Stefan Gerlach, 06 May 2016

Financial markets are increasingly concerned about the outcome of the upcoming Brexit referendum, and considerable attention is therefore focused on surveys of voting intentions. Using a Financial Times dataset covering 201 surveys conducted over the past five years, this column reveals that we can learn surprisingly little from these surveys. While in general they predict the vote will be in favour of remaining in the EU, the organisation that conducted each survey seems to be as important as respondents’ voting intentions in determining individual survey results. Moreover, there is a large number of undecided voters who are likely to decide the outcome of the referendum.

Margherita Comola, Marcel Fafchamps, 08 December 2015

Dyadic social network data – describing relations between two actors – are frequently derived from self-reporting surveys. This column explores how the misreporting problems that are typical of such data can bias estimations. Data on transfers between households in a Tanzanian village are shown to display a high rate of discrepancies within dyads. Failure to account for such misreporting results in a sizeable underestimation of inter-household transfers. 

Prashant Bharadwaj, Mallesh M. Pai, Agne Suziedelyte, 03 July 2015

Fear of stigmatisation might lead to hiding of behaviours and actions. This column presents evidence that stigma concerns can play a role in health seeking behaviour in the case of mental health. In particular, survey respondents under-report mental health conditions 36% of the time when asked about mental health conditions, and about 20% of the time when asked about prescription drug use. Conditions such as diabetes and cardiovascular diseases are less likely to be under-reported.

Daniel J. Benjamin, Samantha Cunningham, Ori Heffetz, Miles Kimball , Nichole Szembrot, 02 January 2015

There is growing interest in alternative measures of national wellbeing, such as happiness or life satisfaction. This column argues that a small number of survey questions are unlikely to capture all the aspects of wellbeing that matter to people. Using a stated-preference survey, the authors find several aspects of wellbeing to be important that are not commonly included in wellbeing surveys, such as those related to family, values, and security. This approach could be used to provide weights for wellbeing indices.

Maxim Pinkovskiy, Xavier Sala-i-Martin, 27 February 2014

How many people are poor worldwide? This column explains that the answer depends on whether one uses survey of national-accounts data to anchor country distributions of income. It then argues that night-time lights suggest that national accounts offer a better estimate. Developing world poverty may be as low as 4.5% in 2010, much lower than the path constructed by surveys.

Joachim De Weerdt, Kathleen Beegle, Jed Friedman, John Gibson, 18 February 2014

Whereas the Millennium Development Goal of reducing extreme poverty by half was achieved by 2010, the global hunger rate has only fallen by a third since 1990. Differences in survey design may account for part of this discrepancy. This column presents the results of a recent experiment in which households were randomly assigned to different survey designs. These different designs yield vastly different hunger estimates, ranging from 19% to 68% of the population being hungry.