Economies of scale of large-scale international development interventions: Evidence from self-help groups in India

Abstract

Livelihoods and microfinance programs for women often show reduced impacts after scale-up. Yet, program scale-up may reduce average per capita costs and maintain cost-effectiveness despite lower impact. This paper presents evidence on the association between program scale, costs, and cost-effectiveness by analyzing how the costs of a large-scale Self-Help Group (SHG) program in India changed from its inception in 2007 to its scale-up in 2019. We use expenditure data from program's audit statements of Jeevika - the Bihar Rural Livelihoods Promotion Society - and find that a 1% increase in program membership was associated with a 0.6% increase in annual program expenditures, indicating large economies of scale. Predicted costs from regressions suggest that the annual per capita program expenditures declined from $29 when the program covered 100,000 members to $5 when it reached 10 million members. Previous impact evaluations of Jeevika showed sizeable but smaller substitutions away from high-cost debt after scale-up than during the pilot, but we found that economies of scale led to similar cost-effectiveness ratios for this outcome. We also found that formation of higher-level federations is associated with lower marginal costs than setting up SHGs. However, previous evidence suggests that Jeevika did not generate average impacts on women's agency and asset ownership after scale-up. Building on a rich history of research on Jeevika, we argue that program implementers must identify key success factors in pilot programs to minimize tradeoffs between cost savings and potentially reduced impacts after scale-up. Further, we suggest investments in linking SHGs to federations to improve the cost-effectiveness of SHGs.

Keywords: Cost-effectiveness; Costing; Economies of scale; India; Scaling up; Self-help groups.

Fig. 1

Jeevika Program Scale-Up. Note: The x-axis depicts the start of financial year, and ranges from 2007 to 08 to 2018–19.

Fig. 2

Jeevika Annual non-CIF Expenditure Over Time. Note: The x-axis depicts the start of financial year, and ranges from 2007–08 to 2018–19. Y-axis shows per member combined annual expenditure on Community Institutional Development, Special Technical Assistance Fund, and Project Management. Model 1 assumes that expenditures are divided across all members irrespective of when they joined the program. Model 2 assumes that all members are supported for three years after joining the program.

Fig. 3

Jeevika non-CIF Expenditure Components Over Time. Note: The x-axis depicts the start of financial year, and ranges from 2007–08 to 2018–19. Y-axis shows per member annual expenditure by expenditure component. CID: Community Institutional Development; STAF: Special Technical Assistance Fund: PM: Project Management. Model 1 assumes that expenditures are divided across all members irrespective of when they joined the program. Model 2 assumes that all members are supported for three years after joining the program.

Fig. 4

Change in High-Cost Loan Amount for Every Dollar Spent on the Program. Note: All estimates are in 2018 USD. Error bars represent 95% confidence intervals. Costs are based on Model 1 assumptions –expenditures are divided across all members irrespective of when they joined the program.

Fig. 5

BRAC Microfinance Costs: 2013 to 2017. Note: The chart shows per member annual expenditure on y-axis, and number of members on x-axis. Bangladesh is shown separately from Uganda and Tanzania because of the wide disparity in scale. Scatter plots with polynomial fit lines are shown in both charts.