Trust in politicians and the provision of public goods

Evidence from Germany

Providing public goods is a key role of the state. Politicians are in charge of infrastructure investments, protection against climate change, defense spending, reserach investments and public education spending. However since the benfits of public goods are often not directly clear to voters and projects have a long time horizon, provision of public goods requires trust of voters. When voters distrust politicians, they often less in favor of increased spending on public goods, as the figure below shows.

The left panel shows voters' preferences over the provision of different types of public goods in relation to trust in politicians. Control variables (generalized trust, age, assessment of current economic situation, outlook, and past trend, as well as fixed effects for income brackets, school education, job qualifications and gender.) are included. The right panel shows the share of sentences about different types of public goods in party manifestos in relation to the share of sentences about corruption.

When voters trust politicians they are more in favor of providing more social benefits even if this means tax increases and more in favor of protecting the environment, even if this harms the economy. How do parties and politicians respond to this?

The right panel suggests that parties that talk more about ocrruption in their manifestos (possibly those more concerned with voters’ trust), are less likely to mention technology, environmental issues, culture and education spending in their manifestos – an indication that parties are responding to voters’ preferences.

In this project, I study in detail how politicians respond to low trust voters. I use data from Germany, which similar to many other countries has seen a decrease in trust in politicians recently.

The left panel shows trust in politicians in Germany between 2009 and 2021, seperated by survey respondents in East and West Germany. The right panel shows the regional distribution of average share of low trust voters on electoral district level.

To identify for each politician, how much they work on on topics related to public goods, I collect all written legislative documents from the German parliament (see also my blogpost on BundestagsAPy, a python wrapper for the offical API) and use text analysis to identify documents about public goods. The figure below gives some examples of the words used in documents about e.g. environmental protection.

The figure shows the most common words in 5 topics that I identify as being about public goods.

With this data, I can estimate the probability that a politician who got elected in a low trust district is authoring a document about e.g. environmental protection. The figure below shows the results (note that the estimates are from a 2SLS regression, where the instrument leverages historical trust levels and a type of shift-share structure, or details, see the paper).

The figure shows the results from a 2SLS estimation of the probability that a politicians from a district with given trust authors a document about the indicated topic.

For almost all public goods topics, politicians from high trust districts are significantly more likely to be working on them, than politicians from low trust districts. The result is fully driven by competitive districts and also holds when considering voting decisions instead of legilative writing.

In the paper, I also study the mechanism behind this finding. Using a unique survey of political candidates, I show that my results are mostly driven by the selection of politicians who are intrinsically less in favor of providing public goods, rather than pandering to voters.

Read the full paper.