WORKING PAPER Table 8. Recall of product information in the joint and separate offer groups All offers Joint offers Separate offers p-value Responded incorrectly to survey question “did you buy 29.9% 24.7% 36.2% 0.122 the crop insurance?” Don’t remember premium 70.1% 64.7% 76.8% 0.104 charged Incorrect about covered 68.8% 69.4% 68.1% 0.864 events Loan applicants receiving both the joint and separate offers showed limited recall of the details when surveyed over the phone. Table 8 summarizes the accuracy of respondents’ survey responses on product information in the joint and separate offer groups. Overall, nearly 70 percent of clients did not recall the premium charged for the insurance28 and a similar proportion made errors in describing the covered events. The differences in recall of product information between the two groups are not statistically significant, either when analyzed with a t-test (as shown in Table 29 5.5) or in a regression framework. The averages, however, suggest that people who were assigned to the separate offer group may have recalled less than those receiving the joint offer. Because clients were randomly assigned into the two groups, any differences in recall at the time of the survey should reflect differences in initial understanding (rather than differences in the amount of information forgotten between the offer and the survey). This runs contrary to our hypothesis that the joint offer, through distraction Crezcamos borrower shows his coffee seedlings to and “information overload” associated with the loan researchers. process, might result in clients understanding and retaining less information, and suggests that bundling the offer of several financial products may exploit a moment of attention to a specific activity, in this case, 30 their crop investment. 28. It is likely that many of the remaining 30 percent of respondents, who gave some cost in response to this question, did not recall the exact price (though the price offered could not be verified for those who chose not to purchase). 29. Including standard errors clustered at the branch level. 30. Clients who received insurance in a joint offer may also have been more likely to be reminded of the insurance by promotional materials at the Crezcamos branch when they visited for loan disbursement shortly after the loan application. Separate offer clients also would have been exposed to the same materials, but before their insurance offer when it was likely less salient to them. Those clients may or may not have revisited the branch between the insurance offer and the survey. 26