We will perform our scheduled update at 10:30 AM EST on DEC 13, 2017, for 20 minutes.
Distributional preferences over health and income
Last registered on November 17, 2017


Trial Information
General Information
Distributional preferences over health and income
Initial registration date
March 17, 2017
Last updated
November 17, 2017 8:00 AM EST

This section is unavailable to the public. Use the button below to request access to this information.

Request Information
Primary Investigator
Lund University
Other Primary Investigator(s)
PI Affiliation
Lund University
PI Affiliation
Lund University
Additional Trial Information
In development
Start date
End date
Secondary IDs
The Gini coefficient and the concentration index for socio-economic health inequality are two widely used measures of inequality in their respective fields. Both indices implicitly make the value judgment that individuals that are equally far away from the median are weighted equally when the index is calculated. The generalized Gini coefficient (Donaldson and Weymark, 1980) and the extended concentration index (Wagstaff, 2002) both generalize the standard indices so that the value judgment enters as a separable variable that can be adjusted e.g. giving relatively higher weight to the disadvantaged part of the distribution. The aim of our study is to use surveys to elicit this inequality aversion parameter with respect to income inequality and socio-economic inequality in health. Assuming that health is increasing in income, we first elicit the ν parameter of the concentration index of socio-economic health inequality. Since the value of the ν parameter is dependent on the societies used to elicit it, we estimate ν based on two types of societies, one with high average health and relatively little variance and one with lower average health and relatively large variance. We then elicit the δ parameter of the generalized Gini coefficient (S-Gini) of income inequality. Both the ν and the δ parameters capture the relative weight respondents put on different income groups (socio-economic groups) when assessing the level of inequality in a society. To our knowledge, this is the first attempt to empirically assess the δ and ν inequality aversion parameters using individual data. We also estimate the inequality aversion parameter ε of the Atkinson inequality index for inequalities in income and health, again assuming that health is increasing in income. While δ and ν measures preferences for how to weight income groups when the average health/income is constant, ε measures preferences for how to weight income groups by eliciting willingness to exchange efficiency in health/income for equity in health/income.

External Link(s)
Registration Citation
Gerdtham, Ulf, Hjördis Hardardottir and Erik Wengström. 2017. "Distributional preferences over health and income." AEA RCT Registry. November 17. https://www.socialscienceregistry.org/trials/2108/history/23253
Experimental Details
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
Individual values for the δ, ν and ε parameters. Background information on family status, income, health, attitude to redistribution in the health- and income domain as well as political-, risk- and time preferences.
Primary Outcomes (explanation)

Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
Respondents answer to an online survey about their preferences over societies presented to them. All respondents respond to four surveys that each elicit a given inequality aversion parameter. The parameters are: (i) ν of the concentration index for socio-economic health inequality (ii) δ of the generalized Gini coefficient (S-Gini). (iii) ε of the Atkinson inequality index for health inequality and (iv) ε of the Atkinson inequality index for income inequality.

In each survey, the respondent faces a series of tasks where he or she is presented with two (the surveys eliciting ε) or three (the surveys eliciting δ and ν) societies. The different societies represent the effects of a policy change in taxes (δ survey) or the health care sector (ν survey). The respondent has to pick the society he or she thinks has highest welfare. The societies are presented as a distribution of income or health over the group of poor, middle income and rich citizens of the society.
In the surveys eliciting δ and ν, respondents face two questions, each comparing three societies. Each question locates the respondent on a certain interval which is narrowed down by repeating the task. The first question locates respondents as either δ<2.5, 2.5< δ<3 or δ>3 (same for ν). Respondents with δ <2.5 in the first question get a follow up question that places them as either δ<1.5, 1.5< δ<2 or 2<δ<2.5. Respondents with δ>3 in the first question get a follow up question that places them as either 3<δ<3.5, 3.5< δ<4 or δ>4. Respondents that placed on 2.5< δ<3 in the first question get a control question that elicits the same interval as the first question but with different societies.

In the surveys eliciting ε, respondents compare a series of five “B” societies with a fix “A” society. The ”B” societies have the same level of inequality, but are increasing in mean health/income. If and when a respondent switches from choosing the A society to choosing the B society defines an interval for the value of ε of the respondent.

The order in which respondents answer the four surveys is random. In addition, a randomly selected subsample of respondents answer a fifth survey where the ν of the concentration index for socio-economic health inequality is elicited for a society with health levels considerably lower than in the main survey eliciting ν. Finally, the respondents will respond to a number of control- and background questions on family situation, income, political orientation, time- and risk preferences and health.

A first step is to conduct a pilot experiment with a reduced number of respondents (around 100) and a reduced version of the background questions, excluding questions on health. The main survey including a complete list of background questions will be sent to 1000 respondents. The respondents receive a moderate compensation for answering the survey.

Experimental Design Details
Not available
Randomization Method
Randomization by a computer
Randomization Unit
Individual respondent
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
1000 respondents (100 respondents in pilot)
Sample size: planned number of observations
1000 respondents (100 respondents in pilot)
Sample size (or number of clusters) by treatment arms
500 respondents answer two variants of health inequality survey, 500 respondents answer one variant of health inequality survey. (50/50 in pilot)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB Name
IRB Approval Date
IRB Approval Number