Measuring Income Expectations Using Phone Surveys
Expectations about future income are a crucial determinant of jobsearch behaviour. Such beliefs matter both for structural estimation and as outcome measures for active labour market policies. For our work in South Africa (Carranza et al., 2019), we focused on both aspects, and carefully piloted ways of measuring beliefs about future income. The population of our study was comprised of young, disadvantaged jobseekers in urban Johannesburg, most of whom graduated from high school. Crucially, we wanted to measure both the mean and variance of expected income. We also wanted to distinguish between income expectations for formal and informal work, as well as expected income in case of unemployment, as these are important ingredients in structural models.
We elicited income expectations for three mutually exclusive and exhaustive categories, as displayed in Table 1.
Table 1: Work status categories
Formal, permanent. employment 
Any other work 
Unemployment 

Definition 
You are working fulltime in a permanent, formal job with a regular salary and a written contract. 
You are doing any kind of job or work for payment that is not a permanent formal job. This could include temporary or casual work, piece jobs, selfemployment, and family work for pay. You could do more than one of these activities at once. 
You are unemployed: you are not doing any kind of job or work for pay. 
To aggregate expectations across categories, we also asked about the likelihood of each scenario. Specifically, we asked the following questions:
 What is the chance (out of 100%) that you will be in Situation 1 three months from now? That is, you work in a fulltime permanent formal job with a regular salary and a written contract and not doing any other job?
 What is the chance (out of 100%) that you will be in Situation 2 three months from now? That is, you are doing any kind of job or work for payment that is not a permanent formal job. This could include temporary or casual work, piece jobs, selfemployment, and family work for pay.
 What is the chance (out of 100%) that you will be in Situation 3 three months from now? That is, you will not do any kind of work for pay.
We considered restricting the probabilities to sum to 100 but found this to be impractical over the phone. Instead, we restricted the answer to each question to be between 0 and 100 and then reweighted the probabilities to sum to 100 for each respondent.
For the two employment categories, we then elicited income expectations roughly following the methodology detailed by McKenzie et al. (2013). We asked about the mean, maximum and minimum that jobseekers thought they could earn in each situation in the following way:
 What salary do you think you would earn in this situation? (Takehome or after deductions). Include all jobs.
 What is the lowest possible salary you could be earning? (Takehome or after deductions)
 What is the highest possible salary you could be earning? (Takehome or after deductions)
The only restriction imposed was that the maximum earning had to be larger than the mean and the minimum had to be smaller than the mean earnings.
Finally, we asked the following questions to get an estimate of the shape of the CDF.
 What do you think the chance is (out of 100%) that you would earn between R ${belief_minimum} and R ${(belief_minimum +belief_maximum)/2} in this situation?
This question provides further information about the shape of the expected income distribution.
Based on this data, we then fitted a uniform and a triangular distribution for each category using questions 24. Table 2 shows that the uniform distribution led to slightly higher estimates for both mean and variance of the distribution compared to using a triangular distribution. The direct elicitation fell in between the two detailed elicitations. The correlation between the direct elicitation and the extrapolated values is between 0.86 and 0.92. Taken together, the direct elicitation seems to do a relatively good job at capturing the mean earnings expectation. As we only required the variance estimates for structural estimation, we dropped the more detailed questions after 50% of our survey to limit the burden for our respondents.
Table 2: Comparison of different expectation estimates
Uniform distribution 
Triangular distribution 
Direct elicitation 

Expected earnings (sd) for formal, fulltime employment. 
6752 (1530) 
5868 (1250) 
6272 
Expected earnings (sd) for other work. 
3866 (850) 
3375 (694) 
3640 
Carranza, E., Garlick, R., Orkin, K., & Rankin, N. (2019). Job Search and Matching with TwoSided Limited Information. Working Paper.
McKenzie, D., Gibson, J., & Stillman, S. (2013). A land of milk and honey with streets paved with gold: Do emigrants have overoptimistic expectations about incomes abroad? Journal of Development Economics, 102, 116–127.