Bridges to Prosperity

Bridges to Prosperity staff reviewed this page prior to publication.

Summary

Published: March 2020; Last updated: September 2022 (2020 version, 2021 version)

What is the problem?

According to the World Bank, access to reliable transportation infrastructure (roads, bridges, etc.) is limited in rural areas of many low- and middle-income countries.2

Researchers and policymakers have hypothesized that individuals in areas without reliable transportation infrastructure may face worse outcomes through several mechanisms:3

• They may be less able to access healthcare services.
• They may have lower access to and benefits from schooling.
• They may be isolated from labor markets, which can reduce opportunities for off-farm employment and have spillover effects on farm production as well.

What is the program?

Bridges to Prosperity is a nonprofit that builds trailbridges in rural communities.4 Bridges built by Bridges to Prosperity can support horses, livestock and motorcycles, though they are not designed to hold cars.5 These bridges are intended to connect isolated households to schools, health clinics and markets.6 They may be particularly beneficial in areas where flooding limits transportation during specific parts of the year.7

Bridges to Prosperity has worked in 20 countries in Latin America and sub-Saharan Africa and estimates it has built more than 350 trailbridges.8 The average bridge cost is $80,000, with costs often split between Bridges to Prosperity and government partners.9 Bridges are designed to last 30 years.10 In the past, Bridges to Prosperity has worked with local governments to identify need for bridges, but more recently, it has worked in partnership with national governments to identify sites for bridges. Once sites are selected, Bridges to Prosperity builds bridges in partnership with local contractors and suppliers.11

What is the evidence of effectiveness?

There is evidence from one quasi-experiment, Brooks and Donovan 2020, that Bridges to Prosperity improves income for households in rural Nicaragua. Brooks and Donovan 2020 find that bridges increased annual household income by 25% through increases in labor market earnings and farm profits.

We have a high degree of uncertainty about whether these same income effects are likely to occur in other settings where Bridges to Prosperity plans to operate in the near future. We anticipate that ongoing research from Rwanda may help resolve some of our questions about external validity.

In our cost-effectiveness analysis, we model the benefits of Bridges to Prosperity as operating through increases in household income. There may be additional benefits of bridges built by Bridges to Prosperity, as well as offsetting effects, that we have not incorporated into our cost-effectiveness analysis.

Quasi-randomized study in Nicaragua

Brooks and Donovan 2020 is a quasi-experiment conducted in rural Nicaragua.12

In this setting, Bridges to Prosperity identified villages that met the criteria for needing a bridge, based on village population, likelihood the bridge would be used, proximity to outside markets, and available alternatives.13 Among villages that passed this needs assessment, Bridges to Prosperity then conducted a feasibility assessment, based on whether villages possessed environmental characteristics that made them suitable for building a bridge.14

To measure the effect of bridges built by Bridges to Prosperity, researchers compared villages that passed both the needs and feasibility assessment and received a bridge to villages that passed the needs assessment but not the feasibility assessment and did not receive a bridge.15 They also compare the effects of bridges during weeks with and without flooding, since the bridges were designed to allow rivers to be passable during flooding and, as a result, should have a greater effect during floods.16

Brooks and Donovan 2020 report effects on household income from two sources: increases in off-farm employment income and increases in farm profits. Households in villages that received bridges had weekly labor market income that was 20% higher in non-flood weeks (p-value = 0.004 for main effect of bridges) and 48% higher in flood weeks (p-value = 0.038 for interaction effect of bridges during flood weeks).17 Households in villages that received bridges also had 76% higher farm profits (p-value = 0.020).18 These estimates suggest bridges increased household income by 25%.19

Overall, we view this study as reasonably high quality, though not conclusive or necessarily representative of other settings where Bridges to Prosperity may operate in the future.

While the trial is not a pure randomized controlled trial (RCT), the empirical approach seems reasonable, and households across villages receiving and not receiving bridges appear balanced at baseline.20 The lack of pure randomization may leave some concern that villages that meet feasibility criteria for bridge building could be different in other unobservable ways that might influence income and farm profit. We would weight this concern more heavily if feasibility were determined by villages' need for bridges or interest among individuals in villages in building bridges. However, because feasibility is determined by characteristics of riverbeds that are on average 1.5 kilometers from households, we guess that it is unlikely that riverbed characteristics would be correlated with other, unobservable village characteristics that would simultaneously affect household income or farm profitability.21

In addition to the main effects of bridges on income and the differential effects in flood vs. non-flood weeks, the study reports several additional empirical results consistent with the researchers' theory, which increases our confidence in the study's main findings. However, we have only reviewed these results superficially and do not have as strong a sense of their validity. These additional empirical results include:

• Households in villages that received bridges reported more employment outside the village than households in villages that didn't receive bridges (consistent with the findings on increased labor market income).22
• Wages were higher in villages that received bridges compared to those that didn't (consistent with general equilibrium effects driving equalization of wages inside and outside the village).23
• Households in villages with bridges had higher fertilizer use than households in villages without bridges (consistent with an increase in farm profits).24
• Bridges led to a decrease in household savings, in conjunction with an increase in investment in fertilizer. This is consistent with a model in which households have less need for savings once labor market income is available to smooth consumption and, as a result, choose to invest more in their farm through increased fertilizer use. Additionally, the authors find that decreases in savings were largest among farmers who increased their on-farm investment the most in response to bridges, which is also consistent with this model.25

Limitations of the study, which give us some uncertainty, include: a) the number of clusters for randomization (15) is relatively small,26 and b) the study was not pre-registered.

External validity and ongoing research in Rwanda

Given the nature of this intervention, we have high uncertainty about whether these findings will extend to new settings. The increases in income found in Brooks and Donovan 2020 rely on an adequate supply of labor market opportunities outside bridge-connected villages, a relative lack of labor market opportunities inside bridge-connected villages during floods, rivers that are impassable during floods but become passable when bridges are installed, weather patterns that make flooding sufficiently frequent to affect behavior, and potentially other conditions as well.

Our rough impression is that these conditions may not generalize to new settings. Because our understanding is that Bridges to Prosperity will likely prioritize bridge-building in Rwanda and Uganda in the future (see below), understanding generalizability in these or similar settings will be particularly important.

We are aware of two ongoing follow-up studies of the impact of Bridges to Prosperity in Rwanda. We are likely to place a high value on the findings of these studies because we expect them to provide more information on the generalizability of the findings in Brooks and Donovan 2020.

The main study is a large-scale RCT that will use step-wedge randomization to measure the effect of more than 250 bridges built by Bridges to Prosperity in Rwanda over five years.27 Results are not expected for several years. In preparation for that larger study, researchers have been conducting a pilot study that uses a quasi-experimental approach that is similar to that used in Brooks and Donovan 2020.28 While the purpose of the pilot study is to understand the feasibility of conducting the full-scale RCT, it may also provide evidence on the effect of Bridges to Prosperity in Rwanda. Full results from the pilot study are likely to be available by mid-2020.

We have seen preliminary results from this pilot study, and while our impression is that these results show mechanisms consistent with Brooks and Donovan 2020, we have not undertaken a thorough review since the results are likely to change once all data are available. We also do not yet have a strong understanding of the quality of the pilot study (e.g., statistical power, likelihood of bias from quasi-experimental approach) and so do not know how much the full pilot results will update our assessment of Bridges to Prosperity's impact in Rwanda. We plan to undertake further review of the pilot study once full results are available.

Additional benefits

Bridges built by Bridges to Prosperity may confer other benefits to households, beyond increases in labor market income or farm profits. Bridges to Prosperity has suggested that bridges may also improve access to healthcare services, make it easier for children to attend school, or improve access to government programs.29 We have not seen any direct evidence for or against these effects, and we currently factor these additional benefits into our assessment of the cost-effectiveness of Bridges to Prosperity using a rough estimate.30

Potential offsetting/negative effects

One possible negative effect of bridges built by Bridges to Prosperity is that increases in labor supply from individuals in previously isolated rural villages may drive down wages in villages receiving new workers, which can lower income for individuals in those receiving villages. We have not reviewed in depth how large this effect might be, and our cost-effectiveness model only includes a rough estimate on this effect.31 We have also not thoroughly vetted the safety of the bridges built by Bridges to Prosperity.

How cost-effective is it?

In May 2022, we updated our cost-effectiveness analysis for Bridges to Prosperity as part of an investigation into a grant to support an ongoing randomized controlled trial of Bridges to Prosperity in Rwanda. The updated cost-effectiveness model for this intervention is available here. 32

Note that our cost-effectiveness analyses are simplified models that do not take into account a number of factors. There are limitations to this kind of cost-effectiveness analysis, and we believe that cost-effectiveness estimates such as these should not be taken literally due to the significant uncertainty around them. We provide these estimates (a) for comparative purposes and (b) because working on them helps us ensure that we are thinking through as many of the relevant issues as possible.

The cost-effectiveness analysis models the effect of Bridges to Prosperity in Nicaragua, based on the findings from Brooks and Donovan 2020, and in Rwanda, where there is ongoing research and which may more closely resemble future implementation settings for Bridges to Prosperity (see below).

In early 2022, we updated our cost-effectiveness model for Bridges to Prosperity by:

• Adding a leverage and funging estimate. We adjust our cost-effectiveness estimates to account for the extent to which we believe our funding may influence how other funders spend their funds (leverage) or may crowd out funding that would otherwise have come from other sources (funging). See more details in this blog post. We have added parameters to estimate leverage and funging here.
• Differentiating between effects on households who live close to the bridge and those that live farther from the bridge. We assume that those who live closer to the bridge will have more benefits, but that those who still use the bridge but live farther away will benefit as well.33
• Updating our estimate of effects on household consumption. We updated our cost-effectiveness model to incorporate information from three additional studies. New sources we incorporate include a quasi-experimental study of Bridges to Prosperity in Rwanda (Thomas et al. 2021), and two quasi-experimental studies of rural road building (not bridge-building) programs in Ethiopia (Nakamura et al. 2019) and India (Asher and Novosad 2019).34 We continue to use information from the Brooks and Donovan 2020 study in Nicaragua.
• Adding supplemental adjustments. We also added parameters to our model to estimate (very roughly) effects of bridges on outcomes other than consumption, such as access to other services, as well as downside adjustments.35

Overall, we estimate that Bridges to Prosperity's program in Nicaragua is roughly as cost-effective as direct cash transfers, given the relatively small number of households affected. While Brooks and Donovan 2020 find a large effect on income (25% increase in annual household income) that we estimate endures over 15 years, the number of households affected per bridge appears to be small (40, according to our best guess) and bridge cost is relatively expensive ($88,000 per bridge plus maintenance costs).36 In Rwanda, these updates have led us to decrease our estimates of cost-effectiveness. Our best guess is that in Rwanda, Bridges to Prosperity is just below the range of cost-effectivness of programs we would recommend funding.37

However, these calculations are based on several key parameters, about which we have a high degree of uncertainty and that could affect our cost-effectiveness estimates:

• Number of households affected per bridge. Brooks and Donovan 2020 collect outcome data for individuals within three kilometers of proposed or actual bridge sites38 and estimate that there are 34 households on average per village.39 We have high uncertainty about the extent to which additional households that were farther away and therefore not surveyed might have also experienced income benefits from bridges. Bridges to Prosperity has indicated to us that its estimates suggest substantially more households were affected by bridges built in Nicaragua, but we have not vetted these numbers or sought to estimate what income effects we would expect for these households.40 We have included a rough guess that 20% more households than those surveyed by Brooks and Donovan 2020 saw income effects of the magnitude they observed in the Nicaragua trial.41 However, we have not conducted a thorough investigation into this parameter, and it is possible that this estimate is still too conservative. If there were additional households affected, our cost-effectiveness for Nicaragua will be underestimated. Under the assumption that relatively few households were affected per bridge, the findings from Nicaragua suggest that Bridges to Prosperity is roughly as cost-effective as cash transfers.
• Duration of benefits. Bridges to Prosperity estimates that bridges last 30 years.42 However, we might expect the effect of bridges on income to endure for less than 30 years for a variety of reasons. First, bridges may break down at a higher rate than estimated, due to overly optimistic projections or unforeseen events like soil erosion. Second, it's possible that in the counterfactual world without Bridges to Prosperity, local or national governments would build these bridges regardless.43 In our current model, we assume that effects persist for 15 years. We are highly uncertain about this parameter and view it as a rough guess.
• Bridge cost and leveraging of funding from the government. Bridges to Prosperity estimates that bridges cost $80,000 on average (though cost varies across settings), and that it shares 40%-50% of costs with government partners.44 We have not thoroughly vetted these estimates and take them at face value in our current cost-effectiveness model. We also do not have a well-vetted estimate of maintenance costs and instead include an extra 10% to the $80,000 bridge cost estimate to capture our best guess.45
• Effects in Rwanda. In Rwanda, the number of households per bridge is likely to be much higher.46 As a result, Bridges to Prosperity could have much greater cost-effectiveness in Rwanda, even with much lower income effects. On the other hand, it is possible that income effects could be negligible in Rwanda or that only a small subset of households could be affected, which could cause Bridges to Prosperity's cost-effectiveness to fall outside of the range of programs we would fund. We are likely to substantially update our model once we review results from the ongoing study in Rwanda.
• Additional outcomes. We are uncertain about the effects on households' health, educational, and other outcomes that are not directly related to income. In addition, we are uncertain about the effect on households who are on the side of the bridge that is less affected by the bridge. The upcoming study in Rwanda may update our views on these questions.

Is there room for more funding?

Bridges to Prosperity reports that it has significant room for more funding in the next several years. It estimates that it has room for more funding for an ongoing nationwide bridge-building project in Rwanda (roughly $10 million) and a planned nationwide bridge-building project in Uganda (roughly $14 million to $50 million).47 Bridges to Prosperity is also currently deciding where to target its intervention in future years, which will likely also result in additional room for more funding.

Key questions for further investigation

• What are the appropriate values for the number of households affected by bridges, duration of bridge effects, bridge costs and leveraging of funding from governments?
• What effects on income will the ongoing research on Bridges to Prosperity in Rwanda find?
• Are these effects likely to generalize to other settings where Bridges to Prosperity has room for more funding or may have room for more funding in the future (e.g., Uganda)?
• How safe are bridges built by Bridges to Prosperity?
• What are the key mediating factors that influence the effect of Bridges to Prosperity across settings, and how might we model these in our cost-effectiveness analysis?
• How important are any potential additional benefits from bridges built by Bridges to Prosperity (access to healthcare, education and government programs)?
• What is the effect of bridges built by Bridges to Prosperity on households in areas on the "other side" of the bridge (i.e., areas that were not isolated but are now connected to previously isolated areas)?
• Are there additional long-term effects of bridges that short-term studies miss?

• What type of monitoring and evaluation does Bridges to Prosperity conduct to determine, e.g., the number of individuals utilizing bridges or bridge quality?
• Given the long time horizon projected for the effects of bridges, are there additional components we should incorporate into our cost-effectiveness analysis (e.g., changes in weather patterns and the probability of flooding predicted over time)?

Sources

Document	Source
Adukia et al. 2019	Source (archive)
Asher and Novosad 2019	Source (archive)
Bridges to Prosperity, FAQs	Source (archive)
Bridges to Prosperity, Funding gap analysis 2022-2025	Source
Bridges to Prosperity, Home page	Source (archive)
Bridges to Prosperity, Our approach	Source (archive)
Bridges to Prosperity, Strategic plan 2020-2025 appendix	Source
Brooks and Donovan 2020	Source (archive)
GiveWell, Bridges to Prosperity CEA for Intervention Report (2020)	Source
GiveWell, Update to Bridges to Prosperity CEA for Intervention Report (April 2022)	Source
GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019	Source
GiveWell's non-verbatim summary of a conversation with Professor Wyatt Brooks and Professor Kevin Donovan, January 6, 2020	Source
Iimi et al. 2016	Source (archive)
Nakamura et al. 2019	Source (archive)
Roberts et al. 2006	Source (archive)
Starkey and Hine 2014	Source (archive)
Thomas et al. 2021	Source (archive)
World Bank 2016	Source (archive)
World Bank, Bangladesh: World Bank Supports Rural Roads and Bridges	Source (archive)
World Bank, Transport overview	Source (archive)

• 1
  • “The study employed a matched-cohort design, in which 12 bridge sites were matched to 12 comparison sites.” Thomas et al. 2021, p. 3. We view this type of study as less reliable than, for example, a well-conducted randomized controlled trial.
  • “A major complication during t = 2 data collection was the COVID-19 crisis. This stopped data collection after only some data was collected, and generates an unbalanced panel of households, which will be important for interpreting data from the t = 2.” Thomas et al. 2021, p. 5. We guess this may have limited the power of the study and also may have introduced bias, since households who weren’t reached through data collection were likely not random.

• 2

  See, for example:

  • "In developing countries, particularly in Africa, the vast majority of farmers do not have good access to the local, regional, or global market, and depends on subsistence farming with few advanced inputs. Limited connectivity is a critical constraint in accessing social and administrative services, especially in rural areas where the majority of the poor live." Iimi et al. 2016, p. 2.
  • "The Rural Access Index (RAI), developed by Roberts, Shyam, and Rastogi (2006), is among the most important global development indicators in the transport sector. It measures the fraction of people who have access to an all-season road within a walking distance of approximately 2 kilometers (km). The original work relied on available household surveys. Although there remains some ambiguity about the methodologies used across countries, the RAI was estimated at 68.3 percent, leaving about one billion rural residents unconnected in the world (map 4). There is significant inequality across regions: while nearly 90 percent of the rural population in East Asia and Pacific has 2 km access to the road network, in Sub-Saharan Africa the RAI is estimated at only 33.9 percent (figure 7). In general, the RAI is expected to increase as the economy grows. African countries are clearly lagging behind at any particular level of rural accessibility (figure 8)." World Bank 2016, pp. 1-2.
  • "The new method was applied to eight pilot counties: Ethiopia, Kenya, Mozambique, Tanzania, Uganda, and Zambia in Africa, and Bangladesh and Nepal in South Asia. Rural access varies significantly across these countries, from 17 percent in Zambia to 56 percent in Kenya (figure 2). In total, it is estimated that about 34 percent of the rural population is connected, with roughly seven million people left disconnected." World Bank 2016, p. Xi.

• 3

  Citations from researchers and policymakers describing potential mechanisms:

  • "Transport infrastructure connects people to jobs, education, and health services; it enables the supply of goods and services around the world; and allows people to interact and generate the knowledge and solutions that foster long-term growth. Rural roads, for example, can help prevent maternal deaths through timely access to childbirth-related care, boost girls' enrolment in school, and increase and diversify farmers' income by connecting them to markets." World Bank, Transport overview.
  • "Many of the world's poor people live in rural areas isolated by distance, terrain and poverty from employment and economic opportunities, markets, healthcare and education. Lack of basic infrastructure (paths, trails, bridges and roads) and access to transport services makes it difficult for poor people to access markets and services." Starkey and Hine 2014, p. 4.
  • "Access to employment outside the village may play an important insurance role, and improved access to external health and education services may be valuable; indeed, we find elsewhere that rural roads cause increases in educational attainment (Adukia et al., 2017)." Asher and Novosad 2019, p. 26.
  • "In practice, improving access to transport for rural men and women in low income countries is considered essential to promote rural development, to increase uptake of human development services (educational and health), to facilitate inclusion of different ethnic and other groups, to improve employment opportunities, and to stimulate growth for poverty reduction." Roberts et al. 2006, p. 1.
  • "For example, 18 percent of households in the control group mentioned clinic and maternal health care access as an important service cut off by flooding." Brooks and Donovan 2020, p. 72.
  • "'Rural bridges play a key role in Bangladesh's development, and an efficient rural road network can have a big effect in improving rural livelihoods,' said Qimiao Fan, Country Director for Bangladesh, Bhutan, and Nepal. 'By enabling greater connectivity, these two projects will help rural communities gain safer access to schools, health facilities and markets, reduce transport costs, increase non-agriculture incomes, and expand employment for both women and men.'" World Bank, Bangladesh: World Bank Supports Rural Roads and Bridges.
  • "Connections to new markets should encourage educational attainment if they increase returns to education, or otherwise raise household income or liquidity. However, immediate earnings opportunities for the young could motivate an earlier exit from schooling. As educational investment responds to market integration, it shapes the long-run economic impacts of policies that are increasingly integrating markets in developing countries." Adukia et al. 2019, p. 1.
  • "Increased integration has potentially large benefits in rural areas where household income is derived from both farming and labor markets, a common feature of income-generating activities in the developing world (Foster and Rosenzweig, 2007). [FN: The direct effect is access to higher wages outside their village (Bryan, Chowdhury and Mobarak, 2014; Bryan and Morten, 2018). However, to the extent that wage income allows farmers to relax credit constraints or better manage risk, it may simultaneously decrease farm-level distortions. Mobarak and Rosenzweig (2014) and Karlan et al. (2014), among others, find benefits from formal rainfall insurance, while Jayachandran (2006) and Fink, Jack and Masiye (2017) show how missing credit markets affect agricultural employment and production. Thus, these margins potentially play an important role.] Thus, understanding spillovers between wage work and farm decisions are necessary to understand the full effect of labor market integration." Brooks and Donovan 2020, p. 1.

• 4

  Bridges to Prosperity, FAQs:

  • "B2P is a 501(c)(3) nonprofit organization, ID #54-2031102."
  • "We believe that safe access is a human right. Access is an equalizer, creating equal opportunity for all members of a community. We act on this belief by turning rural isolation into connection. We’re a non-profit organization that connects isolated communities to education, health care, and economic opportunity. We do this by working alongside community members, industry partners, and governments to build trailbridges.​"

• 5

  "Bridges cannot be crossed by cars, but can support horses, livestock, and motorcycles." Brooks and Donovan 2020, p. 8.

• 6
  • "We design, build, and maintain durable and environmentally sustainable bridges to connect the rural last mile to the rest of the world." Bridges to Prosperity, Our approach.
  • "The isolated communities where we work are cut off from schools, health clinics, and markets for months out of the year." Bridges to Prosperity, FAQs.

• 7

  "The rivers our trailbridges span are perilous to cross on foot, and they often become completely impassable during the rainy season. Trailbridges create improved access to essential services and opportunity." Bridges to Prosperity, FAQs.

• 8
  • "We've built more than 350 trailbridges, serving over 1.2 million community members throughout the world." Bridges to Prosperity, Home page.
  • "B2P has worked in 20 countries since 2003." GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019, p. 1

• 9

  GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019:

  • "B2P estimates that its 350 bridge project in Rwanda will cost a total of $28 million, averaging $80,000 per bridge." P. 2
  • "B2P’s aim is to have each government pay for half of the costs of the bridge building projects in that country." P. 4

• 10

  "B2P’s technical advisors use an estimated lifespan of 30 years." GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019, p. 6

• 11

  GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019:

  • "Until a few years ago, B2P primarily identified projects through a local process, one bridge at a time. This often involved leveraging the connections of its board members or approaching local governments to ask where a bridge was needed and seeking to expand its contacts from there. B2P’s current process is to begin by approaching national ministries and advocating for funding at the national level." P. 1
  • "The total cost of one of B2P’s bridges varies significantly depending on:
    • cost of employment in the country that B2P is working in
      [...]
    • local availability of materials… B2P will have materials that are not locally available… shipped in." P. 3

• 12

  "In this paper, we directly study the impact of integrating rural Nicaraguan villages with outside labor markets and show empirically that it has sizable effects on household wage earnings, farm investment decisions, and savings." Brooks and Donovan 2020, p. 1.

• 13

  "B2P takes requests from local village organizations and governments, then evaluates these requests on two sets of criteria. First, they determine whether the village has sufficient need. This assessment is made based on the number of people that live in the village, the likelihood that the bridge would be used, proximity to outside markets and available alternatives." Brooks and Donovan 2020, p. 8.

• 14

  "If the village passes the needs assessment, the country manager conducts an engineering assessment. The purpose of this assessment is to determine if a bridge can be built at the proposed site that would be capable of withstanding a flash flood. To be considered feasible, the required bridge cannot exceed a maximum span of 100 meters, and the crests of the riverbed on each side must be of similar height (a differential not exceeding 3 meters). Moreover, evidence of soil erosion is used to estimate water height during a flood. The estimated high water mark must be at least two meters below the proposed bridge deck." Brooks and Donovan 2020, p. 8.

• 15

  "We compare villages that passed both the feasibility and the needs assessments, and therefore received a bridge, to those that passed the needs assessment, but failed the feasibility assessment." Brooks and Donovan 2020, p. 8.

• 16

  Brooks and Donovan 2020:

  • "During the rainy season, floods cause stream and riverbeds that are usually passable on foot to rise rapidly and stay high for days or weeks. … During these periods, villages are cut off from access to outside markets. However, it is important to emphasize a number of features of this flooding risk that are relevant for interpreting our results. First, floods are intense torrents of water from the mountains, not simply villages situated next to rivers. Thus, crossing the river by swimming, or any other method, entails substantial risk of injury or death. These floods usually generate prohibitively dangerous crossing conditions or a long journey on foot to reach the market by another route." Pp. 4-5.
  • "The bridges we build traverse potentially flooded riverbeds, thus allowing village members consistent access to outside markets. We partner with Bridges to Prosperity (B2P), a non-governmental organization that specializes in building bridges in rural communities around the world." P. 7-8.

• 17

  Table 3, column 1. Brooks and Donovan 2020, p. 34:

  • For non-flood weeks, the effect of a bridge is C$159.424 (p-value = 0.004) from a base of C$783.563. C$159.424/C$783.563 ≈ 20%
  • For flood weeks, the effect of a bridge is C$148.699 (p-value = 0.038) plus C$159.424 (p-value = 0.004) from a base of C$783.563 minus $C143.627. (C$148.699+C$159.424)/(C$783.563-$C143.627) ≈ 48%.
  • Note: Confidence intervals are not reported.

• 18
  • Table 6, column 9. Brooks and Donovan 2020, p. 45. The effect of a bridge is C$1957.61 (p-value = 0.020) from a base of C$2559.20. C$1957.61/C$2559.20 ≈ 76% Note: Confidence intervals are not reported.
  • The authors argue that flood weeks are unlikely to affect farm profits and do not report differential effects by flooding: "Second, it is unlikely that the flooding has any direct effect on village farms, as the average household is nearly a mile (1.5 kilometers) from the river." Brooks and Donovan 2020, p. 5. We have not vetted this claim but our impression is that it seems reasonable.

• 19

  See calculations here.

• 20

  Table 2. Brooks and Donovan 2020, p. 33.

• 21

  "Our identification strategy is based on the fact that many villages need bridges, but construction is infeasible for some villages due to the characteristics of the riverbeds that they aim to cross. Because these rivers are typically distant from the houses and farmland of the village (the average village household is 1.5 kilometers from the potential bridge site), the failure to pass the engineering assessment is orthogonal to any relevant household or village characteristics. We verify this by showing that baseline characteristics are balanced across villages that do and do not fulfill the engineering requirements, which we detail in Section 2." Brooks and Donovan 2020, pp. 1-2.

• 22

  "Men shift employment from inside to outside labor market work. In the average household, the number of males working outside increases by 0.19 (p = 0.000), compared to a 0.12 person decrease (p = 0.128) inside the village. Combined they generate a statistically insignificant net change in the number of males employed." Brooks and Donovan 2020, p. 14.

• 23

  "Next, we find that male daily wages inside the village increase by C$69 (p = 0.092), consistent with general equilibrium effects resulting from the decreased labor supply induced by the bridge. The male wages outside the village do not change (-C$5.6, p = 0.816) because these villages account for a small fraction of labor market activity outside their village. The wage gap between inside-village and outside-village employment, therefore, converges for men." Brooks and Donovan 2020, p. 14.

• 24

  "First, we see a very large increase in intermediate expenditures. Intermediate expenditures increase by C$659.97 (p = 0.048) on a baseline mean of C$890. The changes are primarily accounted for by fertilizer investment, which increases by C$383 (p = 0.026) compared to a statistically insignificant C$167 (p = 0.260) for pesticide." Brooks and Donovan 2020, p. 15.

• 25

  "Bridges improve market access and increase earnings, which may result in greater savings. However, they also smooth the earnings process of workers that can now consistently reach those markets, which may reduce motives for precautionary savings. The key liquid savings vehicle in rural Nicaragua is storage of staple crops. Storage is defined as quantity harvested net of sales, debt payments, gifts, and land payments, measured as a share of total harvest quantity. We measure this in terms of quantities for both maize and beans. Any household with no crop production is given a value of zero in this regression. Table 7 shows how bridges affect savings behavior. Regressions 1 and 3 show the average effect. Farmers save about 9 percentage points less of both their maize harvest (p = 0.014) and their bean harvest (p = 0.052). Columns (2) and (4) again show that the decrease in storage is concentrated among continuing farmers, the same subgroup as those who increase investment. Among continuing farmers, we find decreases of 13 percentage points for maize (p = 0.016) and 17 percentage points for beans (p = 0.056). Among those who did not farm at baseline, we see small and statistically insignificant changes in storage rates across build and no-build villages.
  "Taken together, these results suggest that farmers are selling a greater share of their harvests and using the proceeds to invest more in their farms. We test this directly by correlating changes from baseline intermediate expenditures with changes for baseline storage among treatment households. The correlation is -0.28 when using corn storage and -0.34 for bean storage. Both are statistically significant at the one percent level. This demonstrates that those who are increasing fertilizer the most are also those decreasing their savings the most." Brooks and Donovan 2020, p. 17.

• 26

  The study includes 15 villages. The average number of households per village is 33.5, and the authors report 97 percent participation in the first round of the annual survey. Average sample size in main regressions we use in our cost-effectiveness analysis is 6,443 (effects on labor market income, Table 5, column 1, p. 42) and 1,493 (effect on farm profit, Table 9, column 3, p. 45). Annual data were collected in three waves from 2015-2017 (one wave before bridge construction and two waves following bridge construction), and biweekly data were collected over a period of 64 weeks starting with the rainy and dry seasons before construction through the end of the rainy and dry seasons after construction. Citations below:

  • "We study a total of fifteen villages. Of these, six passed both the needs and feasibility assessments, and therefore received bridges. The other nine passed only the needs assessment and did not receive a bridge." Brooks and Donovan 2020, p. 9.
  • "The data used in our analysis comes from surveys conducted at the end of the main rainy season, in November 2014, November 2015, and November 2016. Bridges were constructed in early 2015. Therefore we have surveys from three years for all villages. For those that receive a bridge, we observe one survey without a bridge and two surveys with a bridge. We refer to these survey waves as t = 0, 1, 2." Brooks and Donovan 2020, p. 9.
  • "The second component of our data is biweekly follow-up surveys conducted by phone with a subset of households. Because floods are high frequency and short term events, this data shows the contemporaneous effect that flooding has on households. We carried out these surveys for 64 weeks, covering the rainy season before construction, along with the first dry and rainy seasons after construction." Brooks and Donovan 2020, p. 10.
  • "On average, there are 33.5 households per village." Brooks and Donovan 2020, p. 72.
  • "Participation in the first round of the survey was very high in general, with 97 percent of households agreeing to participate." Brooks and Donovan 2020, p. 10.

• 27

  "Over the next five years, Professors Brooks and Donovan will be conducting a randomized controlled trial (RCT) on the effect of 267 B2P bridges in Rwanda… The RCT is a stepped-wedge design (i.e. treatment is delivered in "steps" rather than simultaneously)." GiveWell's non-verbatim summary of a conversation with Professor Wyatt Brooks and Professor Kevin Donovan, January 6, 2020, p. 3

• 28

  "Professors Brooks and Donovan have been conducting an initial pilot study in Rwanda over the past year and are continuing to collect data and analyze results. The pilot was intended to test and adapt the survey design and field methods used in Nicaragua to Rwanda in order to ensure operational efficiency during the full study." GiveWell's non-verbatim summary of a conversation with Professor Wyatt Brooks and Professor Kevin Donovan, January 6, 2020, p. 4

• 29
  • "B2P is interested in the long-term impacts of the bridges on:
    • health care
    • education
    • collective efficacy
    • access to social/government programs
    • women’s empowerment (the Nicaragua study documented a 60% increase of women joining the market)
    • climate resilience (flash floods can inhibit access to markets when no bridge is available)."

    GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019, p. 4

  • Brooks and Donovan 2020 also note, "Note, however, that there are potentially other effects we are missing. For example, 18 percent of households in the control group mentioned clinic and maternal health care access as an important service cut off by flooding." P. 72.

• 30See here in our cost effectiveness model.
• 31See here in our cost-effectiveness model.
• 32 Our cost-effectiveness model from 2020 is available here.
• 33See these parameters here and here in our cost-effectiveness model.
• 34These studies all have limitations. We weigh the studies based on how similar they are in program and setting to Bridges to Prosperity's Rwanda program here.
• 35See these adjustments here.
• 36

  See the cost-effectiveness model here for details on calculations of income effects.

• 37See here for our previous cost-effectiveness model, 'Best Guess of Effect of Bridges to Prosperity in Rwanda' column, 'Bridges vs cash' row. See here in our updated cost-effectiveness model, 'Best Guess of Effect of Bridges to Prosperity in Rwanda' column, 'x cash after leverage and funging' row.
• 38

  Brooks and Donovan 2020 collect outcome data for individuals in villages immediately adjacent to proposed or actual bridge sites ("Our strategy was to survey all households within three kilometers of the proposed bridge site on the side of the river that was intended to be connected." Brooks and Donovan 2020, p. 9) and estimate 33.5 households per village ("On average, there are 33.5 households per village." Brooks and Donovan 2020, p. 72). It is not clear from Brooks and Donovan 2020 whether there are other households in other villages that were also likely to have seen income effects from the bridges. However, it is our understanding that in the Nicaragua setting there were likely to be fewer households affected per bridge than other settings.

• 39

  "On average, there are 33.5 households per village." Brooks and Donovan 2020, p. 72

• 40

  Bridges to Prosperity's estimates of households affected are provided in this document.

• 41 See more details in a cell note here in our cost-effectiveness analysis, row "Households affected by each bridge".
• 42

  "B2P’s technical advisors use an estimated lifespan of 30 years." GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019, p. 6

• 43

  It's also possible that income effects may fade out over time, though it may also be possible that benefits compound over time.

• 44

  GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019:

  • "B2P estimates that its 350 bridge project in Rwanda will cost a total of $28 million, averaging $80,000 per bridge." P. 2
  • "B2P’s aim is to have each government pay for half of the costs of the bridge building projects in that country. In Rwanda, B2P is paying about 60% of the expenses. In Uganda, the expenses are split 50/50." P. 4

• 45See here in our cost-effectiveness analysis.
• 46

  "B2P also believes that the number of households affected by each bridge in Rwanda will be much larger than in Nicaragua." GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019, p. 5

• 47

  "B2P’s funding gap in Rwanda is about $9.5 million to $10 million… B2P’s funding gap in Uganda is in the range of $14 million to $50 million." GiveWell's non-verbatim summary of a conversation with Avery Bang and Dr. Christina Barstow, December 12, 2019, p. 7