An Overlapping Generations Economy with Religiosity Laurent Cellariery
Michael Batu
May 10, 2013
Abstract This paper provides strong empirical evidence that religious individuals save more and allocate a larger fraction of their income to tangible religious donations over their life cycle. We also show using macroeconomic data that countries with a large religious adherence population have a higher saving rate than the ones with fewer religious adherents. To analyze both the microeconomic and macroeconomic e¤ects of religiosity, we consider a production overlapping generations (OLG) economy in which a fraction of the household population is endowed with various religious beliefs. In particular, we augment the benchmark OLG framework with both a perceived after-life bene…t function and a religious faith production technology that relies on tangible lifetime donations. In the resulting model, we analyze and compare how beliefs about the afterlife bene…ts and the input-output relationship in the religious faith production function a¤ect lifetime decisions, the competitive equilibrium dynamics and the long-run standards of living. We also identify the characteristics of the religious population that would make our arti…cial economy consistent with the data and/or reach the golden rule steady-state level of physical capital accumulation. JEL Classi…cation: E13, E21, O57, Z12 Keywords: Religiosity, overlapping generations, savings, faith PhD Candidate, Department of Economics and Finance, University of Guelph, Guelph, Ontario, N1G 2W1, Canada. E-mail:
[email protected]. y Associate Professor, Department of Economics and Finance, University of Guelph, Guelph, Ontario, N1G 2W1, Canada. E-mail:
[email protected].
1
1
Introduction
It is well understood that humanity’s existence has been signi…cantly in‡uenced by a system of beliefs centered in the existence of a God(s) or a universal spirit. This system also includes beliefs in the existence of a place where this God resides, “heaven” for instance, and a promise of life after death. These beliefs have been observed amongst human beings since recorded history with the earliest evidence of religious thought based on the ritual treatment of the dead1 . The emergence of complex and organized religions can be traced to the period when humans abandoned their nomadic hunter gatherer lifestyles in order to begin farming during the Neolithic period. The transition from foraging bands to states and empires resulted in more specialized and developed forms of religion that were re‡ections of the new social and political environments. Newer present-day world religions established themselves throughout Asia and Europe during the middle ages2 . Organized religion has become the foundation of everyday life in the middle ages and up to the modern period, setting the standards and rules of society for its members. The in‡uence of religiosity to its members is quite evident in modern times. In a recent Gallup poll (May 2011), more than 9 out of 10 Americans believed in the existence of God. In the same poll, 85% and 75% believe in the existence of heaven and hell, respectively. More than half of Americans surveyed believe that religion remains an important part of their life. Church attendance has been consistently high with 43% of Americans reporting weekly or almost weekly attendance. Religious donations remain largest in terms of overall charitable giving in the US at 32% of overall donations in 2011. Religion is such a complex and elusive topic that has caught the interest contemporary scholars, be it in the …elds of sociology, anthropology or political science. Scholars have made many claims about the economic consequences of religion, but none more in‡uential than those associated with Max Weber (1905). In The Protestant Ethic, Weber argues that capitalism arose in Europe in part because of how the belief in predestination was interpreted. The notion was that not everyone would be saved; there was only a speci…c number of the elect who would avoid damnation, and this 1
According to Pettitt (2011, p. 47-53) religious behaviour is thought to have emerged in the Upper Paleolithic (30,000 years ago) at the latest. The earliest evidence for treatment of the dead comes from Atapuerca in Spain. At this location the bones of 30 individuals believed to be Homo Heidelbergensis have been found in a pit. The earliest known burial of modern humans is from a cave in Israel located at Qafzeh. Human remains have been dated to 100,000 years ago. 2 For instance, Christianization of the Western world; Buddhist missions to East Asia; the decline of Buddhism in the Indian subcontinent; and the spread of Islam throughout the Middle East, Central Asia, North Africa and parts of Europe and India have all contributed to the rise of organized religions we see today.
2
was based sheerly on God’s predetermined will and not on any action a person could perform in this life. O¢ cial doctrine held that one could not ever really know whether one was among the elect. Leaders began assuring members that if they began doing well …nancially in their businesses, this would be one uno¢ cial sign they had God’s approval and were among the saved – but only if they used the fruits of their labor well. However, Weber’s hypothesis has been challenged in current literature often with empirical results suporting the opposite view3 . Although small in number, economists have long been studying the economics of religion. Adam Smith, for example, laid the foundation for an economic theory of religious institutions. Smith emphasized the importance of market structure, describing in detail the di¤erences between state-sponsored religious monopolies and competitive religious markets (Iannaccone, 1991). Some 200 years after Smith, the next systematic attempt was by Azzi and Ehrenberg (1975) where they developed a life-cycle model to analyze individuals’participation in religious activities. The subsequent literature on the economics of religion has since split into several strands which Iannacone (1998) has grouped into those which: …rst, apply microeconomic theory to explain religious behaviour; second, investigate the economic consequences of religiosity; and third, criticize economic policies while invoking religious precepts. Our contribution belongs in part to the …rst and second of these strands. In particular, we somewhat generalize Azzi and Ehrenberg’s (1975) seminal work by ignoring the distinction between male and female representative agents thus having only one type of agent which we call the "household". Our work di¤ers from theirs as we looked only at religious giving while their work featured two ‘devotional’choice variables, church attendance and religious donations. The work that also closely resembles ours is the one by Blomberg, DeLeire and Hess (2006). Their paper features a dynamic consumer optimization model with mortality in which intra-temporal utility stems from both consumption and religious contributions. Individuals also decide how to allocate resources between religious contributions which have both a this-life consumption value and an afterlife investment value and other consumption expenditures. They found evidence that individuals behave as if religious contributions have a value in the after-life, in a manner consistent with the after life-cycle model. Although Blomberg, DeLeire and Hess (2006) laid down the foundation on how afterlife beliefs a¤ect individual’s lifetime choices our contribution carefully studies the afterlife bene…ts in a quite general setting. In particular, we analyze and compare how beliefs about the afterlife bene…ts and the input-output relationship in the religious faith production function a¤ect lifetime 3
See Iannaccone (1999) for a list of studies which supported this opposite view
3
decisions, the competitive equilibrium dynamics and the long-run standards of living. We then test our claims empirically in both in the micro- and macro-levels. Before presenting the model, some of its main results will be previewed. First, we …nd empirical evidence that households that exhibit a high degree of religiosity tend to save more on average. Moreover, if one will decompose religiosity in terms of how households in‡uence their afterlife position through salvi…c merit, we …nd that Jewish and Pentecostals actually save more than other religious groups, including atheists and agnostics. Second, we were able to show theoretically that an economy with a signi…cantly high number of believers (adherents) in the afterlife save more and have a higher standard of living than an economy with fewer number of believers. Finally, we …nd empirical evidence that countries with a signi…cant number of adherents exhibit a high national savings rate. The rest of the paper is structured as follows. In section 2 we present our theoretical model. The empirical evidence of the impact of religiosity in the household and aggregate (national) levels are shown in sections 3 and 4, respectively.
2
The Model
2.1
Religious Households
Let us consider an in…nitely lived overlapping generations economy with …rm production, physical capital and religious households. At every time t 2 Z+ , a new generation of Nt identical households living for two periods denoted by i 2 f1; 2g are born. Pop-
ulation is assumed to grow at a constant geometric rate n 2 ( 1; +1). In the …rst period, households supply inelastically 1 unit of labor services at the real wage wt . We assume that the net labour income can be allocated between consumption: c1t , savings: st and physical religious contribution: d1t . Therefore, the …rst-period budget constraint is: c1t + st + d1t = wt
(1)
In period 2, the representative household born at time t retires and receives the principal plus the interest on his …rst-period savings represented by Rt+1 st . We assume that the net income can be allocated between consumption: c2t+1 and physical religious contribution: d2t+1 . Therefore, the second period budget constraint can be written as: c2t+1 + d2t+1 = Rt+1 st
4
(2)
where Rt+1 stands for the real interest factor. Let us consider that the lifetime objective function of the representative household depends on his consumption and religious contributions over his two periods of life. Let us assume that the objective function V (c1 ; c2 ; d1 ; d2 ) : R4+ ! R is twice continuously di¤erentiable, strictly quasiconcave, increasing in all variables with V11 < 0, V22 < 0,
V33 < 0, V44 < 0 and guarantees that starvation is avoided4 . For simplicity and without loss of generality, let us consider that the lifetime objective function of the representative household born at time t is represented by the following sum: (3)
V (c1t ; c2t+1 ; d1t ; d2t+1 ) = U (c1t ; c2t+1 ) + G(d1t ; d2t+1 )
where U denotes a strictly quasiconcave lifetime utility function, G represents a strictly quasiconcave after-life bene…t function that is expected to prevail at every period in > 0 stands for a weight parameter5 .
the after-life and
The G(d1t ; d2t+1 ) function depends on lifetime religious contributions or devotional goods, which captures the perceived bene…ts associated with the beliefs in the existence of the after-life. The motivation for the speci…cation of the after-life bene…t function can be traced from doctrines of organized religions. For instance, the Roman Catholic Church adopted indulgences which were sold as a way to obtain full or partial remission of temporal punishment. The earliest plenary indulgence was during the reign of Pope Urban II (1088-1099) and later on became a focal point of criticism during the Reformation. Among Muslims, giving away a …xed portion of one’s wealth or zakat, is a sacred duty and is mentioned in more than 30 di¤erent verses in the Quran. Giving zakat is essential for one’s salvation and those who give will receive rewards in the after-life. Devotional o¤erings or bhikshias is also practiced among Hindus. The lifetime planning problem of the representative religious household born at time t consists of maximizing (3) subject to (1) and (2) with wt and Rt+1 given. Let c^1t , c^2t+1 , d^1t , d^2t+1 denote the solution to that objective maximizing problem. The 4
The non-starvation assumption implies lim V1 (c1 ; c2 ; d1 ; d2 ) = +1 , lim V2 (c1 ; c2 ; d1 ; d2 ) = +1. c1 !0
5
c2 !0
The in…nite discounted sum of perceived after-life bene…ts for a household born at time t and evaluated at time t is: 1 X j H(d1t ; d2t+1 ) = 2 g(d1t ; d2t+1 ) j=0
where , stand for subjective in-life and after-life discount factors, respectively. If the above expression can be rewritten as: H(d1t ; d2t+1 ) = G(d1t ; d2t+1 ) where
=
2
=(1
).
5
2 (0; 1), then
…rst-order necessary conditions evaluated at the solution are: U^1 = Rt+1 U^2 U^1 =
^1 G
U^2 =
^2 G
(4)
According to the strict quasiconcavity of the objective function, the second-order su¢ ciency condition is satis…ed and the solution is a maximum6 . The objective maximizing level of consumption, religious contribution are functions of the real wage and the real interest factor. Therefore, the implied level of savings can be written as: (5)
s^t = S(wt ; Rt+1 )
By totally di¤erentiating the system of necessary conditions (4), the e¤ects of the real wage on consumption and religious contribution can be represented as follows: d^ c1t dwt dd^1t dwt d^ c2t+1 dwt dd^2t+1 dwt
= U^1 dRt+1 =0
= U^1 dRt+1 =0
= U^1 dRt+1 =0
= U^1 dRt+1 =0
where ^ U = U^22 U^11
(U^1 U^22 ^U G ^ 212
^2 U^2 U^12 ) G 12
^ 11 G ^ 22 G
2 ^ 11 G ^ 22 + ^ G U^12 G
2 U^12
^ 1G ^ 22 U^11 U^22 (G
^ 2G ^ 12 ) G
2 ^ 11 G ^ 22 + ^ G U^12 G
^ 212 ^U G
(U^2 U^11 ^U G ^ 212
^2 U^1 U^12 ) G 12
^ 2G ^ 11 U^11 U^22 (G
^U G ^ 212
U^11 U^22
(6)
^ 1G ^ 12 ) G
2 ^ 11 G ^ 22 + ^ G U^12 G
^ 11 2U^1 U^2 U^12 + U^12 U^22 < 0 and ^ G = G21 G
U^11 U^22
^ 11 G ^ 22 G
2 ^ 11 G ^ 22 + ^ G U^12 G
2 U^12
U^11 U^22
U^11 U^22
^ 1G ^ 2G ^ 12 + G ^ 2G ^ 2G 1 22
0, f 00 (k) < 0, lim (f (k)=k) < 1 and lim f 0 (k) = +1. k! +1
k! +1
Perfect competition drives factor prices to correspond to their marginal productivity: Rt = Zf 0 (kt ) wt = Zf (kt )
(9) kt Zf 0 (kt )
7
Normality of religious donations is supported by emprical evidence, for instance see Thornton and Helms (2013). 8 The positive e¤ect of real wage on savings is shown by the following expression: d^ st dwt
^2 =U dRt+1 =0
^2 U ^11 U
^1 U ^12 U
^2 G 12
^U G ^2 12
^ 11 G ^ 22 + G ^2G ^ 11 G ^ 11 G ^ 22 + ^ G U ^2 G 12
9
^1G ^ 12 G ^11 U ^22 U
^2 U 12
^11 U ^22 U (7)
See the appendix for the expressions which relate to comparative statics e¤ects of the real interest rate. An increase in Rt+1 raises the price of …rst-period commodity in terms of second-period. Individual households will substitute present co and donations for future period consumption and donations.Since the individual household is a lender, an increase in Rt+1 leads to a positive income e¤ect. If the substitution e¤ect dominates the income e¤ect then we see an increase in competitive level of savings.
7
2.3
Equilibrium
In equilibrium, the demands for physical capital and labour are determined by (9). Labour supply in the current period is Lt = Nt .
There is no market for physical
religious contributions. Assuming full depreciation of capital, the equilibrium condition for the physical capital rental market is: kt+1 =
S(Zf (kt )
kt Zf 0 (kt ) ; Zf 0 (kt+1 )) 1+n
(10)
It is well-known that the preceding …rst order di¤erence equation has two steady states f0; kSS g where kSS is the non-trivial steady state capital to labour ratio (for instance,
see Galor and Ryder (1989)).
3
Example
3.1
Household and Firms Setup
Let us consider an economy populated by households) and 1
proportion of believers (or faithful, religious
non-believers. For believers, let us assume speci…cally that their
preferences are represented by a time-separable lifetime logarithmic utility function: U (c1t ; c2t+1 ) = ln c1t + ln c2t+1 Consider further that their perceived after-life bene…t function takes the form: g(d1t ; d2t+1 ) = for i 2 f1; 2g where
i
1
ln d1t +
2
ln d2t+1
> 0 denotes the weight allocated to religious donations. There-
fore, following the speci…cation in (3), households maximize the sum of their lifetime utility and perceived after-life bene…ts: V = ln c1t + ln c2t+1 + ( where
2
=(1
). The parameter
1
ln d1t +
2
ln d2t+1 )
(11)
represents how individual households attach
weight to their after-life bene…ts. The lifetime planning problem of the representative household is to maximize (11) subject to (1) and (2) with the interest factor and
8
wage rate as given10 . From the solution to the optimization problem, we obtain the individual household’s competitive level of savings: sbt =
1+
+ + (
2 1
+
2)
wt
(12)
It can be easily shown that an increase in the weights allocated to religious donations and to their after-life bene…ts raises the competitive level of savings. An increase (decrease) in
1
raises (lowers) d1t and lowers (raises) c1t , c2t+1 and d2t+1 . On the other
hand, an increase (decrease) in
2
raises (lowers) d2t+1 and lowers (raises) c1t , c2t+1
and d1t . It can also be shown that an increase (decrease) in the weights associated with after-life bene…ts raises (lowers) donations and lowers (raises) consumption in both periods. The e¤ects of an increase in the weight to after-life bene…ts to savings is ambiguous, depending on the relative sizes of the weight parameters to religious donations11 . These e¤ects demonstrate the positive in‡uence religiosity and faith bring on individual household economic decisions. Non-believers have a di¤erent consumption and savings pro…les because for them = 0. Hence, their competitive level of savings is denoted by: st =
(13)
wt
1+
Comparing the two savings rates, it can be easily shown that the competitive level of savings for religious households is greater than the non-believers if the following condition holds:
2= 1
>
. Since religious households believe in the afterlife, they
seem to have a longer planning horizon, which might also in‡uence their decision to transfer current consumption in the future by saving. This condition is also equivalent to the second period religious giving (or ‘tithing’) ratio greater than in the …rst period, that is:
db2t+1 db1t > b c2t+1 b c1t
(14)
We can conclude here that religious households give more devotional goods during the 10
The solution to the religious household’s optimization problem are as follows: b c1t = Rt+1 wt R w w 2 t+1 t 1 t b c2t+1 = 1+ + ( 1 + 2 ) , db1t = 1+ + ( 1 + 2 ) , and db2t+1 = 1+ + ( 1 + 2 ) . 11 Di¤erentiating the savings function for believers with respect to : @b st = @ [1 +
which is postive (negative) when
2
2
+ (
1 1
+
is greater (less) than
9
2 wt 2 )] 1
.
wt 1+ + ( 1 +
2)
,
second period of their lifetime. Moreover, in order for religious households to give more in the period preceding their after-life, saving in the …rst period should be su¢ ciently high. Let us assume that the technology is represented by a Cobb-Douglas production function of the form: Yt = ZLt1 " Kt" where " is the share of physical capital inputs to production. Perfect competition among …rms drives the real wage to correspond to the marginal product of labour: wt = Z (1
") kt" . Similarly, because of perfect competition, the returns to physical
capital is: Rt = "Zkt" 1 .
3.2
Competitive Equilibrium
The physical capital in this economy at each period is formed from the savings of religious and non-religious households. Thus, the stock of physical capital per worker in this economy with heterogenous households is denoted by: )st + sbt
(1 + n)kt+1 = (1
(15)
Substituting the expression for savings for each household type in the steady state: kSS
Z (1 ") (1 + ) + ( 1 + 2 ) + ( 2 = 1+n (1 + + ( 1 + 2 )) (1 + )
1)
1 1 "
(16)
From the above expression for the physical capital in the steady state we can obtain two limiting cases depending on values of non-religious, that is
. If all households in this economy are
= 0, then we have the steady state physical capital akin to the
benchmark case: NR kSS
If all households are religious, R kSS
1 1 "
Z (1 ") = 1+n 1+
(17)
= 0, then we have:
Z (1 ") = 1+n 1+
It can be easily shown that as long as
2= 1
+ + ( >
1 1 "
2 1
+
2)
(18)
R NR then kSS > kSS > kSS . This result
implies that non-secular (or religious) economies should have higher savings, and thus have higher physical capital in the steady state, relative to secular (or non-religious) 10
economies. This relationship is summarized in Figure 1.
4
Household-Level Empirical Evidence
In this section we provide empirical results to support the …ndings from the theoretical model of religious giving and saving of religious households. We begin by describing the source data and how the variables, the religiosity and saving variables in particular, were de…ned. A discussion on savings pro…les and religiosity follows. The section winds up with the analysis of religious and non-religious determinants of saving.
4.1
Data and Variables
To carry out the analysis, we used the 2005 wave of the Panel Study of Income Dynamics (PSID), which provide information for spending, income and religious pro…les of around 9,014 American households. We use the 2005 wave of the PSID for several reasons: …rst, among the various waves, 2005 provides rich information on religious giving, preference and denomination of respondent households; and second, although the 2003 wave has information on religiosity, we may have to account for individuals switching religions, a route which we wish not to pursue. Household savings, the main variable of interest in this study, and its covariates are de…ned in the appendix. The 2005 PSID has information on household saving which is categorical in nature –whether or not the household saved in the previous year12 . Although useful in some studies, the categorical variable for saving does not re‡ect the nature of savings in our theoretical model which is expressed in goods. There is another variable in the 2005 wave which asks for amounts in savings and checking accounts of family members. However, this kind of information does not capture the dissaving of younger or retired individuals, a stylized fact in the life cycle models. Hence, as a crude measure, we construct our “saving” variable as follows: we sum up all of the expenditure items of reported by households; calculate the income tax for each household using TAXSIM913 ; and subtract the expenditures and income tax from the sum of head and spouse (if there is one) for each household. Since we limit our analysis to a single year, we do not make the distinction if saving is active or passive. Instead, we de…ne saving in the 12
Savings is coded as S704 in 2005 PSID: “Do you [or anyone in your family living here] have any money in checking or savings accounts, money market funds, certi…cates of deposit, government savings bonds, or treasury bills, not including assets held in employer-based pensions or IRA’s?”. 13 TAXSIM9 calculates federal and state income tax liability from 21 PSID input variables. See http://users.nber.org/~taxsim/ for further information on the tax simulation package.
11
simplest of terms –the amount of income that is not spent.
4.2
Categories of religiosity and afterlife bene…ts
Before we analyze the savings data and develop the empirical model, we need to understand how religions di¤er in how afterlife bene…ts are obtained and the e¤ect of religious dogma on economic behaviour. McCleary (2007) grouped several religious traditions according to the adherent’s probability of attaining bene…ts in the afterlife (or “salvation”). McCleary referred to this probability as “salvi…c merit”and, depending on adherent’s behaviour in his life, religions can give varying levels of probability of attaining afterlife bene…ts. There are four levels of “salvi…c merit” according to McCleary: high, medium, low and none. Religions belonging to high “salvi…c merit” include Pentecostals, Buddhists, Su… Islam and Judaism. In this category, afterlife bene…ts are gained through moral actions involving work, saving and charity. In the medium “salvi…c merit” category belong the Roman Catholics, Hindus and Sunni Islam. In the low “salvi…c merit”category are other religions like Protestants, Methodists and Calvinists among others. In the low category, religions o¤er salvation only through faith. In other words, although morally sound behaviour is encouraged, moral action in an individual’s lifetime does little in the attainment of afterlife bene…ts. Since belief in the afterlife is unobservable, we use grouping by “salvi…c merit” as a fair representation. We group the religious denominations in the PSID in the same fashion as McCleary’s14 . The attainment of afterlife bene…ts is a result of two distinct elements, according to Thornton and Helms (2013): First, a religion should have a doctrine on afterlife bene…ts or punishment, which includes the means of attaining them; and second, the adherent, in his lifetime and conditional on religious a¢ liation, should be allowed to in‡uence his probability of attaining afterlife bene…ts through moral actions. McCleary notes that there are religions where, although there is a concrete notion of afterlife, the attainment of bene…ts in the afterlife cannot be in‡uenced by moral action. For instance, among Protestants, Martin Luther’s concept of salvation by faith e¤ectively “decouples”moral actions from any afterlife consequences. Thus, to capture the e¤ect of moral action to afterlife beliefs, we use religious donations and religious attendance as proxies. 14 It should be noted that McCleary (2007) does not provide for the grouping for adherents of Judaism. In this regard, we classi…ed Judaism as “high salvi…c merit”following Thornton and Helms (2013).
12
4.3
Pro…les of household saving
Table 1 shows the saving pro…les according to the various religiosity variables. It is evident that adherents who belong to certain religions with high and medium salvi…c merit save more than their counterparts who belong to the low or no salvi…c merit categories. For instance, adherents in the high salvi…c merit category save 20% more than those who do not belong to any religion or agnostics. It is also interesting to note that adherents who are in the medium category (mostly Roman Catholics) save the most relative to the other groups. In fact, as shown in Table 1, adherents in the medium salvi…c category save by around 65% more than those belonging to the low salvi…c merit category. The di¤erences in saving across salvi…c merit groupings are signi…cant, in particular for high, medium and none. There are also large di¤erences in saving rate, de…ned as the ratio of saving to disposable household income, across households and salvi…c merit levels. As expected, high salvi…c merit households save more. Surprisingly, the level of saving of low salvi…c merit households is less than atheists and agnostics. However, in terms of saving rate, households with no salvi…c merit - atheists and agnostics - save the least overall. In terms of variables on moral action, we …nd that households who go to church on a weekly basis save the least (19% saving rate). However, di¤erences attributed to moral actions – religious giving or church attendance –are not statistically signi…cant. The interactions between moral action variables and salvi…c merit provide some more interesting insights on their e¤ects to saving. As shown in Table 1, households who are religious tend to save twice as more if they go to church often, and these di¤erences are signi…cant at varying degrees depending on their level of salvi…c merit. For instance, households with high salvi…c merit and goes to church once a week tend so save close to twice as much those with high merit who don’t attend church regularly. The same trend can be observed on households with other levels of merit, although their levels of signi…cance vary. There is also evidence that those who go to church often save more than those who don’t. The relationship between age and saving has always been a nonlinear one as shown in Figure 2. In the …gure, we ran a regression on saving on age, its square and other exogenous variables (which were held constant in the …gure). As predicted in the theoretical literature, younger and older households tend to have negative or low saving while households in their working age save. However, remarkable in this result are the signi…cant di¤erences in saving by religiosity. Medium salvi…c merit households save the most but in their old age save less than adherents in the high salvi…c merit
13
category. Moreover, those in the high salvi…c merit category, on average, save more than other groups across age groups. This re‡ects the importance of moral actions in this case, saving – by religions belonging to this group15 . One should be careful, however, in drawing conclusions whether religiosity drives these results. For example, Keister (2003) notes that Jews have generally high levels of educational attainment which suggest that wealth accumulation is likely to exceed that of other groups by a considerable margin. Another notable result from Figure 1 is the low saving for households belonging to the low salvi…c merit category, which is in direct contrast to those in the high salvi…c merit category. In fact, saving for these households – who mostly belong to Protestant groups - is even lower than atheists and agnostics. Lehrer (1999) notes “that traditional attitudes among Protestants has traditional attitudes have translated into relatively high fertility rates that are likely to decrease wealth accumulation by diluting both material and nonmaterial resources during childhood, reducing inheritance, and making saving more di¢ cult in adulthood”.
4.4
Religious Determinants of Savings
Table 2 presents the full results using various estimation methods, OLS and IV. In columns 1 and 2 we report the results using OLS where we estimate the e¤ects of religiosity and socio-demographic variables to household saving. The estimated coef…cients for salvi…c merit categories were all positive and signi…cant, relative to those belonging to no salvi…c merit category. The coe¢ cient for the high salvi…c merit is higher than medium and low, which re‡ects what we have observed earlier. On the other hand, the coe¢ cient for religious donations in column 2 is negative and not significant. Note that these indicators of religiosity are dummy dependent variables, and are signi…cantly di¤erent from zero and positive, thus they represent upward parallel shifts in the regression line. Households belonging to high salvi…c merit religions save $9,423 more than atheists and agnostics. Similarly, medium and low salvi…c merit households save more than those without salvi…c merit. The results for remaining explanatory variables in columns 1 and 2 follow conventional wisdom about saving behavior. The relationship between saving and age is nonlinear. Married households tend save more than single, widowed or divorced. Households with at least college education save more than those whose completed education is at most high school. The marginal propensity 15 For instance, Keister (2003) notes that Jewish religious a¢ liation has a direct e¤ect on wealth accumulation of its members. In Jewish culture, accumulation is seen as an indicator of success, and Jews have been shown to have a highly positive cultural orientation to education and occupational status.
14
to save is at $0.15 which is statistically signi…cant and positive. It is possible that there might be a reverse causality between religious donations and savings. For instance, households donate more to their religion because they save more or it could also be possible that they save more in order for them to give more to their faith. Either way, there could be an endogeneity problem which can be addressed using the IV method. We follow Thornton and Helms (2013) and use federal and state marginal tax rates as instruments to religious donations. The intuition for using marginal tax rates is that it can be used as proxy for the opportunity cost of religious giving16 . “The ability to deduct charitable contributions from ordinary income serves as an implicit subsidy from the government for charitable activities”. The results of the IV method highlight the fact we have seen earlier – household with low salvi…c merit save less than atheists or agnostics. Low salvi…c households save less by around $11,843 relative to those with no salvi…c merit. Moreover, the coe¢ cient for religious donations is positive and signi…cantly di¤erent from zero; for every one dollar increase in religious donations, households save approximately $61, ceteris paribus.
5
Cross-Country Empirical Evidence
In this section we provide empirical evidence in the aggregate level. To carry out the analysis we construct a panel of socio-economic variables and those which relate to religiosity for the four time periods and 204 countries. Our dependent variable is the average of shares of real savings to GDP for the four periods: 1975-1984, 1985-1994, 1995-2004 and 2005-2011. Socio-economic data is taken from the World Development Indicators database and the description of data is listed in the appendix. Data on religiosity was taken from various sources described in the next section.
5.1
Religiosity Data and Variables
Variables related to religiosity as used in this paper are grouped into three: Religious beliefs, adherence and policy. Beliefs on the existence of God, life after death, heaven, and hell were grouped under religious beliefs. Data on religious beliefs were primarily sourced from the various waves of the World Values Survey (WVS): 1981-1984, mostly 1981; 1989-1993, mostly 1990; 1994-1999, mostly 1996 and 1996; and 2005-2007. We supplement the WVS with religious data from the European Values Survey (EVS) 16
According to Thornton and Helms (2013) if a donor faces a marginal tax rate of t, then each dollar of deductable contributions only will cost that donor (1–t) dollars.
15
because not all countries have information on religious beliefs from the WVS. We included data on religious beliefs for 1975-1984 and 2005-2011 periods from various EVS waves: 1981-1984, mostly 1981; and 2008-2009, mostly 2008. Barro and McCleary (2003, p. 762) …rst assembled cross-country data for religiosity and was able to have information for 59 countries. In our update, and putting various sources of data on religious beliefs together, we were able to …nd data for a maximum of 68 countries, depending on the measure of religious beliefs. The second group, religious adherence, was also assembled from various sources. Religious adherence is the religion by which a respondent adheres to as professed in various surveys. We follow the religious groupings in Barro and McCleary (2003, p. 764): Christians, Jews, Muslim Hindu, Eastern and Other religions. Data for 1980 and 1990 adherence shares were sourced from Durlauf, Kourtellos and Tan (2011) for 1975-1984 and 1985-1994, respectively. For the rest of the time periods, 1990 and 2000 adherence, data were sourced from the World Christian Encyclopedia and Robert Barro’s website, for the 1995-2004 and 2005-2011 time periods respectively. The third group, religious policy, pertains to religions and institutions. There are three variables in this group: presence of o¢ cial religion, government regulation of religion and religious pluralism. Data for o¢ cial religion and government regulation of religion, which take in the form of a dummy variable (1 if present, 0 otherwise), were sourced from Durlauf, Kourtellos and Tan (2011). 57 countries have o¢ cial or state religions and some form of regulation on religion. The index of religious pluralism was constructed from Barro and McCleary’s (2003, p. 764) speci…cation. From the various religion shares we can calculate the Her…ndahl index of religious concentration for each country17 . Religious pluralism is one minus the Her…ndahl index, with a low value indicating the presence of a dominant religion18 . Consistent with Barro and McCleary (2003) and Durlauf, Kourtellos and Tan (2011) we also control for additional covariates. These variables measured respectively for each period are: real GDP, consumer price in‡ation rate, ratio of exports minus imports to real GDP, fertility rate, infant mortality rate, life expectancy, ratio of people aged 0 to 14 plus 65 and over to total population, Freedom House measure of property rights, freedom of elections, civil liberties and economic freedom. Despite the lack 17
The Her…ndahl index is the sum of the squares of the religious adherence shares. It can be interpreted as the probability that two randomly selected individuals belong to the same religion (Barro and McCleary, 2003). 18 Countries with low pluralism index include: Spain (0.16), Ecuador (0.11), and Colombia (0.11), Catholic countries; Mauritania (0.02), Afghanistan (0.04), and Libya (0.08), Muslim; and Greece (0.15), orthodox. Countries with high pluralism index include the US (0.72), Canada (0.76), Singapore (0.74) and South Korea (0.82).
16
of correspondence between the religious beliefs, adherence, policy and other control variables, and the time periods in our dataset, we argue that the use of these variables might still be satisfactory since they are typically slow moving and exhibit a high degree of persistence19 . Table 4 in the appendix shows the descriptive statistics for the di¤erent religiosity variables. Overall, on average, 84% of those surveyed across countries believes in the existence of God or some supreme being. Around 60% believes in the existence of the afterlife and heaven, while a smaller number (45%) believes in the existence of Hell. Even with the update from the EVS our descriptive statistics for religious beliefs are close to results in previous studies20 . In terms of religious adherence, our sample tend to be more Christian (58%) and richer than average.
5.2
Cross-country regression results
Table 5 in the appendix presents the results of the cross-country panel regressions with random e¤ects21 . Column (1) reports estimates without the beliefs in hell and heaven, column 2 reports results from the full model while column 3 omits religious belief variables. E¤ects of religiosity in aggregate savings rate are somewhat mixed. The coe¢ cient for belief in god is highly signi…cant but negative across columns (1) and (2). This result indicates that countries with a large number of believers in the existence of God tend to save less. The coe¢ cient for the belief in the existence of the afterlife is positive and highly signi…cant in column (1) and signi…cant at 10% in column (2). These results, although somewhat puzzling, provides for some evidence that belief in the afterlife is an important determinant of savings rate in the aggregate. Belief in heaven and hell are not signi…cant across the three regression models although their signs seem to agree with results from previous studies. Other determinants like religious adherence and policy are not signi…cant in columns (1) and (2). Column (3) reports the regression results controlling for religious adhrence and policy. In this model there is evidence that higher adherence rates result to higher savings rate. For instance, a higher proportion of adherents for muslim, eastern and other religions tend to raise savings rates. To deal with the possible endogeneity of the determinants of savings rate, we follow 19
This was the same argument of Durlauf, Kourtellos and Tan (2011). The data from Barro and McCleary (2003, p. 766) had 80%, 58%, 55% and 38% for the beliefs in the existence of God, afterlife, heaven and hell, respectively. 21 We conducted a Hausman test comparing estimates from random and …xed e¤ects; the result of the test is in favour of random e¤ects. 20
17
Barro and McCleary (2003) and instrument the religious belief variables with the the dummy variable for state regulation of religion, dummy variable for the presence of state-sanctioned religion and the measure of religious plurality. We also include the adherence shares for each religious grouping as instruments and use the same independent (control) variables as in the random e¤ects regressions. These variables are highly signi…cant in explaining religious beliefs and this we are not concerned that these instruments are weak. Table 6 in the appendix shows the results of the instrumental variables regressions. Column (1) presents the results of the instrumental variables regression of savings rate with all religious belief variables - beliefs in the existence of God, the afterlife, heaven and hell - as endogenous variables. None of these religious beliefs are signi…cant although some of the instruments are valid. We suspect that the insigini…cance of these variables is a result of them being correlated with each other. Hence, to address this issue, take only two of the religious belief variables at a time, that is, belief in God and afterlife in column (2), and beliefs in heaven and hell in column (3).
The regression results dramatically improve when we take one set of
variables at a time. From column (2) the coe¢ cients for belief in God and the afterlife remain to be highly signi…cant negative and positive, respectively. On the other hand, from column (3) we …nd that belief in heaven and hell are negative and positive at 10% signi…cance, respectively. These results mirror what we have seen earlier in the random e¤ects models in Table 5. Therefore, we can say that belief in the afterlife is an important determinant of savings rate.
6
Conclusions
In this paper we were able to construct a theoretical model which can explain the impact of religiosity in economic decisions of households. We augmented the benchmark overlapping generations framework with both a perceived after-life bene…tt function and a religious faith production technology that relies on tangible lifetime donations. In the resulting model, we were able to analyze and compare how beliefs about the afterlife bene…ts and the input-output relationship in the religious faith production function a¤ect lifetime decisions, the competitive equilibrium dynamics and the longrun standards of living. We were able to show theoretically that an economy with a signi…cantly high number of believers (adherents) in the afterlife save more and have a higher standard of living than an economy with fewer number of believers. Next, we determine if the results from the theoretical model is supported by empirical evidence. We …nd that households that exhibit a high degree of religiosity tend 18
to save more on average. Moreover, if one will decompose religiosity in terms of how households in‡uence their afterlife position through salvi…c merit, we …nd that Jewish and Pentecostals actually save more than other religious groups. Also, we …nd empirical evidence that countries with a signi…cant number of adherents exhibit a high national savings rate. The results are robust even if we consider the possible endogeneity of religious beliefs and its impact to the savings rate.
19
References Azzi, C. and R. G. Ehrenberg, (1975): "Household allocation of time and church attendance", Journal of Political Economy 83(1), 27-56. Barro, R. and R. McCleary, (2003): "Religion and economic growth across countries", American Sociological Review 68, 760-781. Blomberg, S. B., DeLeire, T. and G. Hess (2006): "The (After) Life-Cycle Theory of Religious Contributions", CES-ifo Working Paper No. 1854. Durlauf, S. N., Kourtellos, A. and C. M. Tan (2011): "Is God in the details? A reexamination of the role of religion in economic growth", Journal of Applied Economics doi: 10.1002/jae.1245. Gallup Poll, May 5-8, 2011. Poll questions retrieved on April 2013, from http://www.gallup.com/poll/147887/americans-continue-believe-god.aspx. Galor, O. and H. Ryder, (1989): "Existence uniqueness and stability of equilibrium in an overlapping generations model with productive capital", Journal of Economic Theory 49:360-375. Iannaconne, L. (1998): “Introduction to the economics of religion,”Journal of Economic Literature, 36, 1465–1496. Keister, L. (2003): "Religion and Wealth: The Role of Religious A¢ liation and Participation in Early Adult Asset Accumulation", Social Forces. 82:173-205. Lehrer, E. (1999): "Religion as a Determinant of Educational Attainment: An Economic Perspective", Social Science Research, 28(4): 358-379. McCleary, R. (2007): "Salvation, Damnation, and Economic Incentives", Journal of Contemporary Religion 22, 1: 49-74. Pettitt, P. B. (2011): "The Palaeolithic Origins of Human Burial", Routledge, New York USA. Simon, P. and L. Blume (1994): “Mathematics for economists”, W.W. Norton and Company, New York USA. Thornton, J. and S. Helms (2013): "Afterlife incentives in charitable giving", Applied Economics 45 (19), 2779-2791. 20
Weber, M. (1930): "The Protestant Ethic and the Spirit of Capitalism", Allen and Unwin, London, UK.
21
7 7.1
Appendix Proofs Comparative statics
By totally di¤erentiating the system of necessary conditions (4), we can express the comparative static e¤ects of the real wage on consumption and religious contribution as follows: d^ c1t dwt dd^1t dwt d^ c2t+1 dwt dd^2t+1 dwt
= U^1 dRt+1 =0
= U^1 dRt+1 =0
= U^1 dRt+1 =0
= U^1 dRt+1 =0
where ^ U = U^22 U^11
(U^1 U^22 ^U G ^ 212
^ 11 G ^ 22 G
2 ^ 11 G ^ 22 + ^ G U^12 G
2 U^12
^ 212 ^U G
^ 1G ^ 22 U^11 U^22 (G
^U G ^ 212
^ 212 U^1 U^12 ) G
U^11 U^22
^ 11 G ^ 22 G
2 ^ 11 G ^ 22 + ^ G U^12 G
^ 2G ^ 11 U^11 U^22 (G
2 U^12
U^11 U^22
^ 2G ^ 12 ) G
2 ^ 11 G ^ 22 + ^ G U^12 G
(U^2 U^11
^U G ^ 212
^2 U^2 U^12 ) G 12
^ 1G ^ 12 ) G
2 ^ 11 G ^ 22 + ^ G U^12 G
^ 11 2U^1 U^2 U^12 + U^12 U^22 < 0 and ^ G = G21 G
(19)
U^11 U^22
U^11 U^22
^ ^ 1G ^ 2G ^ 12 + G ^ 2G 2G 1 22
0, U2 > 0, U12 = U21 > 0, U11 < 0, 23
^2 G 2
2 ^ 11 G ^ 22 + ^ G U^12 G
The bordered Hessian matrix associated with the utility function in (??) is: 2
^ 11 U^ U^1 U^12 + G
i
2 U^12
U^1
U^11 U^22
U22 < 0, G1 > 0, G2 > 0, G12 = G21 > 0, G11 < 0 and G22 < 0. It can be shown that the …rst leading principal minor, H1 , is negative: (U1 )2
H1 =
The second leading principal minor, H2 , is positive: H2 =
U11 U22 + 2U1 U2 U12
U22 U12
2 The third leading principal minor, H3 , is negative if U11 U22 > U12 :
H3 = G11 H2
2 U12 G21
U11 U22
2 The last leading principal minor, H4 , is positive if U11 U22 > U12 and G11 G22 > G212 :
H4 = where
U
= U11 U22
U
G11 G22
G212
2U1 U2 U12 + U22 U12 and
G
G
U11 U22
= G11 U22
2 U12
2U1 U2 G12 + G22 U12 . Thus,
2 the leading principal minors are guaranteed to have alternate signs if U11 U22 > U12 and
G11 G22 > G212 are satis…ed, and consequently, V is quasiconcave.
24
Figure 1: Capital accumulation and steady states contingent on the proportion of religous households, . As long as the religious giving ratio in the second period is larger than in the …rst, we can observe that religious households tend to save more than the non-religious.
1
Figure 2: Average saving by age groups (in 2005 prices). Savings response functions generated from a quadratic …t of age of head of household, by age groups. The curve in blue represent high salvi…c merit religious groups (Judaism and Pentecostals), medium salvi…c merit in red (Roman Catholics and Muslims), low salvi…c merit religions in green (other Protestants) and no salvi…c merit is orange (atheists and agnostics).
2
1
13,494.10 21,471.30 19,936.40 18,425.20 11,036.10 11,688.00 23,813.40 14,661.30 24,500.20 13,347.80 17,219.50 7,236.70 11,437.90 7,614.10 16,580.10 12,144.40
0.31 0.32 0.28 0.27 0.22 0.27 0.3 0.36 0.3 0.24 0.28 0.21 0.18 0.23 0.22 0.25 0.22
Total
12,663.20
9,288.90 10,802.30 14,703.30 15,970.00 12,000.50 10,540.00 20,672.90 7,030.20 18,937.70 14,485.50 16,219.10 8,355.00 14,913.70 7,195.50 17,858.10 11,633.70
12,084.30 16,537.00 11,119.70
10,065.00 15,812.20 11,039.00 15,323.10
Median
4,015
61 122 251 599 441 875 83 100 421 429 558 758 970 316 959 2,258
1,286 1,929 2,574
183 850 1,316 1,666
N
1.78 1.78 -0.2 -0.2 0.2 0.2 -1.8 -1.8 -2.3 -2.3 -3.1 -3.1 -1.1 -1.1 -1.4 -1.4
1.08 -1 -1
-1.1 -2.7 -2.3 -2.8
T statistic
* * ** ** *** ***
* *
**
** ***
Dummy variables for salvi c merit are: high salvi c merit (1 if household head is an adherent of Judaism and Pentecostals, 0 otherwise); medium salvi c merit (1 if household head is Roman Catholic or Muslim, 0 otherwise,); low salvi c merit (1 if household head belong to any other Protestant denomination, 0 otherwise); no salvi c merit (1 if household is atheist or agnostic, 0 otherwise). Dummy for moral action variables are: church attendance (1 if household head attends goes to church every week, 0 otherwise) and religious giving (1 if household gave money for religious causes or purpose, 0 otherwise). Saving rate is the ratio of household saving to disposable (after tax) income. The column for the T statistic tests the null hypothesis of equality of means between two groups. Signi cance levels: *** 1%, ** 5% and * 10%.
a Note:
10,498.30 13,994.30 11,588.20
0.19 0.2 0.25
12,619.00
18,812.30 18,871.50 11,469.50 15,720.30
Mean
0.32 0.28 0.25 0.24
Share of saving to disposable family income
Levels of salvi c merit High salvi c merit Medium salvi c merit Low salvi c merit No salvi c merit Moral actions Weekly church attendance Positive religious donations Zero religious donations Interactions High salvi c merit x attends church weekly High salvi c merit x does not attend church weekly Medium salvi c merit x attends church weekly Medium salvi c merit x does not attend church weekly Low salvi c merit x attends church weekly Low salvi c merit x does not attend church weekly High salvi c merit x positive religious giving High salvi c merit x zero religious giving Medium salvi c merit x positive religious giving Medium salvi c merit x zero religious giving Low salvi c merit x positive religious giving Low salvi c merit x zero religious giving Attend church weekly x positive religious giving Attend church weekly x zero religious giving Does not attend church weekly x positive religious giving Does not attend church weekly x zero religious giving
Religiosity variables
Table 1: Household saving (in 2005 calendar year prices) and religiosity variablesa .
OLS 1
* ** *
*** ** *** *** * *** *** ***
SE
*** *** *** *** * *** *** ***
* ** *
OLS 2
Coe cient 9482.75 6557.34 4594.57
0.15 894.42 -14.44 -12538.97 -10760.56 -9848.51 -16867.97 -6327.15
5577 2957.5 2568.5
0 402.7 4.1 3033.3 5537.7 3238.8 5267.3 2201.2
2096.43
-0.9
9521.8
4502 0.08
5371.6 30845.75 -11843.69
Coe cient
**
***
*** **
0.03 967.29 8.64 9441.39 14477.77 10462.2 12740.85 7026.69
11.2
10920.77 7260.54 5833.99
IV SE
60.96
*** *** *** *** * ***
SE 5576 2964.8 2572.2
0.07 -2214.64 4.07 28062.88 42386.39 36178.05 24631.09 24370.73
0.5 0 403.6 4.1 3054.2 5556.8 3263.9 5279.3 2217.7
18957.25
22268.14 4502
9521.8
Table 2: Household saving response functionsa Coe cient 9423.08 6909.83 4356
2389.18
0.15 849.3 -14.17 -11949.73 -9989.26 -9180.54 -16265.71 -5881.64
Constant
4502 0.08
Salvi c merit categories High Medium Low Moral action Religious giving Socio-demographics Disposable income Age Age squared Single Widowed Divorced Other marital status High school
N R squared
a Note: Columns 1 and 2 were estimated using OLS and instrumental variables for column 3. The de nitions for dependent variables are described in the appendix. The de nitions of variables used in the analysis are described in the appendix. For column 3, we use the federal and state marginal tax rates (frate and srate) as instruments, with p values of 0.001 and 0.000 in the rst stage regression, respectively. Asterisks are signi cance levels: *** 1%, ** 5% and * 10%.
2
3
Single Widowed Divorced Other marital status
High school
Weekly church attendance
Religious preference
Household savings Religious donations Age Age squared Married Religious denomination
Income tax
Household disposable income
Description Sum of household expenditure items in 2005 prices: housing, home repairs, home furnishing, utilities, clothing, vacation, recreation, health, food, transportation, education, and childcare. Includes donations to dierent causes (i.e. needy, community, religious, etc.). Sum of income of head of household and spouse less federal, state and FICA tax liabilities. Sum of federal, state income and FICA tax liabilities from 21 input variables. Calculated from TAXSIM9, NBER's FORTRAN program for calculating liabilities under US Federal and State income tax laws from individual data. Household income less expenditures and income tax. Amounts in 2005 prices of charitable giving towards religious causes. Age of head of household in years Square of Age. Dummy variable: 1= if head of household is married, 0=otherwise. Religious denomination of household head from 21 dierent Christian traditions: Baptist, Lutheran, Methodist, Presbyterian, Episcopalian, Mormons, Jehovah`s Witnesses, Christian, Unitarian, Christian Science, Seventh Day Adventists, Pentecostal, Amish, Quaker, Church of God, United Church of Christ, Reformed, Disciples of Christ, Churches of Christ, Other and None (including atheists and agnostics). Religious preference of head of household: Catholic, Jewish, Muslim, Eastern Orthodox, Other and None (including atheists and agnostics). Dummy variable: 1= if head of household attends church services at least once a week, 0=otherwise. Dummy variable: 1= household has at most completed high school education, 0=otherwise. Dummy variable: 1=household is single, never married, 0=otherwise. Dummy variable: 1=household is widowed, 0=otherwise. Dummy variable: 1=household is divorced, 0=otherwise. Dummy variable: 1=household has other marital status, 0=otherwise.
Table 3: Variables used in the analysis sourced from 2005 PSID.
Variable name Household expenditures
Table 4: Descriptive statistics for religiosity variables used in the analysis Mean Standard Deviation Number of countriesa 0.84 0.19 21, 11, 32, 68 0.63 0.21 22, 13, 32, 67 0.6 0.25 52, 49, 32, 47 0.45 0.27 52, 49, 32, 47 0.58 0.26 103, 103, 162, 162 0.01 0.06 172, 172, 187, 187 0.24 0.36 172, 172, 187, 187 0.02 0.1 172, 172, 187, 187 0.05 0.18 172, 172, 187, 187 0.1 0.19 191, 191, 187, 187 0.26 0.44 57, 57, 57, 57 0.34 0.47 57, 57, 57, 57 0.45 0.23 188, 188, 187, 185
God Afterlife Heaven Hell Christian Jews Muslim Hindu Eastern Other O cial religion Government regulation Pluralism
a Note: Figures in the third column are number of countries for each of the four waves where data is available: 1975-1984, 1985-1994, 1995-2004, 2005-2011
4
5 0.009 0.02 0.003 0.019 -0.023 0.046 0.409 68
-0.004 0.018 0.005 0.017 -0.06 0.041 0.447 80
-0.059 0.017 0.05 0.028 0.038 0.03 -0.022 0.027 -0.777 2.337 -0.006 0.049 -0.042 0.11 0.059 0.055 -0.46 0.381
***
***
-0.213 2.025 0.008 0.034 -0.066 0.103 0.062 0.045 -0.211 0.181
-0.057 0.013 0.064 0.014
2
*
***
0.001 0.011 0.016 0.01 -0.003 0.022 0.304 451
-0.089 0.055 0.06 0.016 0.083 0.037 0.118 0.027 0.06 0.035
2
*
***
**
***
Figures in beneath the estimates are standard errors of coe cients. Other explanatory variables were included but not shown. The dependent variable is the average savings rate - the share of real savings to GDP - for each of the four 10-year periods: 1975-1984, 1985-1994, 1995-2004, 2005-2011. The explanatory variables not shown are real GDP, consumer price in ation rate, ratio of exports minus imports to real GDP, fertility rate, infant mortality rate, life expectancy, and ratio of people aged 0-14 plus 65 and over to total population in 1980, 1990, 2000 and 2010; also included are Freedom House measures of property rights, freedom of elections, civil liberties and economic freedom; and the religion variables are described in text. Asterisks are signi cance levels: *** 1%, ** 5% and * 10%.
a Note:
R squared Observations
Pluralism
Government regulation
Religious Policy O cial religion
Other
Eastern
Hindu
Muslim
Religious Adherence Jews
Hell
Heaven
Afterlife
Religious beliefs God
1
Random Eects
Table 5: Eects of religiosity in real savings as percentage of real gross domestic product a
Coe
SE
Coe
-0.075 *** 0.025
136
Coe
-0.034 0.037
0.09
SE
God 0.016 0.052
-0.048 * 0.028
111
0.165
SE
Afterlife
-0.019 0.061
0.052 * 0.024
0.245
11.509
*** 0.027
Heaven
0.034 0.046
Table 6: Instrumental variables regressions on eects of religiosity in real savings as percentage of real gross domestic product I II III
a
Hell
0.286
5.431
91
R squared
2.298
N
Test of overidentifying restrictions
a Note: Figures in beneath the estimates are standard errors of coe cients. Other explanatory variables were included but not shown. The dependent variable is the average savings rate - the share of real savings to GDP - for each of the four 10-year periods: 1975-1984, 1985-1994, 1995-2004, 2005-2011. The explanatory variables not shown are real GDP, consumer price in ation rate, ratio of exports minus imports to real GDP, fertility rate, infant mortality rate, life expectancy, and ratio of people aged 0-14 plus 65 and over to total population in 1980, 1990, 2000 and 2010; also included are Freedom House measures of property rights, freedom of elections, civil liberties and economic freedom; and the religion variables are described in text. Asterisks are signi cance levels: *** 1%, ** 5% and * 10%.
6
