Published on Development Impact

A $1 is a $1 is a $1, or is it?

This page in:
Development Impact logo

There is a large body of evidence that cash transfers in low and middle income countries result in improvements in consumption, poverty, food security, savings, investments/asset accumulation, education, health, etc… Among other reviews, see Bastagli, Hagen-Zanker, Harman, Barca, Sturge, Schmidt 2021 and Ralston, Andrews, and Hsiao 2017. Of course, there can be substantial heterogeneity in the impacts (by outcome, across contexts). And for some domains, health and women’s empowerment, results are more muted. In a previous blog with Emanuela, we note that improving child health (especially anthropometrics) can require much more than delivering cash to households. It can be greatly dependent on broader infrastructure conditions (local area water, sanitation and hygiene), behavioral change, and age-sensitive responses. Nonetheless, the accumulated evidence leads one to agree with Bastagli et al who conclude that cash transfers “contributed to progress in the selected indicators in the direction intended by policymakers”.

Over the last several years, as these programs have grown, I have wondered if cash transfers from social protection programs deliver differently than what we would see from other sources of income gains. That is, how much does the source of this extra income matter in terms of outcomes? Does it matter to household outcomes if this extra money comes from the government or an NGO, than from a productivity improvement, better farm gate prices, or remittances? 

I have not seen it done, but wondered if looking at cross-sectional data could be suggestive. If cash transfer recipient households spend x% more on food of every $1 they receive (a calculation that Ralston et al do in their meta analysis of studies from Africa), is this in the range of what cross-sectional data show. Likewise for the increase in school enrollment or savings etc. Do cash transfer recipients move along the status-quo local Engel curve? (If anyone has seen work with such calculations, please share it in the comments section, thanks!)…

And if they move differently in our RCTs than what we see in household survey data, why? For example, might this be evidence of the effect of “soft-conditions” (the messaging that sometimes comes with unconditional cash, such as suggesting using the transfer specifically to buy nutritional foods or education supplies), or a response to the effort involved in obtaining income (see below), or something else….

A recent paper brought this idea back to mind last week. Lee, Morduch, Ravindran, and Shonchoy 2024 examine the social meaning of money in Bangladesh, in reference to sociologist Viviana Zelizer’s The Social Meaning of Money, where money is not fungible but gets earmarked depending on how it was obtained. This work remined me of the paper by Christiaensen and Pan 2012 who explore the fungibility of money by income sources in Tanzania and China, drawing on the idea of mental accounting from economist Richard Thaler. Both look at where money comes from, whether the source matters. (A related strand of study, which not my focus here, is the topic of where funds are directed, and flypaper effect and whether ‘money sticks where it hits’).

Arguably Lee et al 2024 do not set out to address the question of whether where money comes affects how it is spent.  Rather they set out to explore whether “the form of money (rather than the functionality of mobile money) change[s] perceptions and choices? Is there something importantly different about holding 20 dollars on your mobile phone rather than holding a 20 dollar banknote in your hand?”  Here I focus instead on another aspect of their paper, that related to the idea of households spending differently because of the income source (all else equal).

They assess this indirectly by conducting a survey to elicit willingness to purchase certain goods, randomizing assignment in whether asked about using mobile money or using cash, with measures of total expenditure, quantity and grams of protein. Their sample includes urban migrant households and migrant-sending rural households. Rural households report a lower willingness to spend (total value and quantity) and lower protein when hypothetically purchasing with mobile money compared to cash. Of course, many things might explain this. For example, rural households might not be able to buy frequent purchases (like food) in local markets using mobile money (which the paper does not address). Or households with less experience using mobile money might show lower willingness to spend it on regular purchases like food (they do not find evidence of this). A third option is whether this is driven by the household’s earmarking or mental accounting by source of income, even when the money is not earmarked per say. The difference in willingness to spend mobile money v. cash is significantly larger for households that received remittances in the past and reported that those remittances were earmarked for non-expenditures (education, health, savings, business, assets, and repayment of debt). This is suggestive that the source of income affects how it is spent.

Christiaensen and Pan more directly study my question of interest, looking at “the coding of income based on the effort involved in obtaining” it. They note that this might matter for the design of social protection: “would potential differences in spending on food and non-food from earned and unearned incomes make social protection measures that provide employment in times of crises be more effective in improving nutritional outcomes than equivalent amounts of cash transfers, which may be easier to administer”. They use panel surveys from Tanzania and China (so they can control for time invariant household heterogeneity) to estimate marginal propensities to consume from earned and unearned income. The latter includes transfers, remittances, gifts, and pensions. The former consists of farm and wage income, as well as business income. To isolate whether results are driven by income being seen as unanticipated/transitory v. permanent, they further look at earned and unearned income decomposed into anticipated and transitory portions.

Households in both Tanzania and China spend relatively more unearned income on less basic consumption goods such as alcohol and tobacco, non-staple food, transportation and communication, and clothing. Earned income spending is relatively higher on staple foods and education.  They hypothesize that people apply emotional tags to income based on the effort in obtaining it, and spend more “wisely” on money they worked harder to get (such as farm income). However, the study can’t do more than speculate to unpack if this is what is going on.

So, these two studies conclude that $1 is not the same as a $1 but depends on the source of the $1. 

Kathleen Beegle

Research Manager and Lead Economist, Human Development, Development Economics

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000