Do in-kind transfers for social assistance lead to ‘deadweight losses’ by restricting consumer choice? This article presents findings from an experiment in Maharashtra, which involved offering low-income respondents the choice between a free quantity of rice and varying amounts of cash to elicit their willingness to pay for rice. It finds that women with higher bargaining power within the household are more likely to choose cash over rice.
Standard economic theory favours cash over in-kind transfers of equivalent values as they avoid the inefficiencies associated with restricting consumer choice – often referred to as deadweight losses. Yet, in-kind transfers, such as the provision of free or subsidised food, remain widespread in social programmes across the globe (Alderman, Gentilini and Yemtsov 2017). The debate over whether to prioritise cash or in-kind transfers continues to stir significant discussion, particularly in social assistance policies in many countries. In India, this issue has been central to policy debates surrounding the country’s large-scale public distribution system (PDS) that provides subsidised food to more than 800 million people (Khera and Somanchi 2020). The PDS has been criticised for corruption and ‘leakage’ of a substantial fraction of subsidised product which is illegally diverted to the open market, leading critics to call for cash transfers to replace the wasteful in-kind transfers. On the other hand, survey data have sometimes shown that households prefer food to equivalent cash transfers when offered a hypothetical choice between the two (Khera 2014).
‘Incentivised’ experimental evidence on this issue has, however, been lacking. Our new research addresses this gap by conducting a field experiment where respondents were given a real choice between a fixed quantity of rice and alternative amounts of cash (Abbink et al. 2024). The experiment involved incentivising the respondents to reveal their willingness to ay (WTP) for rice. A comparison of WTP with the market value of rice then provides a measure of the deadweight loss (or gain) associated with the in-kind transfer.
The experiment
We conducted our experiment with respondents from 250 randomly selected households across 10 slums of Nashik, a city in the state of Maharashtra. We conducted three rounds of the experiment in March, May and August 2019. Our respondents were targeted as the adult member in each household responsible for food purchases, which (unsurprisingly) were overwhelmingly women, comprising 90% of our sample. In each round, we offered respondents a choice between 5 kilograms (kg) of rice (of PDS-comparable quality) and nine alternative cash amounts ranging both below and above the going market value of rice. At the time of the experiment, the market value of 5 kg of rice was Rs. 160, and our cash offers ranged from Rs. 50 to Rs. 500. The 5 kg quantity of the in-kind transfer was chosen to ensure that the respondents faced an intra-marginal choice insofar as it represented less than one-third of their average monthly consumption of rice.
We incentivised the experiment by informing the respondents that we will record their choices, rice or cash, for each of the nine cash amounts. Following this, they will randomly draw one slip out of a bag with nine slips, each bearing the cash amount associated with one of the nine choices. The experiment will then end with the respondent receiving cash or rice according to their previously recorded preference against the cash amount they had randomly drawn from the bag.
As the respondents are offered increasing cash amounts against 5 kg of rice, the cash option becomes increasingly attractive. Our experiment aimed to identify the switch point where the cash option becomes preferable to rice for each respondent. This switch point offers a measure of the respondent’s WTP for 5 kg of rice. The difference between the market value of rice and the WTP then offers a measure of the deadweight loss (DWL). If the difference is negative, then the respondent experiences a deadweight gain (DWG).
Findings
Given that 90% of our sample respondents were female, our results focus on the subsample with female respondents to avoid confounding effects of respondent’s gender, though the key findings also generalise to the full sample.
Figure 1 presents the distribution of respondent choices against each of the cash offers, pooled over all three rounds. As expected, at higher cash amounts, a greater proportion of households opt for cash rather than rice, ranging from 32% choosing cash when offered the minimum amount of Rs. 50 to 85% when offered the maximum of Rs. 500. This pattern holds for all three rounds.
Figure 1. Share of respondents choosing cash against each cash offer, pooled across rounds
Based on the choices made by respondents, we can distinguish three types: first, the ‘single-switch respondents’ as those who made a single switch from rice to cash as higher cash amounts were offered; second, ‘rice-only respondents’ as those who chose rice for all nine cash offers; and third, ‘cash-only respondents’ as those who always chose cash.
Table 1 presents the estimates of WTP and DWL by respondent type. Across respondents, we find a combination of positive and negative DWL. By construction, DWL for cash-only respondents is positive and that for rice-only respondents is negative. It also turns out that cash-only respondents account for more than twice as many cases as rice-only respondents. A more striking result is that the average DWL for single-switch respondents is also negative with an estimate of Rs. -171. Also, for our sample as a whole, the average DWL is negative (that is, DWG overall), with a point estimate of Rs. 192 equivalent to 12% of the market value of rice.
Table 1. Distribution of willingness to pay and deadweight loss, by respondent type
Respondent type |
Number of cases |
Percent of cases |
WTP*** |
DWL*** |
(Rs) |
(Rs) |
|||
Cash-only |
208 |
32.7 |
25 |
135 |
Single switch |
341 |
53.5 |
177 |
-17 |
Rice-only |
88 |
13.8 |
550 |
-390 |
All |
637 |
100 |
179 |
-19 |
Notes: (i) We define WTP for rice for a respondent as the midpoint of the cash choice interval at which the respondent switched to rice to cash. (ii) DWL for respondent i is defined as DWL1 = 160-WTP1 where Rs. 160 is the market value of 5 kg of rice. (iii) Respondents with multiple switches are not included. (iv) For rice-only respondents, we assume WTP to be Rs. 550. (v) For cash-only respondents, we assume WTP to be Rs. 25. (vi) *** all reported values of WTP and DWL are significant at 1% level.
The puzzle of deadweight gain: Intra-household bargaining
The prevalence of DWG is widespread in our sample and is observed for about 45% of all cases. A priori, this result appears puzzling. The design of the experiment rules out transaction costs and quality differences as possible explanations for DWG. Crucially, both cash and rice are redeemed by the same process at the same local grocery shop, and the quality of rice offered was comparable to PDS rice. A lack of trust in the cash option can also be ruled out. The experiment was preceded by a pilot that was successfully implemented. Furthermore, none of the respondents reported any concerns or difficulties with redeeming the vouchers for cash or rice at the local shop.
The dimension that cannot be ruled out, however, is the role of intra-household bargaining in shaping respondent choices between cash or rice. Survey-based and qualitative evidence points to the potential role of intra-household inequality and gender in influencing the choice over in-kind or cash transfers (Khera 2011, 2014, Ghatak, Kumar and Mitra 2016). If women have a given allowance for the household food budget, provision of free or subsidised rice can augment their budget more than a cash transfer that would have to be shared with men. However, this consideration is likely to be less relevant for women who have greater bargaining power in the household. This motivates us to compare women respondents in male-headed households with women respondents in female-headed households, where they would have greater bargaining power.
This comparison uncovers an important finding: women respondents in female-headed households3 have a significantly lower WTP for rice than those in male-headed households (Rs. 151 versus Rs. 190). As a result, in contrast to a DWL for women from female-headed households of Rs. 9 on average (5% of the market value of rice), we observe a DWG for women from male-headed households of Rs. 30 (19% of the market value of rice) (Figure 2). These differences between male- and female-headed households in WTP and deadweight loss are statistically significant.
Figure 2. Deadweight loss for respondents from male- and female-headed households (Rs.)
The likelihood of choosing cash varies both by amount of cash offered as well female-headship. The right panel of Figure 3 shows that the probability of choosing cash is increasing in the cash amount for both male- and female-headed households, but is systematically higher for female-headed households at every cash option.4 Testing for statistical significance, we find that the marginal effects for female-headed households remain significantly higher up to the cash amount of Rs. 200 though not thereafter.
Figure 3. Marginal effects of cash amount on probability of choosing cash, for respondents from male- and female-headed households
Our main findings thus indicate that women in male-headed households (in light of their lower bargaining power) are more likely to choose rice than women in female-headed households so long as the difference between the market value of rice and the cash offer is not too large. When subject to conditions of lower bargaining power, women are willing to forgo a certain amount of cash as a strategy to protect their share of the household budget. This is the basis of the overall DWG we observe in our experiment.
We also look into a number of alternative mechanisms underlying this result including those related to learning, intra-household renegotiation, and the use of rice as a commitment device. We conclude that (i) differential learning rates for respondents by male/female-headship are not empirically plausible for our setting, (ii) while there is evidence of intra-household renegotiation over the three rounds of the experiment, this renegotiation occurs for women from male-headed households, which is consistent with our bargaining power-based explanation, and (iii) while women from both male- and female-headed households report using rice as a commitment device, a higher reported rate among male-headed households also points to women’s lower bargaining power as the underlying reason for their perceived greater need for a commitment device.
Conclusion
A key finding of our research is that whether there is a DWG or DWL associated with in-kind transfer depends on the balance of bargaining power within the household. We find DWL among women respondents from female-headed households and DWG among those from households with male heads. Given that the majority of households are male-headed, DWG dominates overall.
Most welfare programmes are designed to provide either only cash or only in-kind transfers. The existence of DWG associated with in-kind transfers, as in our experiment, does not necessarily imply that in-kind transfers must be the preferred policy option. Rather, a key policy insight of our study is that there is a case for offering respondents a choice between cash or kind. The offer of such a choice can be important for those with weaker bargaining power to sustain a measure of control over the household budget, even if for a limited period of time. Not all of the potential gains available through the choice between cash or kind may be dissipated even with a renegotiated control over the household budget.
Notes:
- Statistically different from zero with a p-value of 0.0001. A p-value is the probability of rejecting the null (maintained) hypothesis. For instance, a p-value of 0.05 implies a 5% probability of rejecting the null.
- p-value of 0.007.
- In our sample, 29% of the households are female-headed.
- These are estimates of marginal effects from a regression of choosing cash (rather than rice) on female-headship interacted with the cash amounts offered, while controlling for slum and round fixed effects and several respondent/ household-level attributes.
Further Reading
- Abbink, K, G Datt, L Gangadharan, D Negi and B Ramaswami (2024), ‘Deadweight Losses or Gains from In-kind Transfers: Experimental Evidence’, Economics Discussion Paper 110, Ashoka University.
- Alderman, HH, U Gentilini and R Yemtsov (2018), ‘The 1.5 Billion People Question: Food, Vouchers, or Cash Transfers?’, World Bank Report.
- Ghatak, Maitreesh, Chinmaya Kumar and Sandip Mitra (2016), “Cash versus kind: Understanding the preferences of the bicycle programme beneficiaries in Bihar”, Economic and Political Weekly, 51(11): 51-60.
- Khera, Reetika (2011), “Revival of the PDS: Evidence and explanations”, Economic and Political Weekly, 46(44 & 45): 36-50.
- Khera, Reetika (2014), “Cash vs. in-kind transfers: Indian data meets theory”, Food Policy, 46: 116-128.
- Khera, R and A Somanchi (2020), ‘A Review of the Coverage of PDS’, Ideas for India, 19 August.
Comments will be held for moderation. Your contact information will not be made public.