Religion and Life Outcomes: Looking for Causal Effects

The pervasiveness of US religious belief has seemed to decline in the last 30 years or so, by a variety of measures. Daniel Hungerman investigates \”Religious Institutions and Economic Wellbeing\” in the most recent issue of Future of Children (Spring 2020, pp. 9-28). As a starting point, here\’s a figures from Hungerman showing the fall in US religious belief over time (using data from the General Social Survey). 
The decline in religiosity seems likely to continue, because the shift is especially large among younger age groups. 
Or here\’s a figure showing that the share of charitable giving going to religious organizations was about half back in the 1980s, but it now down to about 30%. 
Why this shift started happening around 1990s isn\’t altogether clear. The sex scandals afflicting the Catholic Church, for example, are not especially prominent in the broad public eye until the early 2000s. There\’s some nondefinitive evidence that increasing levels of education may tend to reduce religious belief. There\’s also a sense that in  many people\’s minds, religion has become more entangled in politics, which has led some people to react by drawing back from religious belief.  
But Hungerman\’s main focus is about the effects of this change. It\’s well-known that religious people often typically report being happier, and are more likely to vote, less likely to use drugs or commit crimes, and so on. But correlation isn\’t causation. Ideally, one might want to do an experiment where some people randomly become committed to religion while an otherwise identical comparison group does not do so, but this particular experiment seems impractical. Indeed, it seems plausible that those who choose to participate in religion may differ in some underlying way from those who do not. So how can a researcher try to find evidence, one way or another, of causal effects? 

I\’ll just say up front that there\’s no perfect way to find a pure causal effect from religion to economic outcomes. Instead, like a lot of social science issues, one instead tries to tackle the question from a variety of angles in specific research studies, and then see what overall pattern starts to emerge. Thus, the array of examples also gives a sense of how the minds of social scientists work. For example, here are some of the examples and approaches discussed by Hungerman:  

Matching people who are equivalent in the non-religious variables we can observe

Dehejia, DeLeire, and Luttmer examine whether religious individuals’ consumption and self-reported wellbeing appear to be relatively less sensitive to income shocks—that is, whether religion helps “insure” people against negative shocks. … [T]he authors use a variety of methods, such as applying a procedure that matches each religious person in a sample to an observationally similar nonreligious person, so that the final data sample contains a similar distribution of observable characteristics across religious and nonreligious individuals. They find that religiosity does indeed insure against negative shocks.

Looking at whether more of your neighbors share your religious tradition

Gruber proposes a creative strategy: using variation in the ethnic composition of one’s community to study the impact of religion. Put simply, an American of Italian ancestry may not make much of a distinction between living in a neighborhood full of Swedish individuals versus a neighborhood full of Polish individuals—except that the latter group, like Italians, are Catholic. If living side-by-side with ethnicities that share your religious tradition makes you more religious, but otherwise doesn’t affect your wellbeing, than we can use ethnic composition to learn about the causal effects of religion.
Gruber finds, again, that religiosity leads to better outcomes for a number of economic indicators.

Peer groups in schools

Especially noteworthy is a study by the economists Jane Fruehwirth, Sriya Iyer, and Anwen Zhang. In an approach similar to Gruber’s, they exploit variation in the religiosity of peers across cohorts within a school to identify how religion influences mental health in a sample of US adolescents. They find that religion plays an important causal role in promoting mental health.

A lottery for attending the Haj

David Clingingsmith, Asim Khwaja, and Michael Kremer. They look at the effects of attending the Hajj—the pilgrimage to Mecca that Muslims are expected to make at least once during their lifetime. To study how attending the Hajj affects people’s values, Clingingsmith, Khwaja, and Kremer use a Pakistani lottery that allocates Hajj visas; they find that participation in the Hajj leads to greater acceptance of female education and employment. More generally, Hajj lottery winners show both increased Islamic observance and greater belief in equality and harmony among all religions.

Distance from Wittenberg

The great social scientist Max Weber famously considered whether a Protestant ethic for work might drive the difference between economic wellbeing in Protestant and Catholic communities. Becker and Woessmann take up this association in several steps. First, they put it to a careful test in historic Prussia, exploiting the fact that Protestantism expanded from its birthplace in Wittenberg (a previously unimportant town) in a pattern akin to concentric circles. Moving away from Wittenberg, you encounter all sorts of terrain and all types of communities—but places farther from Wittenberg are less likely to be Protestant, all else equal. Becker and Woessmann then confirm that distance from Wittenberg appears unrelated to various controls (such as the presence of schools in the 1500s, before the reformation), but centuries later it does predict income and economic circumstance—being closer to Wittenberg (and therefore more Protestant) is better for economic wellbeing. This suggests that the link between GDP and Protestant affiliation is more than a simple association. Does this mean Weber was right? Not quite. The final step of Becker and Woessmann’s study shows that variation in literacy can largely explain the economic gains of Protestantism. It appears that the Protestant emphasis that everyone should be able to read the Bible (and thus be able to read), rather than a “noncognitive” work ethic, can explain why Protestant societies had higher economic productivity.

Interaction of religion with laws about alcohol and gambling

Looking at the United States, Jonathan Gruber and I investigated this by looking at the repeal of “blue laws” that restrict economic activity on a certain day of the week (often Sunday).40 Most recent blue laws are narrow in focus—for example, alcohol can’t be sold at grocery stores before noon on Sundays. But not that long ago, many states had strong blue laws that prohibited most Sunday economic activity. A Supreme Court ruling in 1961 provided a test by which these laws could be repealed, and many were consequently undone. Gruber and I show that when such laws are undone, religiosity declines, and that risky behavior such as heavy drinking increases— but the increases are driven by those who report having been religious before the repeal occurred.  … Religious rules appear to be effective in discouraging heavy drinking and gambling. The results [from another study] often indicate that the most religious individuals are those who are likeliest to substitute: it’s the most religious groups whose religious giving declines when casinos open or when commerce is allowed on Sundays, and it’s the most religious individuals who are likely to start drinking heavily when the legal drinking age changes.

Figuring out causal connections from religion to life and economic outcomes is a challenging research project. But the weight of the evidence Hungerman discusses–only a bit of which I\’ve mentioned here–is that those who end up being exposed to religion do seem to experience measurable benefits, which in a broad sense take the form of bolstering a person\’s confidence and determination in following a path of learning, saving, and work–and avoiding being derailed by overindulgence in counterproductive habits. 

I\’ll also append here the full Table of Contents for this issue of Future of Children. It seems to have an even higher-than-usual proportion of interesting essays, and I may post some additional commentary about other essays in the next week or so. 

The Productivity That Didn\’t Happen

The bad news is that US productivity growth has been slow for the last 15 years, and in facts for 30 of the last 40 years. But at least other high-income nations are doing worse. Emily Moss, Ryan Nunn, and Jay Shambaugh provide a nice readable overview of productivity fact patterns, along with possible causes and solutions, in \”The Slowdown in Productivity Growth and Policies That Can Restore It\” (June 2020, Hamilton Project at Brookings). 
It\’s conventional to divide US productivity growth since 1948 into four periods, summarized in this figure. There\’s the reasonably rapid post-World War II productivity growth from 1948 to 1973, the slowdown from 1973-1995, a 10-year resurgence from 1995-2004, and the return-to-slowdown since then. 
The US situation doesn\’t look good, but in fact, we\’re going better than other high-income countries. The fact that the productivity slowdown encompasses all the high-income countries has an important implication: it suggests that at least some of the most important causes are not specific to US economic policy or indeed to the policies of any one country, but instead must be causes that would apply broadly across all high-income countries. 
Don\’t forget that the numbers on the figures above are annual rates. For example, US labor productivity drops from an annual rate of 3.1% from 1995-2004 to an annual rate of 1.4% from 2004-2018. This annual change accumulates over time.  To get a feeling for the importance of this accumulation,  let\’s say that labor productivity had continued to rise by 3% per year since 2004. The extra growth could have led to gradually higher incomes every year. As a result of labor productivity growing 1.6% per year faster over the last 16 years, the total US economy in 2020 would be 29% larger. 
Given that US GDP will be about $22 billion this year, 29% works out to $6.4 trillion larger. For perspective, $6.4 trillion works out to roughly an extra $20,000 in 2020 for every US citizen, including adults and children. This is not a one-time boost, but a permanent and ongoing rise. For people, for government, for social problems, for the environment–every problem is a little easier to solve when the financial constraints are relaxed. But because productivity growth slowed down in 2004, those resources never came into existence. 
The reasons for these rises and falls in productivity are largely mysterious to economists. The problem is not a lack of hypotheses, but rather a sense that when you offer lots of possible explanations, perhaps you aren\’t all that sure about any of them. For example, if the more recent productivity slowdown had kicked in after the Great Recession, one could come up with a hypothesis related to the Great Recession–but it clearly started well before that. 
One possible explanation is that this is mostly a mismeasurement problem. The argument here is that GDP doesn\’t capture the economic gains of new technology, so the productivity gains don\’t show up in official statistics. There\’s no doubt that the official statistics are imperfect, but but the question here is whether they somehow became more imperfect circa 2004; that is, they captured the rise in productivity growth and the web pretty well for a decade, but then stopped doing so. There isn\’t much evidence to support that belief. Moreover, the decline in US productivity after the 1995-2004 burst has been quite broad across industries. To put it another way, it doesn\’t seem as if productivity is only down in certain harder-to-measure industries. 
In addition, while it would certainly be nice to believe that our standard of living is rising in all sorts of ways not measured by actual dollars, our household bills and mortgage debt and taxes and government borrowing need to be paid with actual money, not just with a nebulous feeling of being better off. 
While the struggle to explain the timing and breadth of the changes in productivity goes on, some of the intellectual energy has instead shifted toward thinking about what might help reverse the pattern, regardless of its underlying cause. I sometimes like to say that there\’s the basic formula for productivity is well-understood: it\’s a combination of human capital, physical capital, and new technology, interacting in an economic environment with incentives for innovation. Here are a few thoughts from Moss, Nunn, and Shambaugh on these issues, with more in the actual report: 
On the issue of human capital, America\’s rise in average education level has slowed down, and the aging population means slower growth in the prime-age labor force. They write: 

For cohorts born from 1876 to 1951, average educational attainment rapidly increased by 0.8 years per decade, with successive generations receiving about two additional years of education relative to their predecessors. The pace of this increase has now slowed: cohorts born from 1951 to 1987 have added only about 0.3 years per decade …

Slower growth in the prime-age labor force tends to coincide with slower growth in productivity, perhaps because of a reduction in available managerial talent (Feyrer 2007, 2011) or the rate of business formation (Karahan, Pugsley, and Şahin 2019). The aging of the workforce can also place downward pressure on productivity growth by making it more difficult to implement new innovations and processes …
When it comes to private investment, firms don\’t seem to be doing a lot more of it. Here\’s a figure showing investment by firms in the specific area of information processing equipment and software. It\’s probably not a coincidence that the sharp rise starting in about 1990 was followed by rising productivity a few years later, and conversely that the drop in 2000 was also followed by lower productivity a few years later. 
One my my hobby-horses on this blog is the need to increase research and development spending. As the figure shows, total R&D as a share of GDP hasn\’t risen for decades. But what is changing is that the federal share of R&D which tends to focus on basic research and thus on potential breakthrough innovations is falling, while the business share of R&D which is more likely to focus on near-term development of products is rising. 
When it comes to increasing productivity, there are a number of other areas that deserve attention, many of which involve trying to strike a better balance: Can we figure out ways to provide inventors with a return that also encourage follow-up inventions by others? Can we figure out ways to encourage competition and limit anticompetitive behavior? Could adjusting rules related to occupational licensing or residential building help labor to become more productive? Could investments in infrastructure for transportation, energy, and communications help labor and industries be more productive?

The only way for the average person in a nation to consume more in the long-run is for that average worker to have higher productivity in the long-run. Yes, for a time it\’s possible to raise taxes on the rich and transfer to others. It\’s possible for the government and firms and people to borrow money for a time and raise consumption in the present, too. Redistribution and borrowing are useful for specific circumstances, and for certain times and places, but by themselves, they can\’t continually raise consumption for the average person. Only rising productivity can do that.