In a previous post, we discussed the possibility that water scarcity might affect rural producers’ income through means other than simple agricultural productivity. With not enough water to grow corn or beans, we conjectured, perhaps producers’ “mental bandwidth” – attention, executive memory and impulse-control, all fundamental inputs to decision-making – is captured by scarcity, leaving little room for other issues. This follows the logic highlighted by recent research.
An ongoing study joint with Anandi Mani of Warwick University documented that this is precisely the case in the Northeast of Brazil. Surveying 700 farmers without access to irrigation in Ceará, the driest State in the country, the study found that farmers primed about low rainfall outperform others with respect to water-related tasks, but display systematically lower performance in every other task. Using the parlance of this literature, farmers were shown to “tunnel”: water, which fell inside the scarcity tunnel, received increased mental bandwidth, but that extra bandwidth was taken from whatever fell outside this “tunnel.”
How do we claim to know that? With the help of MGov Brasil, we used cell-phones to run cognitive tests. Cell-phones are a unique feature of this study because they allow us to follow individuals over time at high frequency (and at low cost), crucial to documenting precisely the source of the effects (if any) we have described. To give a concrete example, in a study with sugar-cane farmers in India, researchers documented dramatic changes in farmers’ IQ after the harvest – when farmers are “rich” – compared to before the harvest – when farmers are “poor”. However, because “psychological” money-scarcity moves together with “physical” money-scarcity in that setting, we cannot be sure that this effect is not driven by differential nutrition (farmers can eat more calories after the harvest, when they have cash-in-hand), for instance, which might lead to differential performance in cognitive tasks in ways that have nothing to do with the scarcity mindset. If those tests had been run two days apart from each other – say right before and right after the farmer sells their harvest – that alternative story would seem much less plausible.
In our case, we primed certain farmers about water scarcity over SMS and ran cognitive tests only 2 days later. Within that time span, one would find it unlikely that farmers’ expectations would have led them to eat less (or to undertake any other expectation-driven decisions), with immediate effects on mental bandwidth.
We ran such tests with farmers located in two regions of Ceará. Within each of these regions, we disclosed precipitation forecasts over SMS to 50% of the farmers enrolled in the survey, while the remaining 50% received a neutral message about land-preparation techniques.
We also designed a truth-revelation mechanism to have farmers disclose their subjective rainfall forecast. At baseline, farmers were asked to provide their forecast (“Normal”, “Below normal” or “Above normal”), and were informed that, if their forecast proved to be accurate by the end of May, they would be given USD $5 in airtime credit. Farmers were given the opportunity to update their forecast after receiving the SMS.
Within each region, only 40% of farmers agreed with the precipitation forecast. While we will only know who was right by the end of May, the incredibly high dispersion of expectations among very homogeneous farmers is potentially very costly: “over-optimistic” (“over-pessimistic”) farmers, those who expect normal (below-normal) rainfall in a below-normal (normal) forecast municipality, are 6 percentage-points more (less) likely to purchase fertilizer compared to “realistic” farmers in the same region. As farmers fail to take advantage of profitable opportunities or undertake costly investments with low expected return, making sure accurate information reaches them would be crucial.
Following administration of the cognitive tests, those who received the SMS with a positive outlook performed better in all tests compared to individuals in the same region who received the neutral message. Conversely, those who received the SMS with a negative outlook did no worse in any test, compared to individuals in the same region who received the neutral message – perhaps because most farmers might already “tunnel” on water to some extent – but do significantly better in remembering the content of the SMS, and in finding the word “water” among garbled words (one of the cognitive tests).
Because the content of the SMS was randomly assigned, we can say that the differences in performance are caused by the content. While one might conjecture this to be the effect of good vs. bad news in general – hence, not specific to water scarcity – the improved performance of the latter group in finding the word “water” in a series of garbled words is exactly in line with the theory of tunneling, and is not consistent with this alternative interpretation.
Most importantly, the table bleow documents that those who choose to update their forecast do not drive the results (we also find that the latter are not driven by those who remember receiving the SMS). This fact – that scarcity captures unconscious, rather than conscious attention, executive memory and impulse-control – opens a promising avenue for documenting its effects on income-relevant decisions, separately from expectations. Put another way, if cognitive effects worked exclusively through conscious processes, then how could one separate (A) distortions brought about by captured mental bandwidth from (B) expectation-driven decisions guided by anticipation of lower future returns?
This is why the “bottom-up” result backs up the project’s next step: the move from cognitive to income effects. In doing so, we are motivated by the more challenging – but more fundamental – question: does facing scarcity (water-related, or else) cause poverty?
Anecdotal evidence suggests this is likely to be true: almost 50% of farmers in our sample fail to enroll in Government’s insurance program, which costs only USD 6 and disburses 5 installments of USD 80 on the event that at least half of municipality’s harvest is lost due to inclement weather. While traditional explanations would attribute low take-up to farmers being credit-constrained or present-biased, several farmers point out they failed to sign-up simply because they forgot. This should not be surprising after one acknowledges that their attention is captured by water scarcity. Moreover, simple and low-cost policies, such as SMS reminders, might be able to “get into the tunnel”, limiting the negative consequences of captured mental bandwidth.
We hope to shed further light on these and other issues as the project moves forward.
Notes: (1) “Digit span” is a memory test, in which respondents had to correctly type a 5-digit number right after hearing it; “Stroop” is an attention test, in which respondents have to type the number of times they heard a specific number in a sequence; the “penalized” versions adjust respondents’ scores by the length of the interval they took to answer it; “Remembers SMS” is a memory test which assesses whether respondents remember the content of the SMS received; and “Garbled words” is an attention test, in which respondents must indicate whether they heard the word “husband” (a neutral word) or “water” in a sequence of four metalized-speech words, where respondent’s score is given by the differential performance in the water-related task versus the neutral task. (2) Regression controls are gender, age, schooling, crops, animals, whether individuals believe in a rule of thumb about rainfall determination, enrolee status in Bolsa-Família (Brazil´s flagship conditional cash transfer program) and Garantia-Safra (national inclement weather insurance program), income bracket, ownership of checking and savings account, plot property rights, ownership of plot-related assets, and access to irrigation. (3) The number of observations in each column is the intersection of the number of individuals who answered all questions for the control variables and completed the cognitive test in this pilot wave.
Image Credit: Mark Shellhase via Wikimedia Commons.
