their respective [B.sub.i], i = B or C, but divided by the total number of observations in the full sample and sub-sample, respectively. However, the number of Protestants/Catholics is much smaller than the number of Buddhists, so the sum of the former in the numerator when estimating the mean is much smaller than that of the latter. When their own number of observations is used as the denominator while calculating the average, the averages for [B.sub.B] and [B.sub.C] are 1.444 and 2.034, respectively. From this, it seems that Protestants/ Catholics are more likely to believe in the existence of an afterlife than Buddhists. This observation might prevail in all societies, but it also probably reflects the fact that Protestantism/Catholicism is a minor religion in Taiwan, and people with a strong belief in their religion will choose to be a minority in their society. (7) South Korea's case might lead us to believe that the former is the case. South Korea is a unique case in which Protestants/Catholics and Buddhists each account for about the same proportion of the population. Kim (2003) presents the religious population in South Korea in 1995 in his Table I. The proportions of Buddhists and Protestants/Catholics in relation to its population are 23.2% and 26.3%, respectively. In South Korea, Protestants/Catholics are certainly not a minority. However, a 1997 Gallup survey of South Korea showed that 44.9% of Buddhists believe in God, while 89.0% of Christians believe in God. It is likely that the higher degree of belief in the case of Taiwanese Protestants/Catholics is not due to their being a minority, but due to the religion itself. This issue will be discussed again later.

Of the sub-sample, about 78% of those surveyed are Buddhists. Most of the respondents think that their religion is helpful, especially for improving their family situation, the respective percentages in relation to this being the highest--around 64% and 75% for the full sample and sub-sample, respectively. Male and female respondents are more or less equally divided in both samples. The average age of the respondents is about 40 years old and the percentage of married respondents is around 79% in both samples. The mean of the respondents' years of education is 11 years. The average personal wage is around NT$29,000 in both samples. The average amounts of total family income for the sub-sample and the full sample are NT$69,000 and NT$64,000, respectively. For comparison purposes, statistics of non-religionists and minority religionists are also presented in the last column. It seems that individuals in this category are more likely to be younger, male, single, more educated, and richer.

Table 2 summarizes the behavior of those sampled in terms of making contributions and being a volunteer, and the current gains in the present life from religion. The contributions and the volunteer frequency in relation to each religion presented in Table 2 exactly coincide with the economic incentive previously mentioned. That is, religion which promises higher afterlife rewards seems to induce greater contributions and higher volunteer frequency. However, Table 2 also reveals that religion that promises higher afterlife rewards leads to greater gains in the religionists' present life. Specifically, Protestants/Catholics gain more from their religion in the present life than do Buddhists, and Buddhists gain more from their religion in the present life than do folk religionists. This result is consistent with South Korea. Kim (2003) found that Christians are more satisfied with life than are Buddhists. Therefore, at this stage, we cannot distinguish whether religionists' contributions and their behavior as volunteers are a reflection of economic incentives arising from afterlife rewards or are simply feedback resulting from gains derived in the present life. The importance of such gains to the present life validates the inclusion of the gains from religion in the present life in the empirical model.

Furthermore, one might argue that Table 2 reflects nothing but the order of the depth of religious foreignness, and is irrelevant to economic incentives in relation to afterlife return. Presumably, a stronger degree of commitment would be required to commit to a more foreign religion. Protestantism/Catholicism is, of course, the most foreign religion in Taiwan, while Taiwanese folk religion is the most indigenous religion. Recall that DEV is the dummy variable if a respondent is a devotional religionist. The degree of commitment of a respondent to his/her religion is controlled by this dummy variable. Moreover, [B.sub.B] and [B.sub.C] are proxies of subjective beliefs for the afterlives of Buddhists and of Protestants/Catholics, respectively. They might also partially reflect the degree of commitment of a respondent to his/her religion. Therefore, religious foreignness will not be entangled in the economic incentive brought about by an afterlife return in the empirical analysis, since DEV, [B.sub.B], and [B.sub.C] are all in the model. Again, South Korea's case shows that Protestants/Catholics are not more devotional because their religion is more foreign.

The Empirical Model

According to Equations 1, 2a, and 2b, the optimal contributions and time devoted to religion are simultaneously determined, and are functions of [m.sub.f], [t.sub.f], x, r, L, B, M, and T. Since r and T are the same for all individuals, they can be dropped from the simultaneous equations below. By adding the error terms, [epsilon] and u, in each equation, respectively, the empirical models are presented as

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (3a)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (3b)

where [[??].sub.f] and [[??].sub.f]r are instrumental variables for [m.sub.f] and [t.sub.f] respectively. It should be noted that to make a consistent comparison among religions with respect to the religionists' time devotion, [t.sub.f] only measures the frequency of voluntary activities devoted to their religions, excluding time spent weekly in worship or at a church service or Mass. This is because there is no corresponding ceremony for Buddhists and folk religionists. Moreover, different religions usually require their adherents to go to their church or temple more frequently or less frequently, and the time needed for a ceremony is usually different. (8) One might argue that ignoring the time devoted to religious activities other than being a volunteer for their church or temple might fail to take into account the degree of devotionalism of religionists, but is taken into consideration by the devotionalism dummy DEV in the regressions.

The data regarding the contributions ([m.sub.f]) and the time spent in being a volunteer on behalf of the religion ([t.sub.f]) are not the actual amounts of time but rather the intervals that the respondents choose. The midpoint of every interval is used. (9) Though ordered probit or other econometric methods might be appropriately applied, Wooldridge (2002) is highly supportive of the use of the midpoint. It is important to note that the regression model (Eqn. 3b) does not include a dummy for folk religionists. As shown in Table 2, the frequency of being a volunteer is zero for folk religionists. Therefore, there seems no reason to estimate the frequency of being a religious volunteer from the standpoint of folk religionists.

A two-stage Tobit model is applied both for the contributions and the voluntary frequency, since the contributions and voluntary frequency are simultaneously determined and cannot be less than zero, and there are observations gathered at zero.

Because the ritual of worshipping ancestors engaged in by folk religionists usually takes place at home, most folk religionists do not regularly go to temples. Unlike Protestants/ Catholics and Buddhists, it is not a common occurrence to see folk religionists volunteering for some temple. As expected, our data show that none of the sampled folk religionists were volunteers for any temples. Therefore, the simultaneity problem between contributions and volunteer frequency does not exist in the case of folk religionists. However, it is not uncommon for Protestants/Catholics and Buddhists to be volunteers for their churches or temples. This heterogeneity gives rise to complexity when using a two-stage Tobit model to estimate (Eqn. 3a) and (Eqn. 3b). When estimating (Eqn. 3a), our strategy is to estimate the instrumental variable [[??].sub.f] for Protestants/Catholics and Buddhists, and to assume that [[??].sub.f] = 0 for folk religionists. This method ensures that the contributions of these three kinds of religionists can be estimated together. By contrast, it is meaningless to estimate the volunteer frequency (Eqn. 3b) for folk religionists since it is known that there are no volunteers among the folk religionists. (10) Accordingly, folk religionists are excluded as (3b) is estimated.

4. Empirical Results and Comparisons

The regression results are shown in Table 3. (11) Before turning to the religious variables, we first discuss the results of the demographic variables. For the purposes of comparison, contribution regressions both without volunteer frequency (Model I) and with volunteer frequency (Model II) are analyzed. The coefficient of SEX is significantly negative, implying that females contribute and volunteer more than males. This might suggest that females are more risk averse about their afterlife, or more philanthropic than males. This result is consistent with the views of most fundraisers and policy-makers (see the introduction of Andreoni and Vesterlund 2001). It is also consistent with Rooney et al. (2005). (12) The coefficients of AGE are positive and strongly significant in all four models. This finding is consistent with our theoretical setup--the higher the age, the higher the marginal return on religious investment. Therefore, aging individuals are more willing to invest in their afterlife. Married individuals are no different from singles in terms of their volunteer frequency, but are more likely to contribute. It is possible that the family bond limits the time that couples have to engage in being religious volunteers, leading to an insignificant result. In Models I and IV, education is positive and significant at the 5% level, but is only close to being significant at the 10% significance level in Models II and III. The reduction in the significance of education in Model II probably reflects the positive correlation between education and volunteer frequency, and omitting volunteer frequency results in the overestimation of the effect of education on contribution. The low significance of education in Model III might be due to the collinearity between education and the wage since the standard deviation of education substantially declines, from 1.23 in Model III to 0.71 in Model IV (the standard errors are not shown in Table 3), as WAGE is dropped from Model IV. Both the wage and family income are positively significant in terms of contributions, but are insignificant in terms of volunteer frequency. The more one earns, the higher the cost of one's time. This might explain the insignificance of WAGE and FINC (family income) in relation to volunteer frequency. Additionally, the marginal utility of money for high-income individuals is low, and their marginal cost of time is high. The optimal way for them to invest in an afterlife is therefore to contribute more money, and not more volunteering time.

As already mentioned, the insignificance of both the wage (WAGE) and the contribution ([??]f) in relation to volunteer frequency is likely to be because a higher wage implies that the cost of volunteering more is higher, leading to a significantly negative coefficient for the wage. In a way that violates the intuition, the coefficient of the wage is not only insignificant, but is also positive. Since religionists with a higher wage are, other conditions being held equal, usually inclined to contribute more to their religions, it is possible that the wage and the contribution will give rise to a collinearity problem. Hence, in Table 3, Model IV drops the wage to see if collinearity is a possible problem, and the contribution turns out to be significant at the 10% level. The standard error of the coefficient of the contribution falls from 1.98 in Model III to 0.66 in Model IV (the standard errors are not shown in Table 3), which more or less verifies the possibility of a collinearity problem. As can be seen, the significant independent variables are the same in both Model III and Model IV, except for education.

We now turn to the religious variables. Let us first of all recall the first-order conditions (Eqn. 2a) and (Eqn. 2b). Since the ranking of the magnitude of the promised afterlife rewards is, in descending order, Protestants/Catholics, Buddhists, and folk religionists, it is expected that [[beta].sub.C] >[[beta].sub.B]>[[beta].sub.F] and [[alpha].sub.c], > [[alpha].sub.C]>[[alpha].sub.B].

If religionists are more sensitive to rewards promised in the afterlife, significantly positive [[beta].sub.BB], [[beta].sub.BC], [[alpha].sub.BB] and [[alpha].sub.BC] will be obtained. This implies that the more the religionists believe in the existence of an afterlife, the more they will contribute and devote themselves to their religions. Similarly, significantly positive [[beta].sub.i] and [[alpha].sub.i], where i = HB, HN, and HF, demonstrate that the gains in the present life from religion encourage religionists to contribute more and devote more voluntary time.

As expected, a significantly positive coefficient of DEV means that devotional religionists are inclined to contribute more. This conclusion is consistent with Regnerus, Smith, and Sikkink (1998). (13) In our empirical setup, the degree of devotionalism is given while religionists optimize the amount of their contributions and volunteer frequency. One might argue that the decision whether or not to be a devotional religionist could also be an endogenous decision. That is, given the afterlife rewards of different religions, religionists will decide whether or not to be devotional believers, and will optimize their contributions and volunteering frequency. Alternatively, it is also possible that participating religious volunteers might deepen their beliefs, thereby strengthening their devotionalism. In other words, the degree of devotionalism is also an endogenous variable. Iannaccone (1990) extended the model of Azzi and Ehrenberg (1975) with "religious human capital." He argued that the stock of religious experience derived from one's past religious activities is very specific. Since we agree with Iannaccone's (1990) point, we will agree that belief in the same religion is a long-term behavior, and the degree of devotionalism does not frequently vary. Equivalently, the degree of devotionalism to their religions is given at a certain time point when we observe religionists' contributions and volunteer frequency. To ensure that the degree of devotionalism is not an endogenous variable, an endogenity test is conducted. The years of belief is used as an instrumental variable for the degree of devotionalism, since the years of belief variable is certainly exogenous and is correlated with the degree of devotionalism. The chi-square value for the Wu-Hausman test is close to 0 for estimating contributions, and is 2.38 for volunteer frequency. The corresponding p-values with 1 degree of freedom are 0.99 and 0.12, respectively.14 The null hypothesis of no correlation between the degree of devotionalism and the error term cannot be rejected.

The coefficient of the dummy for Protestants/Catholics (Dc) is larger than that of the dummy for Buddhists ([D.sub.B]), and the coefficient of the dummy for Buddhists is larger than that of the dummy for folk religionists ([D.sub.F]). This is likely to imply that Protestants/Catholics contribute more than Buddhists, and that Buddhists contribute more than folk religionists, given that Protestants/Catholics, Buddhists, and folk religionists are at the lowest level in terms of believing that heaven or reincarnation exist.

The subjective probability of believing in the existence of reincarnation for Buddhists ([B.sub.B]) is not significant, while the subjective probability of believing in the existence of heaven for Protestants/Catholics ([B.sub.C]) is positively significant at the 1% level. The significance demonstrates, along with (Eqn. 2a), that promised rewards in the afterlife affect the way that Protestants/Catholics contribute. The insignificance of [B.sub.B] is likely to be a consequence of the Buddhists' low sense of return in relation to the afterlife.

Interestingly, the magnitudes of the coefficients of [B.sub.B] and [B.sub.C] are consistent with the order of their promised rewards of afterlife. To confirm that the promised rewards across religions affect the sizes of the contributions, Table 4 presents the differences in terms of contributions between Protestants/Catholics and Buddhists and their corresponding t values for both models, with the same level of subjective probability in terms of believing in the existence of an afterlife. The differences in volunteer frequency in Table 4 will be explained later. Consistent with our expectations, the higher the subjective probability, the larger and more significant will be the differences in contributions between Protestants/Catholics and Buddhists.

We have verified that the promised rewards of an afterlife affect the way in which religionists contribute to their respective religions. We now wish to investigate whether current rewards from religion also affect religionists' contributions. Recall that [D.sub.HB], [D.sub.HN], and [D.sub.HF] correspond to gains in business/study, in social networks, and in family relationships, respectively, due to religion. Both [D.sub.HN] and [D.sub.HF] are insignificant, but [D.sub.HB] is significant at the 5% level. It is interesting to find that when religion helps religionists make more money, they will directly contribute more to their religion in return. Table 3 also shows that devotional, female, old, and married religionists who have higher wages and higher family income, and who volunteer less ([??]f), tend to contribute more.

In terms of the volunteer frequency regression, Model III in Table 3 shows that both religious dummy variables are strongly significant, and that the coefficient of the dummy for Protestants/Catholics ([D.sub.C]) is larger than that for Buddhists ([D.sub.B]). It follows that Protestants/ Catholics are inclined to volunteer more than Buddhists. This result is also consistent with the conclusion that higher afterlife rewards induce more current investment in terms of either contributions or volunteer frequency. In addition, current gains in terms of improving the social network ([D.sub.HN]) and family relationships ([D.sub.HF]) from religion encourage religionists to volunteer more time to their religion. It seems that the gain resulting from improving the social network and family relationships induces religionists to devote more time to volunteering. It is very interesting to note that religionists choose a distinct way of paying for their appreciation for different types of present gain. When religion helps religionists in their business, they contribute more in return, while when religion helps them to build a good social network and family relationships, they tend to engage in more voluntary work in return.

The subjective probability of believing in the existence of an afterlife ([B.sub.B] and [B.sub.C]) is not significant, though as expected the coefficient of the Protestants/Catholics' subjective probability ([B.sub.C]) is larger than that of the Buddhists ([B.sub.B]). This means that the promised rewards of an afterlife are not as sensitive to behavior in terms of being a religious volunteer as they are to being a contributor. It could be that time is scarce and is equally limited to all religionists. There is no apparent boundary for the wage. Thus, the marginal utility for the last dollar of a high wage is not high. In such circumstances, people are more willing to contribute more if their subjective belief is higher. On the contrary, all people are endowed with the same amount of time. They do not have much more additional time to invest in their religion. This explains why religionists prefer to choose to contribute, instead of donating their time, in order to invest in their afterlife.

It is worth noting that both volunteer frequency and contributions are significantly negative in both the contribution regression and the volunteer frequency regression, respectively. This implies that religionists use both contributions and volunteer frequency as substitutes for each other. At first sight, it seems that our conclusion goes against that of Sullivan (1985), in which a complementary relationship between the amount of the contribution and the time spent in worship was found. However, in this study the volunteer frequency indicates the extra time devoted to religion, in addition to worship.

As a result of the significant subjective probability in terms of believing in the existence of an afterlife for Protestants/Catholics in the contribution regression, it is verified that the promised rewards of an afterlife matter within the same religion, at least in terms of their contribution behavior, as long as the promised rewards are sufficiently large. It seems that when the afterlife rewards are not sufficiently large--for example, in Buddhism--the subjective probability of believing in an afterlife is not an important factor in manipulating religionists' contribution behavior. This might be because the moderate afterlife rewards give rise to only limited differences between the high and low subjective probabilities of believing in the existence of an afterlife in terms of the expected returns in the afterlife. In addition, the results of the volunteer frequency regression suggest that even the afterlife rewards are huge for Protestants/Catholics. When their present resources, such as time, are highly constrained, they will not choose the highly constrained resources in the present life as their religious investment cost, but will choose those resources that are only slightly constrained, such as income, as their religious investment cost. In addition, Table 4 shows that the higher the subjective probability of believing in the existence of an afterlife, the greater will be the differences across different religionists' contributions and volunteer frequency. In Table 4, one might wonder why there exists a significant difference between Protestants/Catholics and Buddhists in volunteer frequency even when both kinds of religionists do not believe in an afterlife existence (the level of subjective probability is equal to 0). The significant difference in Table 4 is likely to reflect the difference in terms of present life gains between these two religions. Recall that religionists will devote more time to their religions if their religions improve their social networks and family relationships.

5. Further Confirmation

Protestantism/Catholicism, Buddhism, and Taiwanese folk religion certainly have their own unique doctrines. The findings that Protestantism/Catholicism induces more fervent contributors might be due to its doctrine and characteristics, in particular its view regarding philanthropy, rather than to its promise of an abundant afterlife. Both the doctrines of Protestantism/Catholicism and Buddhism emphasize philanthropy. The Bible states "love your neighbor as yourself" many times. (15) Taiwan's Buddhism is based on Mahayana Buddhism, where "Mahayana" means "great vehicle." That is, one of the central doctrines of Mahayana Buddhism is to benefit all people just as if one were taking all the people in the great vehicle to the Pure Land (Elysian Fields). (16) Kawamura (1998) thought that Buddhism's philanthropy was the closest to the one of no quid pro quo. (17) Buddhism strongly renounces egotism, and this is equivalent to stressing philanthropy. (18)

The Bible, in fact, proposes that an exact proportion (one tenth) of its believers' income be donated to their organizations. (19) The exact donation suggestion does not guarantee that Christian organizations will necessarily raise more contributions than Buddhist organizations. (20) Though there is no exact income proportion proposed for contributing to Buddhist organizations, this does not mean that Buddhists must contribute less than Protestants/ Catholics. Recently, some Taiwanese Buddhism foundations revealed their strong fundraising capability. For example, the famous Tzu Chi Foundation, based in Taiwan, which has branches and helps poor people all over the world, received contributions amounting to more than US$150 million in 1993. (21)

These discussions seem to imply that Protestantism/Catholicism and Buddhism have similar perspectives in regard to philanthropy. Surprisingly, two of Buddhism's leaders (Thich Nhat Hanh 1995; Bstan-dzin-rgya-mtsho and Dalai Lama XIV 1996) and one Christian writer and researcher (Borg 1999) postulate that both religions have very close creeds, and each of them has written or edited a book to advocate their respective viewpoints. As stated by Borg (1999): "Thus, despite differences in language and imagery, the way taught by the Buddha and the way taught by Jesus strongly resemble one another." Therefore, as one compares the contribution and volunteering behavior of Protestants/Catholics and Buddhists, it is reasonable to agree that the doctrines of these two religions in relation to giving are not the key to differentiating between the contribution and volunteering behavior of adherents of both religions.

Protestantism/Catholicism is monotheistic, while Buddhism and Taiwanese folk religion are polytheistic. However, this does not necessarily mean that a religion that is monotheistic must be characterized by more moral restraint. Both Protestantism/Catholicism and Buddhism possess formal written doctrines, while folk religion does not. Though not all devotional Buddhists in the world are vegetarians, devotional Taiwanese Buddhists have to be vegetarian. They practice the strict Buddhist doctrine in their daily lives.

However, in the previous discussion, we have pointed out that even though Protestants/ Catholics are not a minority in South Korea, Protestants/Catholics in South Korea still have a much stronger belief in God than Buddhists. It seems that something else, other than the creed of giving, is the key to manipulating the believers' behavior. In spite of similar doctrines on giving between Protestantism/Catholicism and Buddhism, the folk religion doctrine is very different from the other two. The folk religion doctrine does not emphasize philanthropy. In addition, the gods in different religions are defined as having different capabilities. Christ is omnipotent, the King of the universe, and is above nature. By contrast, Buddha is a god in the universe and in nature. It seems that the power of Christ is greater than that of Buddha. Christ created the world, but Buddha was created in the world. The omnipotent Christ might be a reason why Protestants/ Catholics are more devotional than Buddhists. Compared to Protestants/Catholics, Buddhists and folk religionists do not have regular worship, and do not have group worship either. Consequently, interactions among Protestants/Catholics are far greater than those among Buddhists and folk religionists. The group worship of some Christian denominations strongly portrays their feelings toward God and toward other Christians. Because of these frequent and emotional interactions either among the believers themselves or between individual believers and God, Protestants/Catholics usually strongly identify themselves with Christian groups. The regular, group, and interactive worship among Protestants/Catholics might be another reason why Protestants/Catholics are more devotional than Buddhists. Iannaccone (1998) referred to these effects as "churches as clubs." Furthermore, folk religionists have less strict moral codes than Protestants/Catholics and Buddhists. However, it is difficult to measure the strictness of moral codes and the differences in God's capability, or to measure how the process of worship affects either the contributions or volunteering engagements of religionists. Nevertheless, if these factors affect the extent of the devotionalism of the religionists and of their beliefs in heaven or reincarnation, their effect will be captured by the corresponding independent variables (DEF, [B.sub.B], and [B.sub.C]) already included in the model.

Azzi and Ehrenberg (1975) concluded that median age was positively correlated with the church membership rate using their 1952 data set, their most confident data set. Sullivan (1985) found that weekly contributions to churches on the part of women significantly increased until the age of 73. On the contrary, Duncan (1999) found that age was not a significant factor for charitable contributions. Studies, for example, Hrung (2004), Yen (2002), and Reece (1979), which simultaneously investigated both charitable and religious contributions, found intriguing results regarding the age factor. (22) To sum up, most of the evidence showed that age was significantly and positively related to religious contributions, but was an insignificant factor as far as charitable contributions were concerned. (23) The point is also supported by Chang (2005), who used the 2003 Taiwan Survey of Social Development Trends, conducted by the Directorate-General of Budget, Accounting, and Statistics, to examine the relationships between age and different types of giving. He concluded that the positive relationship between age and the level of religious giving was stronger than that for all the other types of giving. These findings are consistent with our own. Equation 1 in the present study implies that the contributions of religionists will increase with age since the marginal expected utility of religious contributions increases with age. This can be seen as we rewrite the marginal expected utility in Equation 2 as follows

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (4)

[L.sub.j] in Equation 4 is years of life left, and is equal to expected life minus age. The higher the age, the lower the [L.sub.j] and the higher the marginal utility. It is important to note that Equation 4 is only implied by religious contributions, and is not implied by charitable contributions other than in religious matters.

Accordingly, "age" will be a useful factor for further verifying that our conclusions regarding religious contributions and volunteer frequency across religions are appropriate. We can rewrite the age variable in Equations 3a and 3b as an age x religion dummy variable; that is, AGE x DF, AGE x [D.sub.B], and AGE x [D.sub.c]. If Equation 4 truly depicts the behavior in terms of the individual's contribution to religion, it will be observed that religion with a higher afterlife reward will be associated with a higher marginal contribution with respect to age. That is, the coefficient of AGE x [D.sub.C] is the greatest, that of AGE x [D.sub.F] is the smallest, and that of AGE x De lies in between. Model IIA corresponds to Model II, and Model IVA corresponds to Model IV. The regression results are presented in Table 5. Except for marriage status (MAR), the levels of significance of the independent variables in Tables 3 and 5 are similar. It seems that marriage status explains some of the effect of age on contributions when the age effects are assumed to be the same for all religions. The order of the magnitude of the coefficients of these three age x religion dummies is as expected. For religious contributions, the coefficients of AGE x [D.sub.C] ([[beta].sub.AC), AGE x [D.sub.B] ([[beta].sub.AB]), and AGE x [D.sub.F] ([beta].sub.AF]) are 0.89, 0.39, and 0.23, respectively. For volunteer frequency, the coefficients of AGE x [D.sub.c] and AGE x DR are 0.59 and 0.51, respectively. We further test the differences between any two coefficients for the age x religion dummies in the same regression. For religious contributions, the t-values of [[beta.sub.AC] --[[beta].sub.AB], [[beta].sub.AB]-[[beta].sub.AF], and [[beta].sub.AC] - [[beta].sub.AF] are 2.59, 4.00, and 3.50, respectively. All three t-tests are significant at the 1% level. For volunteering, the t value of [[beta].sub.AC] - [[beta].sub.AB] is 0.23, and is not significant, probably due to the physical deterioration of all religionists through aging. The coefficients of the age x religion dummy in the contribution regression further confirm our conclusion regarding the afterlife investment behavior across religions.

Although in Table 5 the order of the coefficients of [D.sub.C], [D.sub.B], and [D.sub.F] is reversed, it does not violate our early expectations. For comparisons between the different religionists' contributions and volunteer frequency one needs to consider both the coefficients of the religion dummy and the age x religion dummy. At an age equal to 40 (about the mean), the marginal contributions for Christian, Buddhist, and folk religionists are -9.19, -13.69, and -16.15, respectively. The order of the marginal contributions for different religions still meets our early expectations.

6. Conclusions

This study compares the behavior in terms of making contributions and being a religious volunteer both across religions and within the same religion in the cases of Protestantism/ Catholicism, Buddhism, and Taiwanese folk religions in Taiwanese society. It is clear that the Bible claims to promise eternal life in the afterlife to Protestants/Catholics, while the creed of Buddhism promises only a good initial condition in the next life. There is no promised reward of an afterlife for Taiwanese folk religionists. Without a doubt, the promised rewards in descending order of importance are those that relate to Protestantism/Catholicism, Buddhism, and Taiwanese folk religions. Both the contribution and volunteer frequency behavior of different religionists correspond to the order of the rewards in the afterlife promised by these different religions. It is found that a religion promising higher afterlife rewards tends to induce more contributions and volunteer frequency than other religions. It seems that the behavior of religionists in terms of making contributions and being a volunteer corresponds exactly with economic incentives. This implies that maximizing present-life utility cannot sufficiently depict religionists' behavior. Afterlife utility must be taken into account to fully capture human behavior, especially for religionists.

In addition, religionists with different subjective probabilities of believing in the existence of an afterlife do not make significantly different contributions or volunteer differently, except in the case of the contribution behavior of religionists with sufficiently large afterlife rewards. This is because only huge afterlife rewards can give rise to significantly distinct expected returns from religious investments in the case of religionists with different levels of belief in an afterlife existence. This can explain why contribution behavior is not significantly different among religionists within the same religion (Buddhism) with moderate afterlife rewards. The insignificantly different behavior in terms of being a volunteer among religionists with different subjective probabilities in the same religion, regardless of how large or small the afterlife rewards are, is likely to be due to an equal time constraint affecting all religionists. It is also possible that the difference in terms of the afterlife between a religion with a high return (Protestantism/Catholicism) and one with a low return (Buddhism) is greater than the difference in the expected afterlife rewards of the same religion's believers (Protestants/ Catholics) with different subjective probabilities of believing in the existence of an afterlife. Thus, we observe significantly different behavior between Protestants/Catholics and Buddhists in terms of being a volunteer, but we do not observe significant differences among Protestants/ Catholics with different subjective beliefs in the afterlife.

It is also found that current gains from religion in daily life (business/study, social networks, and family relationships) encourage religionists to return more to their religion. Therefore, the amounts of contributions and the frequency with which they volunteer are also dependent on how many current gains religionists have received in the present life. This implies that part of the religionists' contributions and their being a volunteer serves as a kind of payoff in terms of what they have gained from their religion in the present life, and part is related to their investment in a better but uncertain afterlife. It is interesting to note that religionists contribute more in return if their religion helps their businesses, and engage in more volunteer work in return if their religion helps them to build a good social network and family relationships.

Furthermore, it is also possible that different religions possess distinct worship processes, organizing capabilities or inclinations towards centralization in regard to their gods or religious leaders. The religionists' contributions and their being volunteers might be a reflection of these factors, in addition to the promised rewards of an afterlife. However, these institutional factors are difficult to quantify and measure. Investigating the marginal contribution to religion with respect to age circumvents these difficulties. The results are consistent with our expectations, and also make our conclusions more reliable.

Religious belief is a very subjective perception. This is why in the same society a religion with conspicuously promised afterlife rewards does not necessarily destroy other religions with only moderately promised afterlife rewards. However, we still do not know, ceteris paribus, whether a religion with stronger beliefs in terms of promised rewards in the afterlife is more competitive in the religious market. To explore this issue more fully, data collected from a multireligious society would be needed. Furthermore, the promised afterlife rewards of these religions must surely differ significantly. However, nowadays, most societies are dominated by a single religion or similar religions.

Received January 2005; accepted March 2006.

References

Andreoni, James, and Lise Vesterlund. 2001. Which is the fair sex? Gender differences in altruism. Quarterly Journal of Economics 116:293-312.

Azzi, Corry, and Ronald Ehrenberg. 1975. Household allocation of time and church attendance. Journal of Political Economy 84:27-56.

Bolton, Gary E., and Elena Katok. 1995. An experimental test for gender differences in beneficent behavior. Economics Letters 48:287-92.

Borg, Marcus. 1999. Jesus and Buddha: The parallel sayings. Berkeley: Seastone, Ulysses Press.

Bstan-dzin-rgya-mtsho, Dalai Lama XIV. 1996. The good heart: A Buddhist perspective on the teachings of Jesus. Somerville: Wisdom Publications.

Chang, Wen-Chun. 2005. Religious giving, nonreligious giving, and after-life consumption. Topics in Economic Analysis & Policy 5:1-31.

Clain, Suzanne H., and Charles E. Zech. 1999. A household production analysis of religious and charitable activity. American Journal of Economics and Sociology 58:923-46.

Daniels, Peter L. 2005. Economic systems and the Buddhist world view: The 21st century nexus. Journal of SocioEconomics 34:245-68.

Duncan, Brian. 1999. Modeling charitable contributions of time and money. Journal of Public Economics 72:213-42.

Hrung, Warren B. 2004. After-life consumption and charitable giving. American Journal of Economics and Sociology 63:731-45.

Iannaccone, Laurence R. 1990. Religious practice: A human capital approach. Journal for the Scientific Study of Religion 29:297-314.

Iannaccone, Laurence R. 1998. Introduction to the economics of religion. Journal of Economic Literature 36:465-1496.

Johnston, Jack, and John DiNardo. 1997. Econometric methods. 4th edition. New York: McGraw-Hill Company.

Jordan, David K. 1985. Gods, ghosts & ancestors: Folk religion in a Taiwanese village. 2nd edition. Taipei: Caves Books Ltd.

Kawamura, Leslie S. 1998. The Mahayana Buddhist foundation for philanthropic practice. In Philanthropy in the world's traditions, edited by Warren F. Ilchman, Stanely N. Katz, and Edward L. Queen, II. Bloomington, IN: Indiana University Press.

Kim, Andrew E. 2003. Religious influences on personal and societal well-being. Social Indicators Research 62:149-70.

Reece, William S. 1979. Charitable contributions: New evidence on household behavior. American Economic Review 69:142-51.

Regnerus, Mark D., Christian Smith, and David Sikkink. 1998. Who gives to the poor? The influence of religious tradition and political location on the personal generosity of Americans toward the poor. Journal for the Scientific Study of Religion 37:481-93.

Rooney, Patrick M., Debra J. Mesch, William Chin, and Kathryn S. Steinberg. 2005. The effect of race, gender, and survey methodologies on giving in the US. Economics Letters 86:173-80.

Sider, Ronald J. 1999. Dare to give: Will we ignore the world's poor and Jesus' words on materialism? Moody Jan/ Feb:34-5.

Smith, Adam. 1759. 2000. The theory of moral sentiments. Amherst: Prometheus Books.

Sullivan, Dennis H. 1985. Simultaneous determination of church contributions and church attendance. Economic Inquiry 23:309-20.

Takakusu, Junjiro. 1956. The essentials of Buddhist philosophy. 3rd edition. Honolulu: Office Appliance Co.

Thich Nhat Hanh. 1995. Living Buddha, living Christ. New York: Riverhead Books.

Tomes, Nigel. 1984. The effects of religion and denomination on earnings and returns to human capital. Journal of Human Resources 19:472-88.

Wooldridge, Jeffrey M. 2002. Econometric analysis of cross section and panel data. Cambridge: The MIT Press.

Yen, Steven T. 2002. An econometric analysis of household donations in the USA. Applied Economics Letters 9:837-41.

Yokota, John. 1987. Shin Buddhist ethics. The Pure Land, New Series 4:18-19.

Zaleski, Peter A., and Charles E. Zech. 1992. Determinants of contributions to religious organizations: Free riding and other factors. American Journal of Economics and Sociology 51:459-72.

(1) In this empirical study, we combine Protestants and Catholics into one group (Christians) since they both promise exactly the same afterlife rewards. Neither Protestantism nor Catholicism is a dominant religion in Taiwan; therefore, combining Protestants and Catholics has the advantage of increasing the number of observations in this group.

(2) For more details, see Takakusu (1956).

(3) Here we combine folk religionists and Taoists since their beliefs are similar and most Taiwanese cannot distinguish one from the other. In our sample, there are 551 folk religionists and 212 Taoists.

(4) The spirit of folk religion explains why in tradition most Chinese or Taiwanese think that couples without children have been severely cursed.

(5) For more details on Taiwanese folk religions, see Jordan (1985).

(6) Income and contributions are measured in NT$'000s, where NT$ denotes the New Taiwan dollar, and US$1 is around NT$32 to NT$33.

(7) The means (standard deviations) for the degree of devotionalism for Protestants/Catholics and Buddhists are 0.699 (0.461) and 0.600 (0.491), respectively. Though the Protestants/Catholics have a greater mean and a smaller standard deviation than the Buddhists, the former is not significantly greater than the latter. Likewise, the means (standard deviations) of the belief in the existence of heaven for Protestants/Catholics or reincarnation to the next life for Buddhists are 2.027 (1.081) and 1.448 (1.089), respectively. Though Protestants/Catholics have a greater mean and a smaller standard deviation than Buddhists, again, the former is not significantly greater than the latter.

(8) The questionnaire asks Protestants/Catholics how often they participate in religious activities, including worship, Mass, and volunteering. However, the questionnaire only asks Buddhists and other religionists how often they voluntarily help in their temple activities.

(9) The largest interval does not have an upper limit. It is assumed that there is an equal range with an adjacent interval so to specify a midpoint.

(10) Even if the folk religionists are included in estimating the time expended in a voluntary capacity, the Tobit model does not converge.

(11) Age squared is also tried, but it leads to both age and age squared being insignificant in the regression. The standard error of the coefficient of age in Model II with age squared is more than five times that in the same model without age squared. It is likely that multicollinearity creates the insignificance of both age and age squared.

(12) Some experimental studies reveal mixed or more complicated results. For example, Bolton and Katok (1995) found no evidence for gender difference in relation to generosity, and Andreoni and Vesterlund (2001) concluded that "when altruism is expensive, women are kinder, but when it is cheap, men are more altruistic." However, gender difference is not the focus of the present paper, and so the discussion does not proceed further here.

(13) Regnerus, Smith, and Sikkink (1998) concluded: " ... the degree of importance to which people hold their religious beliefs also matters" (pp. 490).

(14) All of the procedures adopted for the Wu-Hausman test follow pp. 257-259 in Johnston and DiNardo (1997).

(15) For example, Mark 12:29-31: '"The most important one,' answered Jesus, is this: 'Hear, O Israel, the Lord our God, the Lord is one. Love the Lord your God with all your heart and with all your soul and with all your mind and with all your strength.' The second is this: 'Love your neighbor as yourself.' There is no commandment greater than these."

(16) The other is Hinayana Buddhism. "Hinayana" means "little vehicle." Simply put, "little vehicle" means taking oneself to the Pure Land.

(17) Kawamura (1998) cited Samyukta Nikaya (The book of the kindred sayings) (I. 105) to support her point: "Monks, go and travel around for the welfare of the multitudes, for the happiness of the multitudes, out of sympathy for the world, for the benefit, welfare, and happiness of gods and humans" (pp. 98-99). Kawamura (1998) also quoted Yokota (1987) to support the point: "Mahayana Buddhist social ethics is based on the reality of emptiness and specifically on the implications of the reality of no-self and no possessions. The basis of all ethical theory in Buddhism, especially from Mahayana perspectives, is compassion" (pp. 100).

(18) We quote Daniels (2005) to depict how the renunciation of egotism is equivalent to stressing philanthropy in Buddhism: "Hence, two obvious economic behavioral implications of Buddhist philosophy include the frugality and moderation of material consumption of the Middle Way and the renunciation of self-interest in following the Eight-Fold Path. This diversion from self or "I-ness" would entail new social definitions of status and directives for action that are no longer based on greed and material possession devoid of the implications for the welfare of others" (pp. 249).

(19) In Malachi 3:10: "Let your tenths come into the store-house so that there may be food in my house, and put me to the test by doing so, says the Lord of armies, and see if I do not make the windows of heaven open and send down such a blessing on you that there is no room for it."

(20) Sider (1999) mentioned: "In 1968, typical church members gave about one third of a tithe (3.14% of their income). It has been dropped virtually every year. By 1995, typical church members gave less than a quarter of a tithe (2.46%)" (pp. 34).

(21) The Foundation did not release the amounts of contributions after 1993.

(22) The results of Reece (1979) are from equation 6 and equation 7 of his Table 2. The former has to do with charitable contributions, while the latter is concerned with religious contributions.

(23) A contrary result is Zaleski and Zech (1992). They found that age is not a significant factor for explaining religious contributions (Equation 1 in Table 5). However, they did not use micro data, but used data for 180 congregations. The "age" they used was median age in relation to zip code, and the age's variance was 4.302, which was much smaller than the variance for the micro data usually used. The insignificance of age could be due to the invariance of age. According to Clain and Zech (1999), age is not considered to be significant for religious contributions, volunteering time and attendance at religious services. However, they admitted that the insignificance of age in relation to attendance at religious services was not consistent with previous studies (their endnote 20).

Hung-Lin Tao * and Powen Yeh ([dagger])

* Department of Economics, Soochow University, 56, Kuei-Yang St., Sec. 1, Taipei 100, Taiwan; E-mail [e-mail]; corresponding author.

([dagger]) Department of Finance, Ching-Yun University, 229, Chien-Hsin Rd., Jung-Li 320, Taiwan, E-mail [e-mail].

The data analyzed in this paper were collected as a result of the research project entitled "1999 Taiwan Social Change Survey (Year 5 of Cycle 3)" sponsored by Taiwan's National Science Council, and compiled by the Institute for Social Sciences and the Office of Survey Research, Academia Sinica. The authors appreciate the assistance of the institutes in providing data. Financial support from Taiwan's National Science Council is gratefully acknowledged (NSC93-2415-H-031-004). The authors would like to thank two anonymous reviewers, and in particular the Journal's coeditor, Julie L. Hotchkiss, for their very helpful comments. Of course, all errors are our own.

Table 1. Definition of Variables and Summary Statistics of Samples and
Non-Religionists

Variable                              Definition

[D.sub.F]     Dummy variable, 1 if the respondent is a Folk religionist
                or Taoist.
[D.sub.B]     Dummy variable, 1 if the respondent is a Buddhist.
[D.sub.C]     Dummy variable, 1 if the respondent is a Protestant or
                Catholic.
DEV           Dummy variable, 1 if the respondent is a devotional
                religionist.
[B.sub.B]     0,1,2,3, in ascending order, represent the extent to which
                the Buddhist respondent believes that people will
                reincarnate to the next life.
[B.sub.C]     0, 1,2,3, in ascending order, represent the extent to
                which the Protestant or Catholic respondent believes
                that heaven exists.
[D.sub.HB]    Dummy variable, 1 if the respondent thinks religion is
                helpful to his/her business or study.
[D.sub.HN]    Dummy variable, 1 if the respondent thinks religion is
                helpful to his/her social network.
[D.sub.HF]    Dummy variable, 1 if the respondent thinks religion is
                helpful to his/her family.
SEX           Dummy variable, 1 if the respondent is male.
AGE           The respondent's age.
MAR           Dummy variable, 1 if the respondent is married.
EDU           The respondent's years of education.
WAGE          Monthly wage of the respondent (NT$'000s).
FINC          Total income of the respondent's family. (NT$'000s).
Observations

                     Full Sample (a)       Sub-Sample (a)

                             Standard               Standard
Variable             Mean    Deviation      Mean    Deviation

[D.sub.F]            0.597       0.491        --

[D.sub.B]            0.315       0.465      0.781       0.414
[D.sub.C]            0.088       0.284      0.219       0.414

DEV                  0.534       0.499      0.621       0.486

[B.sub.B]            0.455       0.908      1.130       1.133

[B.sub.C]            0.179       0.659      0.445       0.980

[D.sub.HB]           0.505       0.500      0.590       0.492

[D.sub.HN]           0.495       0.500      0.637       0.481

[D.sub.HF]           0.638       0.481      0.751       0.433

SEX                  0.506       0.500      0.464       0.499
AGE                 40.196      11.313     40.350      10.977
MAR                  0.793       0.405      0.786       0.410
EDU                 10.674       3.961     11.322       3.766
WAGE                28.736      29.813     29.369      30.600
FINC                64.202      44.063     68.961      45.561
Observations         1278                    515

                      Non-Religious or
                    Minority Religionists

                             Standard
Variable             Mean    Deviation

[D.sub.F]

[D.sub.B]
[D.sub.C]

DEV

[B.sub.B]

[B.sub.C]

[D.sub.HB]

[D.sub.HN]

[D.sub.HF]

SEX                  0.513       0.501
AGE                 37.853      11.780
MAR                  0.705       0.457
EDU                 12.307       3.458
WAGE                33.186      34.175
FINC                75.708      49.172
Observations          339

(a) The full sample includes Folk religionists, Buddhists, and
Protestants/Catholics, while the sub-sample only includes Buddhists
and Protestants/Catholics.

Table 2. Contributions, Volunteer Frequency, and Helps in the Present
Life for Believers

                                                       Volunteer
                                       Contribution    Frequency
                        Observations   (NT dollars)   (Times/Year)

Protestants/                113          11,668          11.84
  Catholics
Buddhists                   402            6481           2.39
Folk Religionists/          763            2579           0
  Taoists
Total observations/        1278            4610           1.8
  Mean

                         Business/        Social         Family
                         Study (a)       Network       Relations

Protestants/                2.12           2.37           2.39
  Catholics
Buddhists                   1.32           1.44           1.71
Folk Religionists/          1.17           1.11           1.40
  Taoists
Total observations/         1.30           1.32           1.58
  Mean

(a) The scale of religious help in regard to business, social networks,
and family relationships ranges from 0 to 3. That is, no help = 0, not
much help = 1, some help = 2, and a lot of help = 3.

Table 3. Tobit Regression Results of the Contributions and Volunteer
Frequency

                                            Contributions

                                          Without Volunteer
                                          Frequency (Model I)

Variable                                Coefficient    t-Ratio

[D.sub.F]                                   -32.19    -9.40 ***
[D.sub.B]                                   -30.42    -7.77 ***
[D.sub.C]                                   -29.97    -6.19 ***
DEV (devotionalism)                           4.18     3.89 ***
[B.sub.B] (belief in reincarnation)           0.89     1.08
[B.sub.C] (belief in heaven)                  4.07     2.69 ***
[D.sub.HB] (help in business)                 2.73     2.19 **
[D.sub.HN] (help in social network)           1.41     1.04
[D.sub.HF] (help in family relations)         1.85     1.32
SEX                                          -2.54    -2.22 **
AGE                                           0.32     5.34 ***
MAR                                           3.21     2.08 **
EDU                                           0.44     2.58 **
WAGE                                          0.10     4.21 ***
FINC (family income)                          0.05     3.48 ***
[[??].sub.f]
[m.sub.f]
Sigma                                        16.83    41.94 ***
Observations                               1278
Log-likelihood                          -10,287.59

                                             Contributions

                                        With Volunteer Frequency
                                              (Model II)

Variable                                Coefficient   t-Ratio

[D.sub.F]                                   -28.11    -7.38 ***
[D.sub.B]                                   -25.01    -5.55 ***
[D.sub.C]                                   -20.6     -3.31 ***
DEV (devotionalism)                           4.37     4.06 ***
[B.sub.B] (belief in reincarnation)           0.95     1.16
[B.sub.C] (belief in heaven)                  5.06     3.24 ***
[D.sub.HB] (help in business)                 2.44     1.95 *
[D.sub.HN] (help in social network)           1.91     1.40
[D.sub.HF] (help in family relations)         1.94     1.39
SEX                                          -3.05    -2.63 ***
AGE                                           0.26     3.99 ***
MAR                                           3.31     2.16 **
EDU                                           0.28     1.53
WAGE                                          0.10     4.30 ***
FINC (family income)                          0.05     3.42 ***
[[??].sub.f]                                 -0.50    -2.39 **
[m.sub.f]
Sigma                                        16.77    41.94 ***
Observations                               1278
Log-likelihood                          -10,284.75

                                          Volunteer Frequency

                                        With Income (Model III)

Variable                                Coefficient   t-Ratio

[D.sub.F]
[D.sub.B]                                  -99.14     -4.71 ***
[D.sub.C]                                  -69.36     -3.19 ***
DEV (devotionalism)                         14.96      2.00 **
[B.sub.B] (belief in reincarnation)          1.35      0.70
[B.sub.C] (belief in heaven)                 5.03      0.89
[D.sub.HB] (help in business)                7.27      1.30
[D.sub.HN] (help in social network)          9.65      1.65 *
[D.sub.HF] (help in family relations)       17.11      2.44 **
SEX                                        -12.03     -1.69 *
AGE                                          0.55      2.84 ***
MAR                                         12.76      1.08
EDU                                          2.01      1.63
WAGE                                         0.08      0.34
FINC (family income)                         0.04      0.89
[[??].sub.f]
[m.sub.f]                                   -1.86     -0.94
Sigma                                       26.14     14.73 ***
Observations                               515
Log-likelihood                            -739.17

                                           Volunteer Frequency

                                        Without Income (Model IV)

Variable                                Coefficient   t-Ratio

[D.sub.F]
[D.sub.B]                                  -94.22     -6.26 ***
[D.sub.C]                                  -63.96     -4.38 ***
DEV (devotionalism)                         12.89      2.99 **
[B.sub.B] (belief in reincarnation)          1.34      0.70
[B.sub.C] (belief in heaven)                 3.40      1.15
[D.sub.HB] (help in business)                6.07      1.39
[D.sub.HN] (help in social network)          8.77      1.67 *
[D.sub.HF] (help in family relations)       16.11      2.54 **
SEX                                         -9.91     -2.91 ***
AGE                                          0.54      2.83 ***
MAR                                          9.38      1.46
EDU                                          1.67      2.34 **
WAGE
FINC (family income)                         0.04      0.91
[[??].sub.f]
[m.sub.f]                                   -1.23     -1.84 *
Sigma                                       26.12     14.74 ***
Observations                               515
Log-likelihood                            -739.23

* Significant at 10% level.

** Significant at 5% level.

*** Significant at 1% level.

Table 4. Annual Contributions and Volunteer Frequency Differences
between Protestants/Catholics and Buddhists, by Subjective
Probability of Believing in the Existence of an Afterlife

 Level of           Contribution Difference (NT$1000)
Subjective
Probability    Model I    t-Value     Model II    t-Value

     0          0.44      0.12           4.42     1.07
     1          3.63      1.46           8.52     2.66 ***
     2          6.81      3.55 ***      12.63     4.07 ***
     3          9.99      3.74 ***      16.74     4.31 ***

 Level of       Volunteer Frequency Difference (times/year)
Subjective
Probability   Model III   t-Value     Model IV    t-Value

     0          29.78     4.15 ***      30.26     4.30 ***
     1          33.45     5.54 ***      32.32     6.46 ***
     2          37.12     3.92 ***      34.37     7.04 ***
     3          40.79     2.50 **       36.42     3.93 ***

All are measured in terms of the values of the Protestants/Catholics
minus the values of the Buddhists.

* Significant at 10% level.

** Significant at 5% level.

*** Significant at 1% level.

Table 5. Regression Results of the Tobit Models with
Age x Religion Dummy

                                             Contribution

                                       With Volunteer Frequency
                                               Model II A

Variable                                Coefficient     t-Ratio

[D.sub.F]                                 -25.25      -6.21 ***
[D.sub.B]                                 -29.44      -6.11 ***
[D.sub.C]                                 -44.73      -5.75 ***
DEV (devotionalism)                         4.02       3.77 ***
[B.sub.B] (belief in reincarnation)         1.13       1.36
[B.sub.C] (belief in heaven)                4.50       2.93 ***
[D.sub.HB] (help in business)               2.53       2.04 **
[D.sub.HN] (help in social network)         1.76       1.31
[D.sub.HF] (help in family relations)       1.87       1.35
SEX                                        -3.2       -2.73 ***
Age x [D.sub.C]                             0.89       4.92 ***
Age x [D.sub.B]                             0.39       4.32 ***
Age x [D.sub.F]                             0.23       3.19 ***
MAR                                         1.99       1.22
EDU                                         0.27       1.55
WAGE                                        0.11       4.53 ***
FINC (family income)                        0.05       3.13 ***
[[??].sub.f]                               -0.36      -1.95 **
[[??].sub.f]
Sigma                                      16.68      41.91 ***
Observations                             1278
Log-likelihood                          -4051.48

                                         Volunteer Frequency

                                      Without Income Model IV A

Variable                                Coefficient     t-Ratio

[D.sub.F]
[D.sub.B]                                 -92.86      -5.76 ***
[D.sub.C]                                 -65.79      -3.95 ***
DEV (devotionalism)                        12.91       2.99 ***
[B.sub.B] (belief in reincarnation)         1.30       0.67
[B.sub.C] (belief in heaven)                3.37       1.14
[D.sub.HB] (help in business)               6.02       1.38
[D.sub.HN] (help in social network)         8.90       1.69 *
[D.sub.HF] (help in family relations)      16.08       2.53 **
SEX                                        -9.98      -2.92 ***
Age x [D.sub.C]                             0.59       2.02 **
Age x [D.sub.B]                             0.51       2.28 **
Age x [D.sub.F]
MAR                                         9.39       1.46
EDU                                         1.67       2.34 **
WAGE
FINC (family income)                        0.04       0.89
[[??].sub.f]
[[??].sub.f]                               -1.23      -1.85 *
Sigma                                      26.11      14.73 ***
Observations                              515
Log-likelihood                           -739.2

* Significant at 10% level.

** Significant at 5% level.

*** Significant at 1% level.