Church Contributions and Church Attendance

Transcription

1 1 1

2 2 Church Contributions and Church Attendance A thesis submitted to the Miami University Honors Program in partial fulfillment of the requirements for University Honors By Christina M. Altrudo May 2003 Oxford, Ohio 2

3 3 ABSTRACT Church Contributions and Church Attendance by Christina Altrudo In this paper, economics and religion encounter each other in the application of microeconomic models to religious non-market behavior in an attempt to explain patterns of church contributions and church attendance. Dennis Sullivan (1985) used a 1963 dataset of over 2,000 Protestants to look at the structural relationship between religious contributions of time and money using the household allocation model, a framework established by Azzi and Ehrenberg (1975). This paper takes a similar approach by modeling contributions of time and money in a simultaneous equations framework using a more recent and comprehensive dataset. The use of dummy variables allows us to look at the selective incentives behind church contributions and church attendance and to consider each separately. Overall, this paper validates most of Sullivan s (1985) findings. 3

4 4 4

5 5 Church Contributions and Church Attendance by Christina Altrudo Approved by:, Advisor Dennis Sullivan, Reader Michael Curme, Reader Harold Forshey Accepted by:, Director Carolyn Haines University Honors Program 5

6 6 6

7 7 List of Tables Table 1 Table 2 Two Stage Least Square Regression Coefficients, Weekly Contributions to Church (in $) Two Stage Least Square Regression Coefficients, Annual Attendance at Church Services Appendix Tables Table A Table B Table C Table D Ordinary Least Square Regression Coefficients, Weekly Contributions to Church (in $) Ordinary Least Square Regression Coefficients, Annual Attendance at Church Services Ordinary Least Square Regression Coefficients, Weekly Contributions to Church (in $), Annual Attendance at Church Services Means and Standard Deviations of Coefficients 7

8 8 Introduction The study of religion as a topic for economic research was begun 200 years ago by Adam Smith (Smith V.1.230) and was left almost untouched until the mid-70s when Corry Azzi and Ronald Ehrenberg (1975) picked it up again. In the past 25 years, economics and religion have encountered each other in three research programs: the application of microeconomic models to religious "non-market" behavior, the economic consequences of religion, and the use of sacred texts to evaluate economic policies (Iannaccone 1998). The first of these lines of research, influenced by Gary Becker, includes multiple studies on household and individual decisions to give time and money to churches. The research questions in this area have considered independent variables such as denomination, church size, church attendance and the demographics of parishioners (Hoge 1994). This paper takes a similar approach to the study of religion by modeling contributions of time and money in a simultaneous equations framework. The goal of the paper is to update previous research relating the two forms of charitable contributions to churches using a more recent and comprehensive dataset. It will proceed with a literature review, a description of the data and empirical strategy, and a discussion of the results. Literature Review Studies on the decision to give time and money to churches borrow from and add to another line of research, the economics of charitable giving. Economists encounter an interesting problem when modeling charity since it is a public good that is financially 8

9 9 supported by both public and private funds. Public funding decisions are made by the government, whose policies are affected by voters that may also contribute private funds. There are a few hypotheses that attempt to explain what motivates individuals to donate private funds to support public goods. James Andreoni (1990) models the decision of a private donor using a simplified economy that only produces one private good and one public good. Individual agents have two consumption decisions: purchasing the private good, x, or donating to the public good, g. The public good receives all of its funding from individuals; there is no government in this simplified economy. The total amount of the public good, G, is the sum of all donations, g, from all agents in the economy. When the donation of individual, i, to the public good is unaffected by outside influences, such as the donations or opinions of other agents, guilt or sympathy, the donor utility function is: U i = U i (x i, G); (1) which represents the purely altruistic model. This utility function is for agents that are only concerned with the overall provision of the public good. Conversely, if an individual makes consumption decisions solely based on the "warm-glow" received from a charitable donation, the donor utility function is: U i = U i (x i, g i ); (2) which is known as the purely egoistic model. Utility is derived solely from the consumption of private and public goods; the total amount of the public good does not increase the agent's utility. The following utility function represents a combination of the purely altruistic and the purely egoistic models: 9

10 10 U i = U i (x i, G, g i ) (3) It is called the impurely altruistic model, and it represents the individual whose utility is enhanced by the consumption of private and public goods as well as the overall level of charity. Donations to the public good, g, are directly and indirectly present in the equation since the total level of the charity, G, is the sum of the individual donations from all members of the economy (Andreoni 1990). When the model includes the government as a contributor to charity, there is an additional option for funding the charitable endeavor and the potential for crowd-out. Crowd-out occurs when an increase in government spending causes a reduction in private spending, which can partially or completely offset the government increase. In our simplified model, only one public good is produced, so we can illustrate a simple crowdout theory. Suppose that government funds, F, and the contributions of others, R, are considered. G is now a function of g, F and R, and that initially, F and R are perfect substitutes in the production of G. In the purely altruistic model of public goods, there will be a dollar-for-dollar crowd-out of g from an increase in F or R, since the individual derives utility solely from the total level of charity. In the purely egoistic model, the crowd-out is zero; agents are not influenced by the government's contributions or the donations of others. In the impurely altruistic model, we expect that the crowd-out is less than dollar-for-dollar. A modification of the impurely altruistic model, the source-ofcontributions model, takes into consideration an agent's preference for F or R, making them imperfect substitutes. For example, an individual may prefer contributions by others to contributions by the government because government contributions may be temporary, 10

11 11 or they may invite more government control (Kingma 1989). Kingma s model also has less than dollar-for-dollar crowd-out, but it differs based on the source of the contribution. Empirical analyses of crowd-out seem most consistent with the impurely altruistic model (Abrams and Schmitz 1984; Kingma 1989; Payne 1998). Burton A. Abrams and Mark D. Schmitz (1984) use data from 1979 itemized tax returns to test whether government aid to non-profits reduces the need of social welfare recipients sufficiently enough to reduce the donations from private agents. They begin with the following hypotheses: contributions to charity or "the purchase of charitable services" act like normal goods, increases in the price of a donation (1 minus the marginal tax rate) decrease the demand, and donors contribute to charities until the marginal benefit and the marginal cost of their contributions are equal. Abrams and Schmitz limit their study by excluding federal funding and only use state and local welfare payments to proxy government aid to non-profits. They find that private donors pick up about thirty percent of any decline in public funding to non-profits, a less than dollar-for-dollar crowd-out effect. In his 1989 analysis of the methodology of the empirical research on charitable contributions to date, Bruce Robert Kingma introduces the distinction between the "substitution effect" and the "crowd-out effect." According to Kingma, the crowd-out effect should only be used in reference to examples in which the government and private investors contribute to the same public good. If public and private funding support different public goods that address the same social concern, a decrease in private funds as 11

12 12 a response to an increase in public funds for a similar public good is a substitution. For example, individual contributors may give money to a non-profit food pantry to supplement government programs that also aim to feed the hungry. If an increase in Food Stamps results in a decrease in individuals' contributions to the food pantry, then government funds have substituted for private contributions. Unless individuals privately contribute to the Food Stamp program, the crowd-out effect is not being analyzed directly. Empirical results from pre-1989 research vary between illustrating complete dollar-for-dollar crowd-out and finding no evidence that government funds crowd-out charitable donations. Kingma s distinction between crowd-out and substitution may explain the discrepancy. In his own empirical study, Kingma uses the example of public television and public radio, both of which are funded jointly by the government and private contributors. He finds that a $10,000 increase in government funding for a public radio station results in a $.15 decrease in contribution per member. Considering that average public radio station is supported by 9,000 members, $1,350 of private investment is crowded-out, and the overall increase realized by the station is only $8,650. This result is less than dollar-for-dollar, and it supports the impurely altruistic model. The substitution effect seems to be the most appropriate way to measure the effect of government spending on the provision of charitable services by churches, since the government rarely supports church efforts directly. There are many examples of similar services provided by the church and the state, such as food pantries and Food Stamps. J. Lipford, et al. (1993) make the case that the church and state produce similar public 12

13 13 goods, such as law and order, and thus analyze the substitution effect for those goods, rather than the organizations that provide them. An increase in the number of churches in any given area may reduce the necessity for state provision of the public good. For example, the church and the state both produce education, such that public and private education can treated as perfect substitutes. It is possible that the state does not have to spend as much money on education in areas with strong parochial schools. Also, states that have a lot of churches with a solid presence have a stronger "moral fiber" and lower rates of anti-social behavior, requiring less government intervention (assuming that the churches preach moral behavior) than states with fewer churches (Lipford, McCormick et al. 1993). The authors measure the moral fiber of a state by the per capita number of abortions, divorces and illegitimate births. Anti-social behavior is measured by the number of murders, rapes and crimes, both violent and non-violent. These are regressed on church membership by state with other controls, such as education, income and race included. The hypotheses are confirmed; ceteris paribus, illegitimate births and crime rates are inversely related to church membership by state. The results confirm that churches produce public goods with positive externalities; thus, there is an opportunity for the government to become a free-rider. Areas with high numbers of churches have lower expenditures on police enforcement, suggesting that there is a lower demand for the state to provide some public goods when churches have a strong bearing in an area (Lipford, McCormick et al. 1993). Dan Hungerman, an economics doctoral student at Duke University, proposes another type of substitution analysis in his working paper, "Are Church and State 13

14 14 Substitutes? Evidence from the 1996 Welfare Reform." The paper integrates the research on religious giving to churches with the research on the substitution effect by looking at the changes in individual contributions to churches and the spending decisions that the churches make with respect to changes in government funding of similar services. Using welfare as the public good, Hungerman takes advantage of a 1996 change in the welfare laws relating to immigrants to come up with an instrument to control for government spending. It allows him to look at donations to churches and church spending pre- and post-1996 to assess the level of substitution. He finds that churches increase their spending by $.27 per member for each one-dollar decrease in per capita spending on Food Stamps. The same is true for spending on Medicaid, although the results are smaller and not consistently significant. His results are consistent with the literature that suggests that the impure altruistic model accurately represents charitable giving with a less than dollar-for-dollar crowd-out effect. Abigail Payne (1998) approaches the charitable contribution crowd-out issue by modeling the ability of individuals to affect the government's funding decisions with their votes. Payne looks at the period from 1982 to 1992 to evaluate the effects of government crowd-out since the value of government grants to non-profits decreased and private contributions increased during that time. She avoids producing a substitution paper by using data from specific non-profits, which allows her to match the amount of private donations with the amount of government aid to each organization. Payne recognizes the importance of using a model that allows private agents to affect charitable organizations on two levels: by donating money and by voting for government policies. She proxies 14

15 15 the voting privilege with two measures: "the representation of the political parties in legislative positions at the state and federal level and the political party affiliation of the state's governor" (Payne 1998). Payne begins with the Andreoni impure altruist model and extends it to encompass three agents that behave in the following manner. The first agent, the government, bases its aid to non-profits on the result of a vote. The second agent, individuals (which includes any private donor, including corporations, estates, etc.), make their contribution decisions with the knowledge of the government contribution and the contributions of all other individuals. Finally, the third agent is nonprofit firms that supply the desired good based on the amount of contributions they receive (Payne 1998). The panel data set for the individual non-profits comes from IRS federal tax returns for a 10-year period. Four social concerns are included in the classification for non-profits: crime or disaster-related, employment or youth, food or shelter, human services. Payne's study reports a one-cent-per-dollar crowd-out for the above-listed non-profits under the OLS specification, and a 50-cent-per-dollar crowd-out under the two-stage least squares estimation. Many empirical studies focus solely on the relationship of government crowd-out on private monetary contributions to a public good. Brian Duncan (1999) predicts that any such study underestimates the effect of crowd-out by up to 27% by neglecting to consider the importance of contributions of time. According to Duncan, all previous research on the effects of charitable contributions of time and money use the private consumption, or purely egoistic model, in which time and money are gross complements. However, Duncan predicts that time and money are best modeled in the public goods, or 15

16 16 purely altruistic model, in which they are substitutes. Since individuals operating in the purely altruistic model value the overall provision of the public good, there is a unique point at which the supply of the charitable good is maximized. In equilibrium, contributors will substitute between contributions of time and contributions of volunteer labor hours to maintain utility maximization. Contributors will not leave the charity constrained by a lack of capital. Since time and money are perfect substitutes, if a charitable firm s capital costs more than the total amount of its monetary donations, contributors will be willing to decrease work hours and increase money. Duncan s predictions also hold in the impurely altruistic model as long as time and money are perfect substitutes. If contributors begin to receive warm-glow utility from the gift itself, time and money cease to be perfect substitutes, and contributors may leave the firm constrained by capital. Dollar-for-dollar crowd-out is consistent with the purely altruistic model if money is the only way to contribute. However, individuals also contribute time to charities, and Duncan (1999) finds that volunteer labor hours actually react more strongly to changes in government policy than monetary donations. Duncan s empirical data comes from the National Study of Philanthropy (1974), one of the few datasets to report on both time and money. He uses the total value of a household s charitable contribution as the dependent variable in his regression. Total value is the sum of the household s wage multiplied by the number of hours volunteered plus the monetary donation. Duncan s results do not find evidence to support dollar-for-dollar crowd-out. Furthermore, by rejecting the crowd-out hypothesis, he concludes that households must derive utility from the 16

17 17 charitable gift itself. Since he also concludes that time and money are perfect substitutes (in support of the altruistic models), it must be true that households derive utility from the total value of their charitable contribution, rather than from time or money separately. Duncan s result is consistent with the standard model of household allocation of time and income. Since households value the total amount of their contribution, we know that time and money contributions are determined simultaneously. This approach has been applied to the study of church contributions and church attendance. Dennis Sullivan (1985) looked at the structural relationship between religious contributions of time and money using the framework established by Azzi and Ehrenberg (1975). In a household allocation model that considers religious activities, time is allocated to leisure, work and religion. Income is allocated to consumption bundles and religious service bundles. From the full income constraint, Ehrenberg develops predictions about the life-cycle effects on attendance and contributions. His model predicts that the age profile of church attendance will be U-shaped, and that the age profile of contributions rises until retirement. The model also predicts that the ratio of attendance to contributions will fall as individuals get older. In his critique of Ehrenberg, Sullivan (1985) points out that it is both useful and possible to estimate the structural relationship between attendance and contributions by making some assumptions about incentives. In other words, it is necessary to pose the question in terms of a specific model, such as the egoistic model in which individuals make contributions because they receive utility from doing so, not because their contribution increases the social good. Sullivan s empirical study considers these incentives, as well as differences in gender using a 1963 dataset of over 2,000 Protestants 17

18 18 in northern California. He uses a two-stage least squares estimating strategy to regress contributions and attendance on demographic variables, incentive variables and on each other. This strategy allows contributions and attendance to be simultaneously determined, which is consistent with Duncan s result that individuals value total amount of their contribution in terms of both time and money. It is the goal of this paper to use the framework set up by Sullivan (1985) to explore the structural relationship between church attendance and church contributions using a more recent and larger dataset. Data and Estimating Strategy In 1992, Daniel Hoge, Charles Zech, Patrick McNamara and Michael J. Donahue, recognizing the need for reliable and current data in order to study religious giving in the United States, received a grant from the Lilly Endowment to conduct an original analysis of five Christian denominations. They chose a random sample of 125 congregations from each denomination representing all nine U.S. census regions. Clergy members supplied data on the income, spending habits, demographic and theological beliefs of each church. Individual giving and attendance habits were tracked for thirty randomly-selected lay members from each parish, who also evaluated the church s teachings and leadership. Hoge et al. use their dataset to derive some empirical results. They find five common factors that influence individual giving in all of the denominations. Individual giving to churches is positively correlated with high family income and high levels of involvement in the church. If the church encourages pledging or tithing or if it is particularly conservative, it is more likely to receive higher gifts from its parishioners. 18

19 19 They conclude their study with some recommendations to churches that want to increase the level of charitable contributions of their congregations, and these recommendations reveal more of Hoge et al. s findings. Churches should allow and encourage the role of lay members in the leadership and financial decisions of the church. They should also make their financial information public and accessible to their congregation. Both of these recommendations increase the level of trust that a parishioner places in the leadership of the church, which will increase his desire to support the church financially. Hoge et al. (1996) note that parishioners are much more sympathetic to missions that affect local causes, rather than national or denominational causes. Finally, they find that lessons in the importance of stewardship and the joy of giving will increase individuals desire to give (Hoge et al. 172). Hoge et al. (1996) find a weak negative correlation between the size of a church and the giving of its members; the larger the church, the smaller the contribution. The size correlation applies only to Catholic Churches, and it is a weak factor. However, freerider theory predicts a positive relationship between free-riding, or taking advantage of a public good without contributing to it financially, and the size of the organization that produces the public good. Jody W. Lipford (1995) notes the relevance that the freeriding hypothesis may have if individuals are operating in an egoistic model: if contributions are directly linked to private benefits, such as afterlife consumption, tax deductions, or desired positions of authority in the church, the perceived and actual costs of free riding may be great (Lipford 299). Lipford (1995) uses 1986 data on three Christian denominations in South Carolina (Baptists, Episcopalians and Presbyterians) to 19

20 20 test the relationship between free-riding and church size. Assuming that churches produce public goods that are not priced, such as sermons or sacraments, Lipford rejects the hypothesis that parishioners are more likely to donate money to their church if it is small. He repeatedly rejects the hypothesis that free riding is more common in large churches than it is in small churches. Lipford finds a positive relationship between church goods and income, and he finds an increased tendency to free-ride in churches where tithing is expected. He finds different rates of giving across the three Christian denominations that he studies. Presbyterians give more than Episcopalians, who give more than Baptists, but interestingly, the rate at which giving increases with income is the same across denominations. Lipford offers an explanation for the surprising result that free-riding does not increase as church size increases. It is possible that churches lessen the effect of free-riding by maintaining a sense of community as the size of the church increases. Sunday school and fellowship groups are examples of the many activities that churches offer in addition to services that add to a sense of belonging. Other studies have shown that people that participate in other activities in addition to regular services are more likely to contribute money to the church than those that do not (Wuthnow 1994). Sullivan (1985) finds a result opposite to Lipford. First, he finds evidence for the free-rider effect in a negative relationship between church size and the size of contributions, similar to Hoge et al. Sullivan also finds a falling income profile of church attendance, indicating that churches do not produce a normal good, opposite from Lipford s finding of a positive relationship between income and attendance. The difference in the results may be due to the difference in the estimating strategies of the studies. 20

21 21 This study uses the dataset compiled by Hoge et al. in 1993, which was made available by the American Religion Data Archive. It reports data on the patterns and motivations of individual giving for denominations representing various polities and tendencies of charitable giving. The denominations include the Assemblies of God, the Southern Baptist Convention, the Roman Catholic Church, the Evangelical Lutheran Church in America (ELCA), and the Presbyterian Church (U.S.A.). Before-tax income is reported by family or household. There are twelve categories, the last of which is openended; a Pareto distribution was fitted in order to model income. Two dummy variables are incorporated to proxy different incentives for religious giving and attendance. Individuals that participate in religious communities that preach the importance of tithing and regularly give 10% of their income to the church have an incentive to donate, in addition to social pressure to donate. Churches that preach tithing do not allow their parishioners to substitute attendance for contributions, so tithing is a selective incentive. Similarly, individuals that participate in religious communities that preach the necessity for faithful Christians to participate in the traditions and the sacraments of the church have an incentive to attend religious services, and do not have the opportunity to substitute contributions for attendance. Thus, the two dummy variables, tithe and duty, represent selective incentive variables that allow us to look at contributions and attendance separately. The estimating strategy of this paper is to allow both contributions and attendance to be determined endogenously, and to incorporate proxies for incentives as instrumental variables, similar to the framework of Sullivan (1985). In addition, one of the benefits of 21

22 22 the 1993 Hoge et al. dataset is the ability to distinguish between denominations, so the denominations can also be included as regressors. The equations for church contributions and attendance were first estimated without the incentive variables using an ordinary least squares regression method (see Appendix A). Most of the coefficients are significant in both cases, except for the marital status dummy variable and the church size variable. The construction of these equations force men and women to have the same coefficients on the other variables, such as age and income. However, there are theoretical reasons to expect men and women to have different responses. For example, given the lower level of labor force participation among women, we might expect less of a response to a change in income than we would expect for men, since men are more likely to be the prime earners of their households. In order to determine if men and women differ significantly in their allocation of time and money to churches, we estimate an OLS regression with variables that interact gender with the life-cycle and incentive regressors. The results in Appendix C reveal a statistically significant interaction between gender and income and between gender and the incentive variables, duty and tithe. This convinced us that it is necessary to consider the gender differences for income and attendance. Results Table 1 reports the two-stage least square regression coefficients for the equation modeling weekly monetary contributions to churches for men and women. The age 22

23 23 profile of contributions rises for both men and women. For men, it rises until around age 52, and for women, it rises until age 81. This is consistent with Sullivan, who predicts that contributions should rise across all working ages, and possibly even after retirement. For both men and women, we find that contributions rise with income. We predict that parents of children under 18 will contribute less than non-parents, based on Sullivan s (1985) study. Since children reduce the discretionary income of their household, we find that the presence of a child in the household has a strong negative effect on the weekly contributions of men, and a statistically insignificant, although negative effect on the contributions of women. In this study, as opposed to Sullivan s 1985 study, evidence suggests that there is a positive relationship between church size and contributions, although it is not a statistically significant relationship. The incentive variable reveals a very strong positive relationship between the belief in tithing and contributions of women, and a positive but weak relationship for men. In this estimation, it is not apparent that attendance has a strong influence on the contribution decision. All five denominations are included in the model with Roman Catholics as the control group. Historically, the Catholic Church receives a low level of weekly contributions (Hoge et al. 2). We find that Lutheran men give significantly more than Catholic men, and Southern Baptist women give significantly less than Catholic women. Overall, the F-statistic reveals that the model is statistically significant for both men and women. Table 2 reports the two stage least square regression coefficients on annual church service attendance. Azzi and Ehrenberg and Sullivan both predict U-shaped age profiles 23

24 24 based on the opportunity cost of time; the opportunity cost of religion falls at first, and then rises after the age at which wages begin to flatten. Sullivan finds this result to be true for men, but he finds a rising age profile for women across all ages due to a lower labor force participation rate and flatter earnings profiles. The empirical results from the Hoge et al. dataset do not correspond to either of these hypotheses. We find an almost flat age profile of attendance that rises across the observed range of ages for men and women. The age coefficients for both men and women are jointly significant, but not individually significant. Azzi and Ehrenberg (1975) and Lipford (1995) predict that there will be a rising income profile of attendance. Like Sullivan, this study finds the opposite to be true. As income rises, individuals chose to attend church less often. We find that married women attend church slightly less than unmarried women, while married men attend more than unmarried men. We also find that the presence of a child in the household encourages men and women to attend church more often. There is a positive relationship between contributions and annual attendance for men and women. Finally, with Roman Catholics still the control group, we find that Lutherans and Presbyterians attend less than Catholics. Conclusion Sullivan (1985) used data from Thirty years later, almost all of his results still hold among Christian denominations in the United States. There are only two areas in which this study found results that differ from Sullivan. First, the age profile of 24

25 25 attendance is not U-shaped; it is slightly increasing for both men and women throughout their lifetimes. Second, this study finds a positive relationship between church size and contributions, but a negative relationship between church size and attendance. However, these results fail to find a statistically significant effect of church size on contributions. Overall, the variables that measure incentives to contribute, tithe and duty, are only significant for women, and they have a positive relationship on contributions and attendance respectively. 25

26 26 Table 1: Two Stage Least Square Regression Coefficients Weekly Contributions to Church (in $) d Men Women Regressor a Coefficient (Standard Error) b c Coefficient (Standard Error) b c Intercept ( ) * ( ) Married ( ) ( ) Child ( ) * ( ) Age ( ) ( ) * Age Squared ( ) ( ) * Income ( ) * ( ) * Income Squared (x 10 6 ) ( ) ( ) Church Size ( ) ( ) Tithe ( ) ( ) * Attendance ( ) ( ) Assemblies of God ( ) ( ) Southern Baptist ( ) ( ) * ELCA ( ) * ( ) Presbyterian ( ) ( ) Adjusted R (F = ) (F = ) Turning Age a. Variable List: MALE = 1 if male, 0 otherwise; MARRIED = 1 if married, 0 otherwise; CHILD = # of dependent children under 18 in the household, AGE = age in years, CHURCH SIZE = # of members, TITHE = 1 if contribution is 10% of income, 0 otherwise; ATTENDANCE = attendance in days. b. Asterisks in this column indicate that coefficients are significantly different from zero at the α=.05 level. c. Daggers in this column indicate that the variable and its square are jointly significant. d. The denominations are jointly significant for men and women. 26

27 27 Table 2: Two Stage Least Square Regression Coefficients Annual Attendance at Church Services d Men Women Regressor a Coefficient (Standard Error) b c Coefficient (Standard Error) b c Intercept ( ) * ( ) * Married ( ) * ( ) * Child ( ) * ( ) * Age ( ) ( ) Age Squared ( ) ( ) Income ( ) * ( ) * Income Squared (x 10 6 ) * Duty ( ) ( ) * Church Size ( ) ( ) Contributions ( ) * ( ) * Assemblies of God ( ) ( ) * Southern Baptist ( ) ( ) ELCA ( ) * ( ) * Presbyterian ( ) * ( ) * Adjusted R (F = ) (F = ) Turning Age n/a a. Variable List: MALE = 1 if male, 0 otherwise; MARRIED = 1 if married, 0 otherwise; CHILD = # of dependent children under 18 in the household, AGE = age in years, CHURCH SIZE = # of members, TITHE = 1 if contribution is 10% of income, DUTY = 1 if primary duty of Christians is sacramental traditions, 0 otherwise. b. Asterisks in this column indicate that coefficients are significantly different from zero at the α=.05 level. c. Daggers in this column indicate that the variable and its square are jointly significant. d. The denominations are jointly significant for men and women. 27

28 28 Appendix Table A: Ordinary Least Square Regression Coefficients Weekly Contributions to Church (in $) Annual Attendance at Church Services Regressor a Coefficient (Standard Error) b c Regressor a Coefficient (Standard Error) b c Intercept ( ) * Intercept ( ) * Male ( ) * Male ( ) Married ( ) Married ( ) Child ( ) * Child ( ) * Age ( ) * Age ( ) * Age Squared ( ) * Age Squared ( ) * Income ( ) * Income ( ) Income Squared 2.29E-08 (1.121E-08) * Income Squared -2.40E-10 (9.07E-11) * Church Size ( ) Church Size ( ) Assemblies of God ( ) * Assemblies of God ( ) * Southern Baptist ( ) * Southern Baptist ( ) * ELCA ( ) * ELCA ( ) * Presbyterian ( ) * Presbyterian ( ) * Adjusted R (F = ) Adjusted R (F = ) a. Variable List: MALE = 1 if male, 0 otherwise; MARRIED = 1 if married, 0 otherwise; CHILD = # of dependent children under 18 in the household, AGE = age in years, CHURCH SIZE = # of members, TITHE = 1 if contribution is 10% of income, 0 otherwise; ATTENDANCE = attendance in days. b. Asterisks in this column indicate that coefficients are significantly different from zero at the α=.05 level. c. Daggers in this column indicate that the variable and its square are jointly significant. a. Variable List: MALE = 1 if male, 0 otherwise; MARRIED = 1 if married, 0 otherwise; CHILD = # of dependent children under 18 in the household, AGE = age in years, CHURCH SIZE = # of members, DUTY = 1 if primary duty of Christians is sacramental traditions, 0 otherwise. b. Asterisks in this column indicate that coefficients are significantly different from zero at the α=.05 level. c. Daggers in this column indicate that the variable and its square are jointly significant. 28

29 29 Appendix Table B: Two Stage Least Square Regression Coefficients Weekly Contributions to Church (in $) Annual Attendance at Church Services Regressor a Coefficient (Standard Error) b c Regressor a Coefficient (Standard Error) b c Intercept ( ) * Intercept ( ) * Male ( ) * Male ( ) * Married ( ) Married ( ) Child ( ) Child ( ) * Age ( ) Age ( ) * Age Squared ( ) * Age Squared ( ) Income ( ) * Income ( ) * Income Squared 2.84E-08 (1.305E-08) * Income Squared -4.07E-10 (1.07E-10) * Church Size ( ) Church Size ( ) Tithe ( ) * Duty ( ) * Attendance ( ) Contributions ( ) * Assemblies of God ( ) * Assemblies of God ( ) * Southern Baptist ( ) * Southern Baptist ( ) ELCA ( ) ELCA ( ) * Presbyterian ( ) Presbyterian ( ) * Adjusted R (F = ) Adjusted R (F = ) a. Variable List: MALE = 1 if male, 0 otherwise; MARRIED = 1 if married, 0 otherwise; CHILD = # of dependent children under 18 in the household, AGE = age in years, CHURCH SIZE = # of members, TITHE = 1 if contribution is 10% of income, 0 otherwise; ATTENDANCE = attendance in days. b. Asterisks in this column indicate that coefficients are significantly different from zero at the α=.05 level. c. Daggers in this column indicate that the variable and its square are jointly significant. a. Variable List: MALE = 1 if male, 0 otherwise; MARRIED = 1 if married, 0 otherwise; CHILD = # of dependent children under 18 in the household, AGE = age in years, CHURCH SIZE = # of members, DUTY = 1 if primary duty of Christians is sacramental traditions, 0 otherwise. b. Asterisks in this column indicate that coefficients are significantly different from zero at the α=.05 level. c. Daggers in this column indicate that the variable and its square are jointly significant. 29

30 30 Appendix Table C: Ordinary Least Square Regression Coefficients Weekly Contributions to Church (in $) Annual Attendance at Church Services Regressor a Coefficient (Standard Error) b Regressor a Coefficient (Standard Error) b Intercept ( ) * Intercept ( ) * Male ( ) Male ( ) Married ( ) Married ( ) Child ( ) Child ( ) * Age ( ) * Age ( ) * Age Squared ( ) * Age Squared ( ) * Income ( ) * Income ( ) * Income Squared 5.45E-08 (1.536E-08) * Income Squared 1.64E-10 (1.43E-10) Church Size ( ) * Church Size ( ) * Male*Married ( ) Male*Married ( ) * Male*Child ( ) Male*Child ( ) Male*Age ( ) Male*Age ( ) Male*Age Squared ( ) Male*Age Squared ( ) Male*Income ( ) * Male*Income ( ) * Male*Income Squared -4.76E-08 (2.273E-08) * Male*Income Squared 1.31E-10 (2.11E-10) Male*Church Size ( ) Male*Church Size ( ) Male*Tithe ( ) * Male*Duty -( ) ( ) * Adjusted R (F = ) Adjusted R (F = 14.91) a. Variable List: MALE = 1 if male, 0 otherwise; MARRIED = 1 if married, 0 otherwise; CHILD = # of dependent children under 18 in the household, AGE = age in years, CHURCH SIZE = # of members, TITHE = 1 if contribution is 10% of income, 0 otherwise; ATTENDANCE = attendance in days. b. Asterisks in this column indicate that coefficients are significantly different from zero at the α=.05 level. a. Variable List: MALE = 1 if male, 0 otherwise; MARRIED = 1 if married, 0 otherwise; CHILD = # of dependent children under 18 in the household, AGE = age in years, CHURCH SIZE = # of members, DUTY = 1 if primary duty of Christians is sacramental traditions, 0 otherwise; ATTENDANCE = attendance in days. b. Asterisks in this column indicate that coefficients are significantly different from zero at the α=.05 level. 30

31 31 Appendix Table D: Means and Standard Deviations of Coefficients Variable Mean Standard Deviation Minimum Maximum Male Married Child Age Income 42, , , ,000 Contributions 1, , ,100 Church Size , ,999 Attendance Duty Tithe

32 32 Bibliography Abrams, B. A. and M. D. Schmitz. "The Crowding-out Effect of Governmental Transfers on Private Charitable Contributions: Cross-Section Evidence." National Tax Journal XXXVII (1984): Andreoni, James. "Impure altruism and donations to public goods: A theory of warm glow giving." Economic Journal 100 (1990): Andreoni, James and John Karl Scholz. An Econometric Analysis of Charitable Giving with Interdependent Preferences. Economic Inquiry, XXXVI (1998): Azzi, Corry and Ronald Ehrenberg. Household Allocation of Time and Church Attendance. The Journal of Political Economy, 83 (Feb. 1975): Brooks, Arthur. "Welfare Receipt and Private Charity." Public Budgeting & Finance (Fall 2002): ). Clain, S. H. and C. E. Zech. "A Household Production Analysis of Religious and Charitable Activity." American Journal of Economics and Sociology 58(4) (1999): Duncan, Brian. Modeling Charitable Contributions of Time and Money. Journal of Public Economics, 72 (1999): Hoge, D. R. "Introduction: The Problem of Understanding Church Giving." Review of Religious Research 36(2) (1994): Hoge, D, C. Zech, P. McNamara, and M. Donahue. Money Matters: Personal Giving in American Churches. Louisville: Westminster John Knox Press, Hungerman, Dan. Are Church and State Substitutes? Evidence from the 1996 Welfare Reform, working paper. Iannaccone, Lawrence. Introduction to the Economics of Religion. Journal of Economic Literature, XXXVI (1998). Kingma, Bruce. R. "An Accurate Measure of the Crowd-out Effect, Income Effect, and Price Effect for Charitable Contributions." Journal of Political Economy 97(5) (1989):

33 33 Lipford, Jody W. Group size and the free-rider hypothesis: An examination of new evidence from churches. Public Choice, 83 (June 1995): Lipford, Jody., R. E. McCormick, et al. "Preaching Matters." Journal of Economic Behavior and Organization 21 (1993): Payne, Abigail A. "Does the government crowd-out private donations? New evidence from a sample of non-profit firms." Journal of Public Economics 69 (1998): Smith, Adam, An Inquiry into the Nature and Causes of the Wealth of Nations. Methuen and Co., Ltd Ed. Edwin Cannan. Library of Economics and Liberty. 3 April < Sullivan, Dennis. Simultaneous Determination of Church Contributions and Church Attendance. Economic Inquiry, 23 (April 1985): Wuthnow, Robert. God and Mammon in America. New York: The Free Press, Zaleski, Peter A. and Charles E. Zech. The Effects of Religious Market Competition on Religious Giving, Review of Social Economy, LIII (1995):