Can we drive development at scale? An interim update on economic growth work
Disclaimer: This is an interim report and the views expressed here do not represent “house views” of our employer Founders Pledge
Introduction
Global health and development is still arguably the most popular EA cause. For example, payouts from the Global Health and Development EA Fund comprise 45 percent of the total amount of money granted from EA Funds. Almost all of this spending supports so-called “randomista”-type development: direct interventions that have strong experimental evidence of effectiveness. This allocation is justified by the claim that these interventions are the most cost-effective way to improve the lives of people in low- and middle-income countries (LMICs).
Earlier this year, John Halstead and Hauke Hillebrandt published an EA Forum post that argued this is likely mistaken. In “Growth and the case against randomista development” they write that past poverty alleviation has overwhelmingly been achieved by economic growth, not direct interventions. They argue that the magnitude of the gains of growth are so large that interventions which can increase growth rates are likely more cost-effective even if there is less evidence of their effectiveness or they have a low chance of success.
The post generated a lot of discussion and commenters raised several important potential objections. These included:
Growth work is not neglected and there are no good marginal funding opportunities
Economic growth does not make people much happier
Economic growth does not help the poorest of the poor
We are clueless about the causes of growth
Implementing better economic policies faces political economy challenges that EA funding cannot overcome
Over the past few months, we have spent between 100 and 150 hours looking deeper into these challenges to try to determine the likelihood of finding concrete, cost-effective funding opportunities to promote economic growth. We conducted a brief literature review, but given the breadth of the subject matter and the uncertainty of the research question we relied heavily on conversations with experts. After about 30 such interviews, we’ve decided to stop looking for concrete funding opportunities for now. So far we haven’t found any growth-focused policies or programs which experts agree would be highly-valuable to support. We think good opportunities probably exist, but identifying them will require thorough evaluations of potential funding opportunities. Since evaluating policy-focused interventions requires considerable investment from both us and representatives from the organisation under investigation, we’re deprioritizing this project for now. We’re posting this wrap-up to share what we’ve learned, get feedback on our current conclusions, and stimulate further discussion on this important topic.
Is growth work neglected?
A key question we looked into was whether or not research and advocacy into economic growth is relatively neglected compared to direct, randomista-style interventions. A complication here is that almost all development programs, including randomista programs, affect growth at some level. This makes it difficult to separate out funding for the policy-focused work we’re interested in. For example, at first glance the International Growth Centre seems like it would be a relevant funding opportunity with a large budget. After further investigation, though, our impression is that the IGC’s work is more in the randomista school.[1]
If one were to simply add up all the money multilateral organisations like the World Bank, Official Development Assistance (ODA) agencies, and large NGOs spend on work they classify as “economic growth”, it would be a large amount—much more than $100 million per year. There are also several organisations like the UN Development Programme, whose budget is $6 billion per year, dedicated to supporting policymaking in LMICs. However, much less money is spent evaluating the effectiveness of this spending to find out what works well and is cost-effective.[2]
The two most important funders in this space are the World Bank and the UK’s Department for International Development (DFID).[3] The World Bank’s Development Economics Department (DEC) manages an annual budget of roughly $100M. The World Bank also makes $30 billion in loans and spends $200 million per year on “analytical and advisory products” to client countries in a typical year.[4] DFID has a joint program with the Economic and Social Research Council to fund growth research. That program has a budget of £25 million ($31 million) per year and has so far “funded a portfolio of 43 grants across four research calls.” Together, these budgets make a conservative lower bound for total funding in this space.
While $130M per year is a lot of money, it’s negligible compared to the value of sparking a growth acceleration or avoiding a growth deceleration. Funding from multilaterals or ODA agencies may also neglect impactful funding opportunities due to political or institutional considerations.[5] We found fewer private funders of growth research than we expected. There are few clear examples of philanthropists interested in this kind of work (the Ford Foundation being one exception). Jim Cust, a World Bank economist, suggested this is because growth interventions are much less “attributable” than direct interventions and that “much of the development industry is risk-averse.”[6] This means interventions which have an impact, but where the influence of a given funder is difficult to see, are likely to be neglected.
While we didn’t conduct a systematic search, we did lightly engage with a few potential funding opportunities to gauge what kind of work additional funding could support. We found more active organisations than we expected. An incomplete list of organisations we identified or spoke with includes:
We also expect that there are some “single-issue” think tanks or organisations working in low- and middle-income countries working in this space, but which we did not come across.
These organisations differ greatly in their work, focus, theory of change, and size. We haven’t evaluated the cost-effectiveness of any specific projects or programs these organisations run. We’re not suggesting that these are the most cost-effective opportunities in this space or that donations to any of them are sure to be more cost-effective than donations to GiveWell’s top charities. Nevertheless, we were struck by the number of potential funding opportunities we found.
Does growth make people happy?
Growth and Subjective Well-being
Income, and hence economic growth, isn’t good for its own sake. We care about it to the extent it brings about other ends that we care about intrinsically. A plausible candidate for something we care about intrinsically is well-being, so it could be important to investigate the extent to which economic growth improves well-being. There is a rich body of literature on subjective well-being, which typically involves asking people how happy they are or overall how satisfied they are with their lives.
Log GDP per capita correlates well with self-reported happiness and life satisfaction across countries[7] and log income correlates well with happiness and life satisfaction across individuals,[8] which suggests we should measure impact in logs, rather than dollars. Surprisingly, however, log GDP per capita doesn’t correlate much with positive and negative affect once we control for other factors such as social support[9] and freedom to make life choices.[10],[11] This doesn’t necessarily imply that increasing GDP per capita wouldn’t improve positive and negative affect though – increased GDP per capita could improve affect via its effects on variables such as social support and freedom to make life choices.
In the World Happiness Report 2020, social support, freedom to make life choices, generosity and perceptions of corruption all had greater regression coefficients than log GDP per capita (in absolute value) in regressions of Cantril Ladder (roughly, life satisfaction), positive affect and negative affect.[12] This suggests that these variables have a greater effect on subjective well-being than income. As a result, we should bear these variables in mind in evaluating and prioritising growth interventions.
The report also found that 32% of the difference in happiness between top 10 and bottom 10 happiest countries is due to log GDP per capita; next largest contributors are social support (27%) and healthy life expectancy (21%).
Life satisfaction and GDP per capita data from the previous World Happiness Report and the World Bank, gathered by Our World in Data, show a strong relationship between life satisfaction and GDP per capita:[13]
It doesn’t follow automatically though that economic growth (i.e. increasing GDP per capita) improves life satisfaction: it could be the case that while life satisfaction tends to be higher in countries with higher GDP per capita, increasing GDP per capita does not cause life satisfaction to increase. The economist Richard Easterlin has famously argued that economic growth does not increase happiness.[14] There now seems to be a consensus that economic growth increases happiness in the short-run but there is little consensus over the long-run effects of growth on happiness. For example, Stevenson and Wolfers (2008) argue that growth does have long-run effects on happiness but Easterlin (2016) maintains that growth only increases happiness in the short-run (less than 10 years).[15] Charles Kenny (Center for Global Development) confirmed that the long-run effects of growth on happiness are unclear.[16]
Rough cost-effectiveness analysis informed by the subjective well-being literature
Lant Pritchett and later Hillebrandt and Halstead have made very rough historical cost-effectiveness estimates of attempts to influence recent growth accelerations. These aren’t detailed, accurate estimates but are intended to make the case for the plausibility of the high cost-effectiveness of trying to accelerate economic growth, to stimulate further research.
Pritchett considers the impact of the Ford Foundation in funding the India Council on Research on International Economic Relations (ICRIER), which was possibly highly influential in India’s recent economic growth episodes.[17] Hillebrandt and Halstead extend this example by considering similar, hypothetical attempts to influence Vietnam’s economic growth acceleration between 1989 and 2010 (the median growth episode identified in Pritchett et al. 2016). They compare this to direct interventions such as the Graduation Approach and funding Malaria Consortium. Both estimates measure the goodness of growth episodes in net present value, which fails to account for the diminishing returns to income of subjective well-being.
Measured this way, growth accelerations in countries that are already rich will look more valuable than accelerations in poor countries. But if well-being is a linear function of log GDP per capita, then equally proportionate increases in income increase well-being equally, no matter the initial value of GDP per capita.[18] Here, we build on aforementioned estimates by measuring impact in income doublings (or changes in log2 income).[19] Measuring impact in this way paints a somewhat less optimistic picture of accelerating growth. The model is here with the main results displayed below.
If per capita income grows at rate over a year, then the number of income doublings per person due to the economic growth in that economy is .[20] In the case of a growth acceleration, countries grow at an additional rate of growth g relative to counterfactual growth rate (i.e. they grow at rate rather than ), where is estimated by Pritchett et al. 2016 for various growth episodes. Then the number of income doublings per person due to economic growth acceleration is:
g′is typically small (on the order of a few percent), so we can fairly safely ignore it and approximate per capita income doublings as . (This is an overestimate by a usually negligible degree.) We then approximate total income doublings by multiplying the per capita income doublings by population size.
Pritchett outlines two scenarios, which Hillebrandt and Halstead adopt: one optimistic and one pessimistic, with respect to accelerating economic growth in India. In both cases, we suppose that the ICRIER received $36 million in funding shortly before it was created. In the optimistic case, this funding increased the probability of the ICRIER being created by 50 percentage points, and the ICRIER in turn increased the probability of India adopting growth accelerating policies by 10 percentage points.[21] In the pessimistic case, these numbers are 10 percentage points and 1 percentage point, respectively.
Hillebrandt and Halstead find that in the pessimistic scenario, and considering Vietnam’s recent economic growth episode instead of India’s, trying to accelerate economic growth is roughly on a par with the Malaria Consortium in cost-effectiveness. In our model, this scenario favours the Malaria Consortium by a factor of about 20. Hillebrandt and Halstead’s optimistic case (with respect to accelerating economic growth) finds trying to accelerate India’s growth to be about 3,000x more cost-effective than the Graduation Approach. In our model, this falls to about 350x. This tentatively suggests that, historically, trying to accelerate growth was about an order of magnitude less cost-effective than previously estimated but that it could still be better than direct interventions such as the Graduation Approach and Malaria Consortium.
The usual caveats apply to these (very rough!) estimates. Don’t take them literally. Since our model is much more speculative and less sophisticated than GiveWell’s detailed cost-effectiveness analyses, we shouldn’t place as much weight on it. Even if the Ford Foundation’s funding of ICRIER was highly cost-effective, it doesn’t follow that there will necessarily be similarly cost-effective opportunities available today. We should also be wary of hindsight bias in choosing parameters of the model. Ex post, it might seem obvious that the Ford Foundation’s funding of ICRIER drastically increased its chances of being created and in turn that the ICRIER significantly increased the chances of India adopting growth-accelerating policies. But presumably, that sequence of events was less predictable ex ante.
Measuring impact in income doublings rather than net present value reduces the lead of growth interventions over randomista interventions but maybe not drastically. To determine whether there are growth interventions that are more cost-effective than randomista interventions, a more detailed analysis of specific growth interventions or case studies to come to more robust cost-effectiveness estimates would be helpful. We haven’t tried to establish here that increasing economic growth is highly cost-effective – this is intended more as a plausibility argument to encourage more in-depth research.
| Ethiopia graduation approach | low | medium | high |
| Additional income per $ | 5.40 | 3.50 | 1.60 |
| Spending on graduation approach (current US$) | 36,000,000 | 36,000,000 | 36,000,000 |
| Additional income | 194,400,000 | 126,000,000 | 57,600,000 |
| Population | 112,078,730 | 112,078,730 | 112,078,730 |
| Additional GDP per capita | 1.73 | 1.12 | 0.51 |
| Ethiopia GDP per capita (current US$) | 857.5 | 857.5 | 857.5 |
| Change in log_2(GDP per capita) | 0.003 | 0.002 | 0.001 |
| Total income doublings | 326,737 | 211,849 | 96,880 |
| Cost-effectiveness (income doubling/$) | 0.009 | 0.006 | 0.003 |
| Growth episode | India 1993-2010 | ||
| Approx total income doublings of growth episode | 414,442,180 | 546,708,462 | 678,974,743 |
| Probability that $36m leads to creation of think tank | 10% | 30% | 50% |
| Probability that think tank affects growth episode | 1% | 5.5% | 10% |
| Expected number of income doublings | 414442.1802 | 9020689.617 | 33948737.16 |
| Cost ($) | 36000000 | 36000000 | 36000000 |
| Cost-effectiveness (income doubling/$) | 0.012 | 0.251 | 0.943 |
| Ratio over Ethiopia Graduation Approach | 1.27 | 43 | 350 |
| Ratio over Malaria Consortium | 0.40 | 13 | 111 |
| Growth episode | Vietnam 1989-2010 | ||
| Approx total income doublings of growth episode | 46,689,012 | 63,339,258 | 79,989,504 |
| Probability that $36m leads to creation of think tank | 10% | 30% | 50% |
| Probability that think tank affects growth episode | 1% | 5.5% | 10% |
| Expected number of income doublings | 46689.01222 | 1045097.756 | 3999475.182 |
| Cost ($) | 36000000 | 36000000 | 36000000 |
| Cost-effectiveness (income doubling/$) | 0.001 | 0.029 | 0.111 |
| Ratio over Ethiopia Graduation Approach | 0.14 | 4.9 | 41 |
| Ratio over Malaria Consortium | 0.05 | 1.6 | 13 |
Does growth reduce poverty and help the poorest of the poor?
A common objection to focusing on economic growth and a possible issue with the above analysis, which only uses the per capita GDP growth data, is that growth may not benefit those living in extreme poverty. We take this concern seriously, and the large literature on “inclusive” or “equitable” growth suggests many economists and policymakers do so as well. However, we did not spend much time looking into this question. This is mainly because we have yet to see a strong counterargument to Pritchett’s claim that this isn’t a major concern due to the strong negative correlation between national median income and national poverty rate.
Do we know anything about what causes growth?
A fourth objection to growth research or advocacy is that it’s a waste of time because the causes of economic growth are unknown and unknowable. “The Empirics of Growth” by Kevin Grier gives a high-level overview of our knowledge about growth. He highlights the fragility of most findings that attempt to link specific policy changes to growth rates. Even the influence of broad measures like education levels or investment rates is controversial, while “the policy variables that seem most robustly related to growth are sound macroeconomic policies (mainly stable and reasonably low inflation), openness to trade, institutional quality (i.e., little government corruption), and financial development.”
Grier summarizes 2004 work by Hausman, Pritchett, and Rodrik:
Ricardo Hausmann, Lant Pritchett, and Dani Rodrik (2004) [...] studied eighty-three cases in which a country rapidly increased its growth rate and sustained the increase for at least eight years. Their most statistically significant results are that a financial liberalization raises the probability of a growth increase by around 7 percent, and that a political regime change toward autocracy (from democracy or less-strict autocracy) raises the probability of increased growth by almost 11 percent. Most growth increases (which they call “growth accelerations”) are unpredictable, however, and as they put it, “the vast majority of growth accelerations are unrelated to standard determinants such as political change and economic reform, and most instances of economic reform do not produce growth accelerations.”
It’s important to note that this finding relates to what proportion of reforms spark growth accelerations. Because growth accelerations are so valuable, such reforms may still have high expected value even if they usually fail to effect a change (so long as the downside risk is also low).
One of the major reasons it is difficult to know what causes growth is that all factors are endogenous—that is, most variables that plausibly affect growth rates tend to go up or down at the same time and there is almost always reverse causation for the variables that are correlated with growth. Cross-country growth regressions can identify these covariates but not causal relationships. As a result, economists have tended to model growth as exogenous—something that either does or does not happen randomly—and then use empirical data to see what goes along with growth.
Given the importance of growth, we find it surprising that so much growth research uses models which cannot help us identify its causes. Doug Gollin (University of Oxford, Structural Transformation and Economic Growth) told us that “It’s not that economists believe that growth happens exogenously; it’s just that for some questions, the sources of growth are less interesting than the consequences of growth. For these questions, exogenous growth models are sufficient. And these models are also nicely transparent. There is of course also an important literature on endogenous growth. In these models, the growth rate is determined within the model – but the sources of growth are essentially baked into the models, so the models don’t add a lot to our understanding.”[22] This seems like clear evidence that there is potential improve these models, which help researchers describe the effects of growth in a country but don’t help inform policy prescriptions to stimulate growth.
Perhaps the most well-known attempt to articulate the conditions for growth was the Washington Consensus, a set of policies that the IMF, the World Bank, and the U.S. Treasury prescribed for debt-burdened developing economies throughout the 1980s and 1990s. These reforms were broadly oriented towards market-based approaches. They included things like fiscal policy discipline, tax reform, market exchange rates, trade liberalization, and privatization.[23] Today the Washington Consensus has a pretty bad reputation.[24] That said, recent work by Easterly, Grier and Grier, and Chari, Henry, and Reyes suggest that policy reforms in line with Washington Consensus prescriptions generally show positive effects. This suggests that we are not completely clueless. Some clearly bad policies such as extreme inflation, negative real interest rates, and trade suppression do tend to reduce growth.[25]
Nevertheless today it is generally recognized that policy advice should consider structural and political differences between countries.[26] To this end, growth diagnostics, an approach introduced in the mid-2000s,[27] has gained some footing among multilateral organisations and consultants.[28] The growth diagnostics method assumes that countries face multiple different constraints on their growth rate, but that the constraints which are most influential or “binding” vary among countries. Researchers may be able to use growth diagnostic tools to identify “the most binding constraints on economic activity, and hence the set of policies that, once targeted on these constraints at anypoint in time, is likely to provide the biggest bang for the reform buck.”[29]
While this sounds promising, our impression is that growth diagnostics has largely failed to take off due to a lack of academic incentives to use the method amid concerns about causal identification. The development economics field has prioritized research methods with stronger claims to causal identification: chiefly randomized controlled trials, but also quasi-experimental methods like instrumental variables estimation and regression discontinuity designs. Proposals which use these methods are more likely to be funded and papers which use them are more likely to be published. Development economists may also be wary of studying growth at the macro scale as this research hasn’t reliably generated useful findings in the past. Angus Deaton writes that in academia and policy circles, “Past development practice is seen as a succession of fads, with one supposed magic bullet replacing another—from planning to infrastructure to human capital to structural adjustment to health and social capital to the environment and back to infrastructure—a process that seems not to be guided by progressive learning.”[30] However, we understand that the growth diagnostics approach has gained some traction with policy makers outside of academia.[31]
Our overall impression is that making confident, country-specific recommendations to affect growth rates is very difficult. It is hard to determine confidently which specific constraints are binding in a given situation. Country policies may vary along many dimensions. And so many factors also affect growth rates that it’s difficult or impossible to detect the effect of a particular policy change. Yet it’s important to also consider Pritchett’s argument that funding policy research or advocacy probably looks quite good in expected value terms. Although “most instances of economic reform do not produce growth accelerations,” the average growth acceleration increases GDP per capita so much that projects which have only a small chance of positively influencing policy look good in back-of-the-envelope calculations.[32]
Could we even get growth-promoting policies implemented?
Another argument that was raised against this work was that, even if we did know what policies to implement to increase growth rates, we shouldn’t or wouldn’t be able to influence policy change in low- and middle-income countries. We think there are important practical and moral questions to consider here. On the practical side, one approach we find compelling is the idea that researchers or advocates could do important preparatory studies so that they have evidence and recommendations ready to go when opportunities for change arise. For example, on the 80,000 Hours podcast Rachel Glennerster suggested that researchers are currently playing such a role in Ethiopia. The country “is going through a tremendous reform, and we really ought to be focusing attention, and helping Ethiopia in that transition. Tremendous potential, because they’re absolutely fundamentally changing policy there in ways that could be really beneficial to the poor. So jump on those opportunities, but you can’t really make them happen. It’s something that the developing country themselves has to decide to do, then help them as much as you can.”
Uncertainties
While we feel we’ve made some progress on these questions, we recognize that we are still highly uncertain about what specific activities would be cost-effective to fund. Some of the major unresolved questions we have are:
Research
What kind of research would additional funding support? What are the major unanswered questions in this space researchers would tackle?
How valuable is research at the current margin? The average value of research is high, but it’s unclear how valuable research in this area is at the margin[33] or whether research priorities can be altered[34]
Policy change
What sort of policies are most likely to affect a country’s growth rate? Growth seems really important, but concrete examples of things to work on usually seem quite small and unlikely to stimulate the large-scale change that makes growth so valuable
Should we prioritize countries and tailor policies to their specific context, or should we prioritize policies and look for countries that are open to policy change? Some experts thought the best approach would be to identify priority countries with large populations or people in poverty.[35] A focused approach would allow a deeper understanding of the local context and influential actors, which would help us better identify good funding opportunities.
Growth episodes
What’s the “hit rate” for growth episodes? It’s easy to imagine that funding growth research in Vietnam would have had very high expected value if it made the country’s growth acceleration more likely. However, it’s equally easy to imagine that lots of money could be fruitlessly spent trying to stimulate growth in the DRC, or other countries where local politics mean that policies will never be changed or the state does not have capacity to enforce policy changes. What is the Vietnam of the 2020s? Can we identify ex ante where funding is most likely to be effective?
Nathan Nunn suggests industrial policy benefits are near zero-sum.[36] That is, some countries will develop (driven by, e.g., demand for certain exports), and a growth acceleration in one country comes at the expense of growth in another. To what extent is this true?
Why we’re not continuing at this time
We’re not continuing this project because we think identifying funding opportunities would be too costly for us and for the organisations under consideration. We do think good funding opportunities likely exist in this space. One concern some experts expressed is that growth research output is not highly elastic to funding because the supply of researchers is fixed in the short-term and there are other large funders active in the space.[37] While the magnitude of growth accelerations means that some growth work is valuable, it does not follow automatically that marginal growth work is valuable. However, our sense is that if Founders Pledge members (or a different funder) were unusually willing to fund research with low attributability they could have a big impact.
We have also identified a range of think tanks that work on these issues and think it’s likely that there are effective think tanks working in developing countries who could use more funding.[38]
It remains the case that the expected value of information of further research is high. We could see this work being much less or much more cost-effective than global health interventions, so reducing uncertainty would be valuable. However, we didn’t find much consensus among the experts we spoke to regarding the best ways to learn more about how to have impact in this space.
In addition to these questions about how valuable more work in this vein would be for funders, our past work evaluating policy advocacy organisations suggests that this project would be burdensome for the organisations under investigation. Establishing a causal link between policy advocacy and policy change is challenging. Evaluating policy advocacy organisations requires us to build a detailed understanding of the organisation’s outputs and the policy ecosystem in which they operate to judge their chance of success. Ideally, we would evaluate this by considering case studies of previous work carried out by the organisation to determine how strong their track record of influencing policy is. Making such judgements requires lots of information from the organisation and relevant experts and referees, which is time-consuming to produce. We’re wary of doing harm by burdening these organisations with a costly evaluation process.
Paying these costs can be worthwhile if we expect to move a lot of money to the top organisations as a result of the evaluation. However, at this time we are very uncertain about how much money we expect to move in this space. We’re uncertain about our members’ donation preferences and the timeline of their liquidity events. It seems very likely that we would directly move less than a few million dollars, and probably much less, to any recommendations in this space. Without stronger assurances of funding, it’s difficult to justify further research when the costs are so high and the expected benefits are fairly low.
Proposed next steps for future work
Future work could carry out in depth evaluations of promising funding opportunities and try to compare such funding opportunities to direct global health and development interventions. We think that evaluating specific funding opportunities would be more valuable than general research aiming to prioritise interventions within the space of economic growth or more broadly aiming to prioritise between economic growth and other areas within global health and development. This is because we’ve found these latter prioritisation problems to be intractable, with little expert consensus. Direct cost-effectiveness estimates would likely provide more useful information.
Acknowledgements
We are extremely grateful to the experts with whom we spoke, some of whom are cited here and some of whom preferred to remain anonymous, for taking the time to help us with this research.
[1] We got this impression from multiple experts. One might also look at a sample of projects they have funded. Currently, on the first page of the projects funded under their “State” research theme are randomista-type projects focused on providing information to jobseekers, incentivizing bureaucrats, and increasing female labour force participation (see here).
[2] This is the impression we got from talking to experts, but seems like a safe assumption (high confidence, 90%).
[3] DFID is now in the process of merging with the Foreign and Commonwealth Office. This may change their funding priorities, but we don’t expect it to change our conclusions here.
[5] Julian Jamison (University of Exeter) also raised this concern. There are “internal and external pressures—for example, to reward the allies or punish the enemies of powerful shareholders, or to “push money out the door” to achieve professional prestige and budget maximization goals—that undermine the credibility of [the World Bank’s] policy conditionality” (Knack et al., p. 4).
[6] Conversation with Jim Cust, 29 April 2020.
[7] See e.g. Table 2.1 World Happiness Report 2020.
[9] “Social support is the national average of the binary responses (0=no,1=yes) to the Gallup World Poll (GWP) question, “If you were in trouble, do you have relatives or friends you can count on to help you whenever you need them, or not?”
[10] “Freedom to make life choices is the national average of binary responses to the GWP question, “Are you satisfied or dissatisfied with your freedom to choose what you do with your life?”
[11] Table 2.1 World Happiness Report 2020.
[12] Table 2.1 World Happiness Report 2020.
[13] Happiness and Life Satisfaction, Our World in Data
[14] Easterlin 1974, Easterlin 1995
[15] “examining the relationship between changes in subjective well-being and income over time within countries we find economic growth associated with rising happiness.” Stevenson and Wolfers, 2008.
“New data confirm that for countries worldwide long-term trends in happiness and real GDP per capita are not significantly positively related.” Easterlin, 2016.
[16] Conversation with Charles Kenny, 11 May 2020.
[17] “There is a narrative in which Ford Foundation, a global philanthropy provides some millions (or tens of millions) of dollars of funding that play some role in creating a think tank that itself then plays some role in providing the conditions in which good policy choices are made that then results in the creation of 3.6 trillion in additional output of Indians.” Pritchett 2018, p. 21.
[18] If per capita income is y, average life satisfaction is ls0 = a + b log y, and income increases by a factor of p, then life satisfaction increases by ls1 – ls0 = (a + b log py) – (a + b log y) = b log py/y = b log p, which is independent of y.
[19] Note that Layard et al. 2008 suggests happiness and life satisfaction diminish somewhat faster than logarithmically. Working with logs makes the maths more straightforward, but further research to test the sensitivity of the analysis to this assumption could be valuable.
[20] If per capita income is y, then the number of income doublings is log2 (1+g)y – log2 y = log2 (1+g)y/y = log2 (1+g).
[21] “Optimistically, suppose this gift increased by 50 percent the chance ICRIER was created and became an effective think tank (perhaps other funding could have come along, perhaps not) and suppose the existence and actions of this think tank increased by 10 percent the odds 22 India adopted growth accelerating policies (my read of the situation is that it was higher).” Pritchett 2018, pp. 21-22.
[22] Personal communication from Doug Gollin, 8 October 2020.
[23] See, e.g., the Wikipedia article on the Washington Consensus.
[24] “While the lessons drawn by proponents and skeptics differ, it is fair to say that nobody really believes in the Washington Consensus anymore. The question now is not whether the Washington Consensus is dead or alive; it is what will replace it” (Rodrik 2006, p. 974).
[25] “(1) policy outcomes worldwide have improved a lot since the 1990s, (2) improvements in policy outcomes and improvements in growth across countries are correlated with each other (3) growth has been good after reform in Africa and Latin America, in contrast to the “lost decades” of the 80s and 90s” (Easterly 2019, p. 1).
[26] “New macro development takes structural differences between economies more seriously.” Conversation with Doug Gollin (University of Oxford, Structural Transformation and Economic Growth), 12 May 2020.
[27] “Policies that work wonders in some places may have weak, unintended, or negative effects in others. We argue in this paper that this calls for an approach to reform that is much more contingent on the economic environment, but one that also avoids an “anything goes” attitude of nihilism” (Hausmann, Rodrik, and Velasco 2005, p. 1).
[28] This impression is the result of our expert consultations, and is weakly held (60% confidence). There is some evidence that growth diagnostics are applied by some academics; for example, Harvard’s Growth Lab has produced 24 growth diagnostic studies. However, our research did not turn up a review of growth diagnostics that evaluated its success, and no experts were aware of any especially influential growth diagnostic studies.
[29] Hausmann, Rodrik, & Velasco, p. 2
[31] Conversation with Ranil Dissanayake (Center for Global Development), 12 May 2020.
[32] Even after adjusting for the fact that income gains are more valuable for people whose incomes start out low, our rough calculations suggest that if a $5M investment in policy advocacy has just a 0.5%, or 1 in 200, chance of sparking a growth acceleration like Vietnam’s (the median acceleration in Haussman, Pritchett, and Rodrik’s sample), it would be more cost-effective than donating to GiveDirectly.
[33] Julian Jamison (University of Exeter) raised this concern, conversation 22 May 2020.
[34] “Inducing a scientist to change their direction by a small amount–to work on marginallydifferent topics–requires a substantial amount of funding in expectation. The switching costs of science are large” (Myers 2019, abstract)
[35] Conversation with Jim Cust (World Bank), 29 May 2020.
[36] “Although each country can gain through this strategy, this is not true for developing countries as a whole. That is, every country can’t be a South Korean success story.This is because much of the gains from industrial policy come at the expense of other countries.Much of the benefits of this strategy are zero-sum. This raises the question of whether this is a successful strategy for economic development. The answer appears to be a clear no, at least not for most countries. However, the fact that these benefits are zero-sum provides insight into what can be done to help developing countries” (Nunn 2019, p. 7).
[37] Conversations with Charles Kenny, 11 May 2020, and Kevin Bryan (University of Toronto), 22 May 2020.
[38] Ranil Dissanayake (Center for Global Development) suggested that this is the case. Conversation with Ranil Dissanayake, 12 May 2020.
- Big List of Cause Candidates by (25 Dec 2020 16:34 UTC; 277 points)
- Does Economic Growth Meaningfully Improve Well-being? An Optimistic Re-Analysis of Easterlin’s Research: Founders Pledge by (27 Sep 2022 16:28 UTC; 258 points)
- Intervention Report: Charter Cities by (12 Jun 2021 13:54 UTC; 126 points)
- Visualizing the development gap by (7 Dec 2022 12:57 UTC; 106 points)
- Further thoughts on charter cities and effective altruism by (20 Jul 2021 19:56 UTC; 78 points)
- Hits-based development: funding developing-country economists by (1 Jan 2022 0:28 UTC; 76 points)
- CEA’s 2020 Annual Review by (10 Dec 2020 23:45 UTC; 69 points)
- EA Updates for November by (2 Nov 2020 12:38 UTC; 36 points)
- What are some (potentially) effective policy advocacy organizations in global health and development? by (28 Apr 2023 15:48 UTC; 16 points)
- 's comment on Let’s Fund Living review: 2020 update by (13 Feb 2021 12:45 UTC; 5 points)
- 課題候補のビッグリスト by (20 Aug 2023 14:59 UTC; 2 points)
Thanks a lot for your research and writeup! Really nice to see follow-up work on this topic.
A few thoughts:
Is growth work neglected? I’m not sure if that’s the right question to ask. After all, “micro” development (direct service delivery) type work isn’t neglected—tons of money go into it each year—but most had no good evidence behind it, which motivated the founding of GiveWell, IDinsight etc. So perhaps whether “effective” work in growth is neglected is the more relevant question. Though I agree with you that it may be hard to assess the field as a whole compared to individual orgs.
The orgs you listed:
As you said, some (like IGC) focus more on “randomista” type work (I think this applies to Y-RISE too though they care more about effects at scale). I’m guessing there are more orgs focused on the more “macro” aspect of growth, e.g. growth diagnostics.
ODI’s fellowship program is a really interesting model, but I’m not sure how effective it is or how much they measure their impact. I’ve met a few former fellows who after finishing an undergrad econ degree went to work in a ministry in a LMIC for some time, and they told me it wasn’t clear they had much impact. I suspect ODI may want to place more experienced people—now they say they only select masters/PhD graduates, but from what I heard they pay very little, so perhaps that’s a constraint on impact too. It’s a really interesting and high-potential mode but I suspect it can be greatly improved. (IDinsight where I currently work has a similar approach of having embedded learning partnerships with LMIC governments, though as one can expect there are a lot of challenges in working with and influencing them. IDinsight for now focuses more on “randomista”/micro topics like health, education, cash transfers, but topics like tax administration and state capacity are on our minds too.)
Relatedly, perhaps an impactful thing is to fund scholarships for bureaucrats in LMICs to study in top policy schools, e.g. Harvard’s MPA-ID. I heard Latin America (e.g. Mexico, Peru) did a lot of this and it has impacted how governments work there, but don’t know much; I also know some Indian IAS officers have done this.
I am not sure if it would necessarily be that much work evaluating the potential impact, or just track record, of an org. One can sometimes establish a credible causal link using case studies. E.g. Open Phil cited a few impressive achievements of CGD and attributed certain results to orgs in their history of philanthropy studies. In fact, I was hoping for some kind of analysis like that for the growth-promoting (or research) orgs. But of course you (authors) would know better than me about your situation!
BTW, have you checked out Nick Bloom’s work on management practice? He shows it’s a significant constraint on productivity in LMICs (of course, maybe not as fundamental as institutions/politics, but could still be an important one). This interview with him is interesting: https://conversationswithtyler.com/episodes/nicholas-bloom/
Thanks for this great comment! I agree with you on neglectedness, I think the field is so broad that by looking at high-level funding we’re probably counting a lot of stuff that isn’t relevant directly to the question of “would there by impactful work that’s currently unfunded?”, which is waht we actually care about.
Agree also our list of potential orgs working in the space is a bit random and probably misses some good, relevant funding opportunities. Thanks for the info about ODI and IDinsight, too.
My concern with wading into specific evaluations is less about establishing a credible/plausible causal link, and more about collecting enough data to build a proper counterfactual. My impression is that substantial policy changes often involve many different organisations, departments, and people, and it’s hard to work out whose presence made a difference. Our decision to stop short at this time was based heavily on our colleagues’ evaluations for climate organisations, which required a huge investment to confidently work out whose impact was truly counterfactual.
In any case, I’d love to speak more about your experience in the field if we take this work further—if you’re interested in that, please feel free to DM me so we can keep in touch.
I see. Let me know if I’m understanding this correctly: Founders Pledge aims to have cost-effectiveness estimate numbers, which involves a lot of work especially for topics like growth and climate change, whereas Open Phil takes a more qualitative approach for such topics with higher uncertainty. (If so, I am also curious about the philosophy behind your approach—I’m really uncertain which one works better, and that’s a bigger conversation.)
Re topics to look into, I second Michael’s suggestions: labor markets, firms, and monetary policy in developing countries. There’s also: trade, infrastructure, industrial policy, legal system, institutions etc. (Nick Bloom whom I mentioned earlier had the hypothesis that improving management practice in LMICs could be pretty impactful, and that requires a type of education/training not commonly discussed in LMICs.)
One thing tangentially related is Emergent Ventures India. (They don’t have a formal website—all updates seem to be posted on the Marginal Revolution blog.) It’s not growth-specific but rather just for innovative ideas that improve welfare. They don’t have any rigorous analysis (so I’m not sure whether it will fly with EAs) but the projects look cool and it could be a high-potential model (if expanded to Africa etc.).
Happy to keep in touch—will shoot you a DM!
Another example is the continential free trade zone currently being set up in Africa.
This framing is weird. Obviously these factors have a positive causal effect on growth. But why would you expect a silver bullet? Conditions change over time, so the constraints on growth will change as well.
Some thoughts regarding your uncertainties:
One answer: labor markets, firms, and monetary policy in developing countries.
I think one of the main benefits would be the collection of new datasets, which would allow us to identify the most important problems (and figure out how to solve them).
One answer, building on the Washington Consensus stuff you cited: having an economist “in the room” to prevent the president from implementing a policy that would cause hyperinflation.
Fixed.
I think that’s a bit too pessimistic! Founders Pledge has made some progress on this (link goes to pdf) and I think we can do pretty well by taking a kind of journalistic approach. For example, we can speak to charities, experts, and government officials and see if the charity’s claims about who they spoke to and when are true, if the timelines match up, and if it seems like the government would have made changes anway. Check out pp. 8-10 of the linked doc.
I do recognize that this is much more difficult than looking at the results of an RCT. We’ll never be as sure that the effect is causal and it takes a lot of time from both us and the organisation we’re looking at. These costs are the main reason we’re not continuing our growth work at this time.
Good post. I have been following worm wars, the case against randomistas, etc. At the risk of being blunt(and as someone with personal ties to randomista), I think it seems pretty certain that growth in almost any form is not what EAs should be focusing on in terms of actual research. So I disagree with the claim that the method of growth is a high impact space to evaluate especially when we haven’t settled whether growth in general is high or positive impact.
The long term effects (and by this I don’t mean if people will be happy ten years after growth occurs) are highly uncertain, and honestly to the best of my intuition, negative. Given the utter lack of any sort of unifying government on this planet, I think we have enough players as is. The topic is obviously a lot more nuanced than that, but I think it’s suffice to say that no one is about to come up with an airtight argument for how dev will make the world better in 100 years. Like many others have pointed out, it continues to surprise me how much we focus on direct or semi direct impacts in EA when most of us have accepted long termism.
The best piece of evidence by far imo, and it’s not even a very good one, is pritchetts claim on income/poverty negative correlation. And honestly, unless someone can completely dismantle that claim, its not difficult to see x-risk as a more effective anti-poverty measure, given the general upward trajectory of our planet. My epistemic status on that claim isn’t super high, but still.
That being said, empirical poverty research presents a very good recruiting tool for finding people who value the EA framework but haven’t had their third eye opened. I wonder if this could bite us in the butt at some point, but I don’t think EA has much of a choice unless it wants to be an even smaller, more idiosyncratic community than it currently is.
Thanks for this thoughtful comment! Thinking about x-risk reduction as giving us more time to grow the economy and alleviate poverty is really interesting.
While I agree the long-term effects are highly uncertain, I think it’s important to distinguish catch-up growth from frontier growth. Most growth accelerations in low-income countries bring them from “super poor” to “still pretty poor”. People in these countries live more comfortably, but they’re usually not getting rich enough to develop geopolitical ambitions that increase x-risk. (China and maybe India being notable exceptions.)
I’m actually not sure it’s true that “most of us have accepted longtermism.” As we say in this post, the Global Health and Development Fund is still the biggest EA Fund. Last year’s EA survey found that Global Poverty was still the most popular cause, and only 41% of respondents would choose the Long Term Future if they had to focus on one cause.
In any case, we might want to continue to have some EAs working on things other than longtermism in order to diversify in the face of moral uncertainty. And, as you say, having something useful and interesting to say about more mainstream causes is important for PR and movement growth. I thought the discussion of this point in the comments of this post was good.
It also seems like this comment could be made on any post that is not about long-termism, so there doesn’t to be anything especially relevant to this post here. If we don’t know whether growth is good in the long-term, then we presumably also don’t know whether eradicating malaria is either.
Also, I think growth plausibly is good from a long-termist point of view because it shortens the time of perils. It also has lots of beneficial political effects as it prevents zero sum rent seeking and encourages socially valuable activity.
Hi John, I’ll define here what I think you mean by “the time of perils”. I’ve heard of it before, but I had to Google it to refresh myself on what it means, and I got this from this forum post by Will MacAskill:
It’s not clear to me how growth “shortens the time of perils” without increasing the extinction risk during the time of perils, which would be bad from a longtermist perspective. If we accelerate economic growth, we would likely accelerate climate change, and we would likely become more technologically advanced faster.
Being more technologically advanced at a faster rate would mean we have less time to research on how to mitigate existential risks from emerging technologies, i.e. how to build safe and aligned AGI. But I’m happy to hear counterarguments to this view!
P.S. I think you can put something like “(the time wherein existential risk is unusually high)” after you use the phrase “the time of perils”, so people not familiar with the term could better understand what you meant!
Leopold Aschenbrenner has written about this here.
“The same technological progress that creates these risks is also what drives economic growth. Does that mean economic growth is inherently risky? Economic growth has brought about extraordinary prosperity. But for the sake of posterity, must we choose safe stagnation instead? This view is arguably becoming ever-more popular, particularly amongst those concerned about climate change; Greta Thunberg recently denounced “fairy tales of eternal economic growth” at the United Nations.
I argue that the opposite is the case. It is not safe stagnation and risky growth that we must choose between; rather, it is stagnation that is risky and it is growth that leads to safety.
We might indeed be in “time of perils”: we might be advanced enough to have developed the means for our destruction, but not advanced enough to care sufficiently about safety. But stagnation does not solve the problem: we would simply stagnate at this high level of risk. Eventually, a nuclear war or environmental catastrophe would doom humanity regardless.
Faster economic growth could initially increase risk, as feared. But it will also help us get past this time of perils more quickly. When people are poor, they can’t focus on much beyond ensuring their own livelihoods. But as people grow richer, they start caring more about things like the environment and protecting against risks to life. And so, as economic growth makes people richer, they will invest more in safety, protecting against existential catastrophes. As technological innovation and our growing wealth has allowed us to conquer past threats to human life like smallpox, so can faster economic growth, in the long run, increase the overall chances of humanity’s survival.
This argument is based on a recent paper of mine, in which I use the tools of economic theory—in particular, the standard models economists use to analyze economic growth—to examine the interaction between economic growth and the risks engendered by human activity.”
It seems plausible that it would have helped to have more rich countries capable of lobbying against the nuclear arms race (in terms of reducing x-risk).