Christian, Glover, Kondylis, Loeser, Ruzzante, and Zwager (2026) study how the shape of agricultural input subsidies (how the subsidy rate varies with the quantity of inputs a farmer buys) affects which farmers use inputs and how much they produce. The paper’s setting is the eVoucher program of the UN Food and Agriculture Organization (FAO) in 10 districts of northern Mozambique, under which registered farmers redeem an electronic voucher at agrodealers for improved seeds, fertilizer, tools, and pesticides, choosing among three packages (Small, Medium, and Large) whose subsidy rate falls with size. Under the standard “Decreasing” schedule, a farmer pays 400 Mozambican meticais (MZN) for 4,000 MZN of inputs (a 90 percent subsidy), 1,200 MZN for 6,500 MZN (82 percent), or 4,200 MZN for 11,000 MZN (62 percent).
For the 2023 primary season, the paper randomly assigned 1,280 farmers in 187 villages to four schedules. The paper studies three alternatives to the standard “Decreasing” subsidy schedule:
Using voucher records and household surveys, the paper finds that the High & Decreasing schedule raised redemption of the Large package by 14.1 percentage points, less than the 19.5 points under Uniform: compared with Uniform, the cheaper Small and Medium packages divert some farmers who would otherwise have chosen Large. Its estimated effect on agricultural production, 229 MZN per farmer, is close to zero and not statistically significant. The paper infers that the diverted farmers were among those who gain most from additional inputs, so that the extra subsidies on the smaller packages largely offset the productive gains of the cheaper Large package. Because farmers who redeem the Large package are considerably richer, the paper concludes that High & Decreasing is both regressive and inefficient.
This entry reports the MVPF of moving from the standard Decreasing schedule to a High & Decreasing schedule. The paper also estimates the MVPF of the two other treatment arms: moving to a Sharply Decreasing schedule, which cuts only the prices of the Small and Medium packages, and moving to a Uniform schedule, which cuts only the price of the Large package.
MVPF = 0.4
The net cost is the increase in subsidy payouts, the money FAO pays agrodealers to cover the value of redeemed inputs beyond what farmers pay themselves. The paper estimates that High & Decreasing raised payouts by 1,253 MZN per registered farmer, the largest increase of the three schedules, relative to an average of 2,533 MZN under Decreasing (100 MZN is about 1.5 USD).
About a fifth of this cost is mechanical: farmers who would have redeemed the Small or Medium package anyway receive an extra 300 MZN each, and the 2.5 percent who would have redeemed the Large package an extra 2,600 MZN, together 237 MZN per farmer. The rest comes from farmers induced to take the Large package, whose subsidy rises to 9,400 MZN, and from farmers induced to redeem a voucher they would otherwise have left unused.
Net cost is the change in subsidy payouts alone: it includes no fiscal externality from any change in the farmers’ production, as the effective tax and transfer rate on that income is effectively zero.
The paper estimates willingness to pay as the change in consumer surplus and impacts on productivity.
Consumer surplus: The paper bounds farmers’ willingness to pay for the price cuts by revealed preference. The lower bound holds choices fixed as under Decreasing, crediting the 300 MZN cut to farmers who took the Small or Medium package there and the 2,600 MZN cut to those who took Large: 237 MZN. The upper bound holds choices fixed as under High & Decreasing, crediting the full cuts to every farmer who takes each package under that schedule: 589 MZN. The bounds are wide because many farmers move into the Large package, and each switcher’s gain could lie anywhere up to the full 2,600 MZN.
Productive impacts: The paper argues that farmers underestimate the returns to inputs and so do not fully weigh production gains when choosing a package; the planner therefore places a direct weight, γ, on the change in production. γ = 0.2 is the product of a discount factor of 0.6 (by the paper’s calculation, an 8 percent required monthly return over a crop cycle of about six months) and a bias of one third, the share of the production gain that farmers are taken to overlook. The bias comes from learning: Uniform’s effect on redemption of the Large package persisted for two years after the extra subsidy ended, which the paper reads as a rise of about 50 percent in farmers’ willingness to pay once they had used the inputs, and so as a sign that farmers had overlooked at least a third of the returns. Applied to High & Decreasing’s estimated production effect of 229 MZN, this is 0.2 × 229 = 45.8 MZN.
Total willingness to pay is therefore between 237 + 45.8 = 282.8 MZN and 589 + 45.8 = 634.8 MZN, with a midpoint of 458.8 MZN. The range reflects partial identification: the experiment bounds consumer surplus rather than pinning it down. It is not a confidence interval.
Dividing by the net cost gives an MVPF between 282.8 / 1,253 = 0.23 and 634.8 / 1,253 = 0.51, the unweighted bounds the paper reports in its Table 7. The headline of 0.37 is the midpoint of these partial-identification bounds.
On the same unweighted basis with γ = 0.2, Table 7 puts Sharply Decreasing at 0.60 to 0.65 (net cost 621 MZN) and Uniform at 1.18 to 1.50 (net cost 1,207 MZN). Without the weight on productive impacts (γ = 0), High & Decreasing’s MVPF is 0.19 to 0.47. The paper’s preferred specification also weights benefits by social marginal utility, with an inequality aversion of 1, which puts High & Decreasing at 0.21 to 0.47. Across the range of planner preferences the paper considers, High & Decreasing is nearly always beaten by one of the other two schedules. Once production carries any real weight, Uniform comes out ahead: High & Decreasing costs slightly more but adds little to output.

Christian, Paul, Steven Glover, Florence Kondylis, John Loeser, Matteo Ruzzante, and Astrid Zwager (2026). “Reshaping Agricultural Subsidies.” Working Paper. Available at: https://matteo-ruzzante.com/assets/pdf/cgklrz.pdf