Randomizing Religion: The Impact of Protestant Evangelism On Economic Outcomes. Gharad Bryan, James J. Choi, and Dean Karlan. Jun 2020. https://economics.harvard.edu/files/economics/files/ms29321.pdf
Abstract: We study the causal impact of religiosity through a randomized evaluation of an evangelical Protestant Christian values and theology education program delivered to thousands of ultra-poor Filipino households. Six months after the program ended, treated households have higher religiosity and income; no statistically significant differences in total labor supply, consumption, food security, or life satisfaction; and lower perceived relative economic status. Exploratory analysis suggests that the income treatment effect may operate through increasing grit. Thirty months after the program ended, significant differences in the intensity of religiosity disappear, but those in the treatment group are less likely to be Catholic and more likely to be Protestant, and there is some mixed evidence that their consumption and perceived relative economic status are higher. We conclude that this church-based program may represent a method of increasing noncognitive skills and reducing poverty among adults in developing countries.
I. INTRODUCTION
A literature dating back at least to Adam Smith and Max Weber has argued that religiosity is
associated with a set of characteristics that promote economic success, including diligence,
thriftiness, trust, and cooperation (Iannaccone 1998; Iyer 2016). More recent research has linked
religiosity to positive outcomes in domains such as physical health (Ellison 1991), crime rates
(Freeman 1986), drug and alcohol use (Gruber and Hungerman 2008), income (Gruber 2005), and
educational attainment (Freeman 1986; Gruber 2005). Other studies have argued for negative
economic effects of some aspects of religiosity due to a focus on otherworldliness (Weber [1905]
1958 in his discussion of Catholicism) and substitution toward church attendance away from
production (Barro and McCleary 2003). Despite extensive research, claims that religion causes
outcomes remain controversial, in part because people choose their religion. Naturally occurring
religious affiliation is likely to be correlated with unobserved personal characteristics, which may
be the true drivers of the observed correlations. Iannaccone (1998) writes that “nothing short of a
(probably unattainable) ‘genuine experiment’ will suffice to demonstrate religion’s causal impact.”
To study the causal impact of religiosity, we partnered with International Care Ministries
(ICM), an evangelical Protestant anti-poverty organization that operates in the Philippines, to
conduct an evaluation that randomly assigned invitations to attend Christian theology and values
training. There are 285 million evangelical Christians in the world, comprising 13% of Christians
and 36% of Protestants (Hackett and Grim 2011).
1 ICM is representative of an important sector
that attempts to generate religiosity while alleviating poverty.
ICM’s program, called Transform, normally consists of three components—Protestant
Christian theology, values, and character virtues (“V”), health behaviors (“H”), and livelihood
(i.e., self-employment) skills (“L”)—taught over 15 weekly meetings (plus a 16th meeting for a
graduation ceremony). Each meeting lasts 90 minutes, spending 30 minutes per component. ICM’s
leadership believes that the Values curriculum lies firmly in the mainstream of evangelical belief.
Between 2009 and 2017, 194,000 people participated in Transform. The basic structure of the
program, using a set series of classes outside of a Sunday worship service to evangelize, is a common model. For example, over 24 million people in 169 countries have taken the evangelistic
Alpha course since 1977 (Bell 2013), and Samaritan’s Purse has enrolled 11 million children in
about 100 countries in its evangelistic Greatest Journey course since 2010 (Samaritan’s Purse
2017). Like Transform, these are courses of approximately a dozen sessions.
We randomly assigned 320 communities (from which we selected 7,999 households) to
receive the full Transform curriculum (VHL), to receive only the Health and Livelihood
components of the curriculum (HL), to receive only the Christian values component of the
curriculum (V), or to be a no-curriculum control (C). We identify the effect of religiosity by
comparing invited households in VHL communities to invited households in HL communities, and
invited households in V communities to households in C communities that would have been
invited had that community been assigned to be treated.
We measure outcomes approximately six months and 30 months after the training sessions
ended and analyze them in accordance with a pre-analysis plan. At six months, we find that those
who were invited to receive the V curriculum have significantly higher religiosity than those who
did not receive the V curriculum, demonstrating that the treatment had its intended first-stage
effect. Examining downstream economic outcomes while correcting for multiple hypothesis tests
by controlling the false discovery rate (FDR), we find that the V curriculum increased household
income by 9.2%, but had no statistically significant effect on total labor supply, assets,
consumption of a subset of goods, food security, or life satisfaction, and it decreased perceptions
of relative economic status within one’s community by 0.11 points on a 10 point scale.
2 Post-hoc
analysis shows that the income effect is strongly concentrated on the Transform invitee and is not
significant for other household members’ labor income, providing further support that the
estimated income effect is not a Type I error.
Exploratory regressions suggest that the religiosity treatment effect operates by increasing grit
(Duckworth et al. 2007)—specifically, the portion of grit associated with perseverance of effort
(and in particular, agreement with the statements “I am a very hard worker,” “I finish whatever I begin,” and “Setbacks don’t discourage me.”). This mechanism accords with Weber’s conception
of the Protestant work ethic. We find no consistent movement in the other potential mechanisms
that we measured: social capital, locus of control (other than the belief that God is in control, which
increases), optimism, and self-control. Furthermore, post-hoc analysis finds that the HL treatment
had no statistically significant effects on income or perceived relative economic status at six
months.
3 Because the HL treatment includes many of the non-religious aspects of the V
intervention (e.g., meeting in a group over a number of weeks), this null finding suggests that the
six-month V curriculum treatment effect primarily captures the impact of altered religiosity.
By 30 months, there is no longer a statistically significant difference in the intensity of
religiosity between the experimental groups. However, individuals who received the V curriculum
are 3.6 percentage points less likely to identify as Catholic and 2.3 percentage points more likely
to identify as Protestant. To put these changes in context, the control group at 30 months is 70%
Catholic and 21% Protestant.
There is mixed evidence on the effects on downstream economic outcomes. Relative to the
no-curriculum control, those who received only the V curriculum have a significantly higher
perceived relative economic status (0.34 points on a 10-point scale) and marginally significantly
higher consumption (7.5% of the control group mean, FDR q-value = 0.062). Exploration of the
mechanisms responsible for these positive effects finds that V curriculum recipients are more
optimistic, even though they do not have higher grit. On the other hand, we find no statistically
significant effects on primary economic outcomes when combining the VHL versus HL and V
versus control comparisons. This difference in findings is driven by the fact that the HL group
appears better off than the VHL group at 30 months. Relative to the no-curriculum control, the HL
group has significantly higher income and perceived relative economic status (in tests that do not
adjust for multiple comparisons).
Interpreting these results requires an understanding of the context and details of the
intervention. ICM operates in a setting where most people claim to be religious. In the six-month
survey, only 2.4% of those who did not receive the V curriculum and 2.3% of those who did receive
the V curriculum indicate that they are “not religious at all.” Our experiment should therefore be understood as measuring the effects of strengthening pre-existing religiosity or changing the
emphasis of pre-existing religious beliefs, rather than the effects of causing the completely
irreligious to become religious. Arguably, these intensive margin effects are the most relevant
ones, since 84% of the world’s population is religious (Pew Research Center 2015). It is also
important to note that ICM targets the ultra-poor within communities, and the communities in our
study (including those in the no-curriculum control) are chosen by pastors who presumably believe
that they would be able to run a successful program there. Most expansions by religious
organizations into a community are probably based on a belief that the community would be
receptive, so these are an externally relevant type of community. It is possible that the ultra-poor
are more receptive to religious outreach than less impoverished individuals (Chen 2010), so ICM’s
outreach may be more effective than comparable outreach to higher-income populations.
In addition, religiosity is not a singular concept, and its causal impact will likely depend on
many factors. Johnson, Tompkin, and Webb (2008) differentiate between “organic” exposure to
religion over a prolonged period of time (e.g., through one’s upbringing at home) and “intentional”
exposure through participation in a specific program targeting a specific set of individuals. Both
are important channels of religious propagation, and the type of religiosity produced may depend
on the channel. Our study is about intentionally generated religiosity of a specific kind (evangelical
Protestant Christian), and a significant aim of our study is to establish, in the context of a
randomized controlled trial, that intentional exposure to a religious program can generate the
critical first stage: an exogenous change in religiosity.
[...]
Our work also relates to a growing number of papers that use instrumental variables or natural
experiments to study the causal effect of religion on economic outcomes.4 Clingingsmith, Khwaja,
and Kremer (2009) find that winning a lottery for hajj visas changes beliefs, values, and religious
practices. Barro and McCleary (2003) conduct a cross-country analysis of economic growth using
the existence of a state religion, state regulation of religion, adherence shares for the major
religions, and a religious pluralism index as instruments. They find that religious beliefs
(“believing”) increase economic growth, whereas religious service attendance (“belonging”)
decreases growth. Because our study does not induce independent exogenous variation in beliefs
versus behaviors, we cannot add further evidence on this “believing versus belonging” hypothesis.
Gruber (2005) uses local ancestral mix as an instrument and finds that religious participation in
the U.S. (which is almost entirely Christian) increases education, income, and marriage rates and
decreases disability and divorce rates. Gruber and Hungerman (2008) exploit the repeal of U.S.
state laws prohibiting retail activity on Sundays and find that Christian religious participation
decreases drinking and drug use. Bottan and Perez-Truglia (2015) study the decline in Catholic
religious participation caused by clergy scandals and find evidence that religious participation
increases charitable giving
[...]
No comments:
Post a Comment