The Social Welfare Functions of Elected Politicians

We conduct a survey where participants submit preference-aggregation and paternalistic-intervention decisions as introduced in Ambuehl et al. (2021) and Ambuehl and Bernheim (2021).
In terms of the former, we focus on profiles that allow us to identify the type of scoring rules subjects use and in terms of the latter we compare choice sets involving options with front-end delay to choice sets involving options without front-end delay to understand whether people in particular intervene paternalistically to prevent present biased choices.
As "variation", we compare elected politicians to a population sample collected by a survey company that is similar to the politicians in terms of their socioeconomic charateristics and to a population sample that is representative for the German population.

2021-05-25

2021-07-31

Aggregation decisions - convexity / concavity and Scoring vs. Condortcet, as analyzed in Ambuehl & Bernheim (2021)
Paternalistic interventions - yes/no and strictness of the intervention, measured as maximal amount that can be chosen early + Reaction to front-end delay as analyzed in Ambuehl et al. (2021)
How do beliefs about cardinal utility / disutility of aggregation options correspond to concavity of scoring rule used?
How do beliefs about Chooser's unrestricted choices correspond to the paternalistic interventions imposed?

Delegation decisions (only population sample) - delegation of aggregation/paternalistic decisions to politicians or members of the population with / without the same demographic characteristics as elected politicians.
How does support for real world paternalistic interventions correlate with the laboratory-type measures?

We compare elected politicians to "the people" in terms of the their preference-aggregation and paternalistic-intervention decisions.

Our experimental design consists of 3 main parts - a prior preferences elicitation, the main survey, and the post intertemporal decisions.

**********************************************
Prior preference elicitation:
Using pollfish, we ask survey participants about their opinions wrt to three political organizations in Germany. The three organizations are the Hans-Böckler-Stiftung, the Bund der Steuerzahler Deutschland e. V. and the Ludwig-Erhard-Stiftung which are rather left, center and conservative when it comes to their political stance. In particular, participants are asked to provide an incentivized ranking ranging from the best to the worst of the three organizations. These preferences are used in the main survey to construct the preference profiles of the groups the participants then shall decide for.

**********************************************
Three surveys with politicians and the population:

We conduct three surveys:

Survey 1
Survey 1 is conducted with politicians and a general population sample that is similar to the politicians in terms of their socio-economic status.
The survey consists of three parts:
Part 1.
- Participants make 6 preference-aggregation decisions knowing that one of the 6 decisions may be chosen to be paid out (the chosen organization receives 30 € if the respective decision is randomly chosen to be paid out). People learn that they are confronted with the preference profiles of 5 people who took part in a previous part of the study (see above) and that it is their task to aggregate the respective preferences into a joint group decision only knowing the preferences profiles (and not the identities of the organzations). The profiles are constructed such that the choices enable us to identify the type of scoring rule used by the respective decision maker.
- In addition, we collect participants' beliefs regarding the WTP of people that took part in the preference elicitation study and know the identities of the organizations, that their respective best, middle, worst organization does (not) receive 30 €.

Part 2.
- Participants make 2 decisions where they can choose to make 3 payment options for a future study participant either available or unavailable. Importantly, the 3 options differ in their total payoff and in the relative amount that the future participant receives on the earlier of two payment dates. The design makes sure that it costly to be impatient. The 2 decisions differ in their parameters and in whether the early option is with or without front-end delay.
- Participants provide beliefs in 2 choice sets about how many of 10 people would choose either of 3 options in case all 3 options are available. Again, the two choice sets differ in their parameters and in the presence of front-end delay on the earlier payment date.
- People are asked if they are in favor of forcing the population to save a certain amount for old age in addition to their pension. (on a 7-point likert scale)

Part 3.
- Standard survey questions regarding risk, time and social preferences are elicited (similar to global preference survey)

Survey 2.
Survey 2 is conducted with a general population sample that is representative for the German population.
The survey consists of 4 parts.
Part 1.
Questions on socio-demographic charateristics.
Part 2.
Identical to "Part 1" of "Survey 1"
Part 3.
Participants learn that they can delegate two preference-aggregation decisions similar to the previous part to another person. They can decide whether this person shall be a randomly chosen politician, a person from the population sample that is similar to the politicians in terms of socio-economic charateristics or from the representative population sample.
Part 4.
Participants provide incentivized beliefs regarding the behavior of 10 people of the 3 groups (politicians, similar population sample, representative population sample) in the two aggregation decisions.

Survey 3.
Survey 3 is conducted with a general population sample that is representative for the German population.
The survey consists of 4 parts.
Part 1.
Questions on socio-demographic charateristics.
Part 2.
Identical to "Part 2" of "Survey 1"
Part 3.
Participants learn that they can delegate two paternalistic-intervention decisions similar to the previous part to another person. They can decide whether this person shall be a politician, a person from the population sample that is similar to the politicians in terms of socio-economic charateristics or from the representative sample.
Part 4.
Participants provide incentivized beliefs regarding the behavior of 10 people of the 3 groups (politicians, similar population sample, representative population sample) in the two paternalistic-intervention decisions.

**********************************************
Post intertemporal decisions:
We collect the decisions of the paternalees who are confronted with their possibly restricted choice-sets and are paid out accordingly.

*********************************************
Design.
The pen-and-paper format of the survey for politicians prevents full randomization. Instead, we block-randomize surveys. In total, we have 16 versions of the survey varying the order of the aggregation decisions, the order of the choice options within the aggregation decisions, the order of the paternalistic-intervention decisions and the parameters of the paternalistic-intervention decisions.

********************************************
Hypotheses.
Central questions we want to adress with the data are the following:
1) Do politicians aggregate preferences differently from the general population in terms of concavity / convexity? How?
2) Are politicians more eager to interfere in others' choice options paternalistically?
3) Are politicians more prone to react to 'present bias' in the sense that they react more strongly to the absence of a front-end delay in paternalistic interventions?
4) Are potential differences between politicians and the population due to selection of those who want to become politicians into this role or due to differences in socioeconomic charateristics?
5) Who do the people want to make such decisions, politicians or the people?
6) We plan to analyze potential sources of heterogeneity wrt to our central variables of interest, such as party affiliation/preferences, gender, age, education and income.

Randomization done by a computer

Individual

No

Number of cluster = number of individuals in our surveys (see below)

1800 - 2100 in survey + unknown number of elected politicians

600 - 700 in Survey 1 ("Similar to politicians")
600 - 700 in Survey 2 (Aggregation decisions and delegation)
600 - 700 in Survey 3 (Paternalistic decisions and delegation)
Survey 1 with politicians: survey sent to all federal and state members of parliament (response quota unknown ex-ante)