The Perceived Government Budget Constraint

How to cite

Nakata, Taisuke; Fueki, Takuji; Hino, Masashi; Katagari, Mitsuru; Katayama, Munechika. 2025. "The Perceived Government Budget Constraint." AEA RCT Registry. AEARCTR-0017458. https://doi.org/10.1257/rct.17458-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Taisuke Nakata* , Takuji Fueki , Masashi Hino , Mitsuru Katagari , Munechika Katayama

* Primary investigator

Registry ID
AEARCTR-0017458
DOI
https://doi.org/10.1257/rct.17458-1.0
First registered
Last updated
Status
Completed
Trial dates
2025-12-19 – 2025-12-25
Countries
Japan ()
Registry record
http://www.socialscienceregistry.org/trials/17458

Abstract

The intertemporal government budget constraint (IGBC) plays a key role in economic theory and policy debate. In this paper, we examine the extent to which the public understands this constraint. For that purpose, we ask households: (i) the expected change in the government debt, (ii) the expected real GDP growth rate, (iii) the expected nominal interest rate, (iv) the expected inflation rate, and (v) the expected primary balance. From the answers to these questions, we can compute the left-hand and right-hand sides of the IGBC and examine their equality. We also provide several pieces of information that may affect left- and right-hand sides of the IGBC and examine whether households revise their forecasts in response to provided information in a manner that is consistent with the IGBC.

Experimental design

Our experiment will be conducted through an online survey. Our main survey will consist of four parts: (i) an attention check, (ii) a pre-treatment block, (iii) treatment, and (iv) a post-treatment block.

1. Setting and Sample Size

We plan to conduct an online experiment in Japan using a marketing company, MyVoice Communications, Inc. We will recruit respondents from their panel. The sample will be approximately representative of the Japanese population in terms of gender and age categories. We plan to collect data from 800 respondents for each of the six groups (five treatment groups and one control group). The total sample size will be 4800 (=800*6).

2. Attention Check

Before asking the main survey questions, we will include an attention check question to ensure that respondents are paying close attention to our survey. This is intended to improve the quality of our data by filtering out those who may not be fully engaged with the survey questions.

To this end, we will ask respondents to select a specific option from a list of choices regardless of their actual opinion. In particular, we will ask them to choose ``Very interested'' to the following question:

``How interested are you in the issues in the Japanese Economy?''

The options will be as follows: (i) Strongly interested, (ii) Very interested, (iii) Somewhat interested, (iv) Not very interested, and (v) Not interested at all. We will exclude respondents who fail to select ``Very interested'' from our survey.

3. Pre-treatment Block

Our pre-treatment block will have two blocks of questions. The first block (perceptions) will ask respondents about their perceptions of the Japanese economy in the past. The second block (prior predictions) will ask respondents about their predictions for the Japanese economy in the future.

In the first block, we will ask respondents about their perceptions of the following six macroeconomic variables in Japan:
1. Average annual real GDP growth rate from 2015 to 2024
2. Average annual nominal interest rate on 10-year government bonds from 2015 to 2024
3. Average annual inflation rate from 2015 to 2024
4. Average primary balance to GDP ratio from 2015 to 2024
5. Government debt to GDP ratio in 2024
6. Average fraction of old people (aged 65 and over) in the total population from 2015 to 2024

For each variable, we will provide a brief explanation of the variable and anchoring values.

Anchoring is intended to help respondents understand the scale of each variable and to reduce the cognitive burden of answering the questions. This helps us to reduce measurement error in our survey responses. Anchoring values will be based on the actual values of each variable in the past. Except for the government debt-to-GDP ratio, we will provide the average value of each variable from 2005 to 2014 as the anchoring value. For the government debt to GDP ratio, we will provide the value in 2014 as the anchoring value.

In the second block, we will ask respondents about their predictions of the same six macroeconomic variables in Japan in the future. Specifically, we will ask them about their predictions of average values for the next ten years (2025-2034) for the real GDP growth rate, nominal interest rate on 10-year government bonds, inflation rate, and primary balance to GDP ratio. For the government debt to GDP ratio and the fraction of old people in the total population, we will ask them about their predictions for the year 2034.

4. Treatment

We have five treatment groups and one control group. The respondents will be assigned to one of the six groups. The treatment groups will receive different information related to the government budget constraint. All groups will be reminded of their answers in the perception block of the pre-treatment survey. The control group will not receive any additional information.

The first treatment group (T1) will receive information on the government debt-to-GDP ratio in Japan. The second treatment group (T2) will receive information on the real GDP growth rate. The third treatment group (T3) will receive information about the nominal long-term interest rate. The fourth treatment group (T4) will receive information about the inflation rate. The fifth treatment group (T5) will receive information on the primary balance to GDP ratio. Except for T1, respondents also receive projections of each variable for the next ten years (2025-2034). The projections are mainly based on one of the scenarios projected by the Cabinet Office of Japan. For the inflation rate, we use the value based on the household survey conducted by the Bank of Japan. They are shown as figures together with historical averages from 2015 to 2024.

We do not provide a forecast for the government debt-to-GDP ratio because it is known that the magnitude of expected changes in the government debt-to-GDP ratio among survey respondents tends to be much larger than reasonable forecasts from the Cabinet Office of Japan (e.g., Fueki, Hino, Katayama, and Nakata, 2025) and that providing such a forecast may lead to confusion among respondents. Thus, we only provide the actual value in 2024.

5. Post-treatment Block

Our post-treatment block will have two blocks of questions. The first block (posterior predictions) will ask respondents again about their predictions for the Japanese economy in the future. The second block (demographic questions) will ask respondents about their demographic characteristics.

The first block will be identical to the second block of the pre-treatment survey. This allows us to measure how respondents update their predictions after receiving the treatment information.

A common concern in asking the same question again without rephrasing the question is the experimenter demand effect. However, this concern is less likely to be important in our study. Unlike in a common information provision experiment, we are not interested in whether a certain piece of information affects people's behaviors, beliefs, or opinions per se. Rather, we are interested in whether there is consistency in their responses from the perspective of an economic theory. Thus, it is highly unlikely that an ordinary person can tell our intention.

The second block will ask respondents about their demographic characteristics---such as gender, age, and the number of children---socio-economic characteristics---occupation, education, household income, borrowings, and political stance, etc. These variables will be coded as follows:

-We code gender as a dummy variable that takes one for female respondents.
-We code age as a continuous variable as well as a set of dummy variables for age groups.
-If the respondent declares to have one or more children, a dummy variable is set to one.
-We code employment status as a dummy variable which takes one if they are employed. It takes zero when the respondent claims that s/he is either one of the following classifications: students, housewives/househusbands, or not employed or retired.
-We code education as a dummy for whether the respondent has at least a Bachelor's degree.
-We code household income as the log of mean income in each interval specified by the respondent.
-We create a dummy variable for whether the respondent has any borrowing. We also create a set of dummy variables for whether the respondent has a particular type of borrowing.
-We create two dummy variables for the left and the right in terms of political views. The former takes 1 if the respondent chooses values from 1 to 4 on the Likert scale ranging from 1 to 10. The latter takes 1 if the respondent identifies her/himself from 7 to 10 on the same Likert scale. We exclude political centrists (5 and 6 on the Likert scale).
-We create a discrete variable for the respondents' view toward the government policy during the COVID-19 crisis.

We code perceptions about the current debt-to-GDP ratio as continuous variables. We winsorize the data based on perceptions to make sure that extreme outliers do not drive our results.

Throughout our analysis, we winsorize our variables to deal with outliers.

Design details

Randomization method
Randomization done by a computer
Randomization unit
Individual
Clusters
4800 clusters
Observations
4800 individuals
Arms
800 for the control group; 800 for each of the five treatment groups.
Primary outcomes
The discrepancy between left-hand and right-hand sides of the intertemporal government budget constraint implied by households' forecasts of (i) the change in the government debt, (ii) the real GDP growth rate, (iii) the nominal interest rate, (iv) the inflation rate, and (v) the primary balance.

Cited by

No citation data available for this registration yet.

Keywords