Willingness-to-Pay for Attributes of High-Profile Jobs

This pre-analysis plan refers to an online choice experiment that allows us to elicit workers' preferences over jobs under exogenous variation in job attributes. Our main contribution relative to previous studies using similar approaches is that we focus on high-profile job offers. A key feature of our design is that we sample workers who actually hold this type of job, or will likely negotiate about a high-profile job offer in the near future. In order to do so, we focus on an important segment of the market of high-profile jobs, namely the market for tenured professorships. In collaboration with the association of professors in Germany (Deutscher Hochschulverband, DHV), we plan to collect data in a sample of high-profile workers. The population of workers to be invited to the survey consists of active associate and full professors (most of them tenured) and (non-tenured) assistant professors, many of whom will likely negotiate about a job offer for a tenured professorship in the near future. The experimental design aims at identifying the willingness-to-pay (WTP) for certain job attributes. The job attributes we study include performance-related pay, the option to negotiate about further pay increases, and mobility requirements (plus other job attributes discussed in the following). Special attention will be given to gender differences in the willingness-to-pay for (avoiding) these attributes.

In many aspects, our experimental design follows Maestas et al. (2018), who use a survey experiment to estimate the WTP of workers for alternative work arrangements and various non-wage characteristics of job offers. The approach is based on the idea of inducing random variation in fictitious job profiles and observing the choices individuals make when facing the tradeoff between these hypothetical job offers with different wage and non-wage characteristics. The resulting data allow us to identify the workers' average willingness to pay for the presence of certain job characteristics.

UPDATE: see PAP

2023-05-11

2024-02-29

Binary variable indicating the subjects' choice between two hypothetical job offers

Before participating in the experiments, each respondent answers a survey about demographics and current job characteristics. We elicit the following:
- current position (Post-Doc, Assistant Professor (W1, no tenure track), Assistant Professor (W1, tenure track), Associate Professor (W2, non-tenured), Associate Professor (W2, tenured), Full Professor (W3, non-tenured), Full Professor (W3, tenured)
- age (if tenured: <40, 40-49, 50-60, >60; if non-tenured: <35, 35-39, 40-44, >44)
- gender
- children of primary school age, or younger, who need some form of care during workdays (yes/no)
- taking into account private and family situation: flexibility in choosing a place of residence (Likert scale from 1 (very unflexible) to 7 (very flexible))
- current workplace in daily commuting distance from main place of residence (yes/no)
- federal state (current position)
- discipline (social sciences, law, natural sciences, engineering, economics and business, medicine)
- if non-tenured: how well informed about negotiations for a professorship and the topics typically raised in such negotiations (Likert scale from 1 (very poorly) to 7 (very well))
- if non-tenured: with how many people in touch regularly regarding academic career, negotiations, and other related topics (nobody, one person, two people, ..., 5 people, more than 5 people)
- if tenured: performance-related bonus in current position (yes/no)
- if tenured: number of past negotiations for a professorship (1, 2, 3, more than 3)

After the survey, we administer a series of ten stated-preference experiments to each survey respondent. In each of these experiments, survey respondents are asked to select between two job offers, each defined by a partially varying set of non-wage job characteristics and the job's monetary compensation. To minimize the risk of differential perceptions regarding unspecified job characteristics, we instruct respondents to assume that any job attributes not mentioned are identical across offers.

The job offers' monetary compensations comprise two components. The first component is a fixed base pay that is given by the regulations regarding the compensation of tenured professors in the respective federal state. For a given participant, this base pay in the experiment does not vary between job offers and is the same across all 10 experiments. The second component is the bonus. We leverage this bonus to induce random variation in monetary compensations. Using a discipline-specific mean bonus, m, the random variation in the bonus is achieved by setting the bonuses of Offer A and Offer B as c_A*m and c_B*m, respectively, where c_A and c_B follow a N(1, 0.075) distribution. We truncate both weights to lie between 0.5 and 1.5 and round the bonus values to full Euro amounts.

The offers' non-wage characteristics vary freely. We consider the following characteristics:
- Mobility requirements, measured by whether or not the job's location is within commuting distance of the preferred place of residence for the respondent and her family
- Academic reputation, measured by whether or not the university offering the job has the status of an "Exzellenz-Universität" in the German system of higher education
- Child care options, measured by whether or not the university offers guaranteed placement in a child care facility
- Share of women among professors at the university department offering the job (10%, 25%, or 40%)
- Performance-related pay, measured by whether or not the job features a bonus that is contingent on the job holder reaching certain pre-defined goals
- Option to negotiate further pay increases, measured by whether or not there is an option to negotiate about a further bonus after three years

When creating hypothetical Offers A and B, we randomly select two of these non-wage attributes to vary across the two offers (in addition to the monetary compensation, which always varies between offers). Within each of the two randomly selected attributes, we choose corresponding attribute values at random sequentially for both offers without replacement. This makes sure that Offer A and Offer B actually vary in the selected attributes. We adapt the strategy used by Maestas et al. (2018) to limit the number of job pairs in which one of the jobs dominates the other on all varying dimensions.

In addition to the 10 choice experiments, we include one further survey question that serves as an attention check. When facing this question, which appears randomly between the fourth and the last choice experiment, respondents are instructed to respond in a specific way (mark two specific options from a choice menu), irrespective of what they believe is the true answer to the respective question. The attention check question allows us to estimate the share of inattentive participants and test the robustness of our findings with respect to excluding inattentive respondents.

In terms of implementation, in each experiment we display the hypothetical job offers with all characteristics side by side. We instruct respondents to either select ``Prefer Offer A,'' or ``Prefer Offer B.'' Each respondent makes the binary decision between Offer A and Offer B in 10 distinct sequential experiments.

UPDATE: see PAP

Not available

Randomization by a computer

Individual choice between two hypothetical job offers

No

We will invite about 23.800 subjects to participate in the survey experiment. We hope to be able to collect data from at least 2400 subjects. Each subject will make 10 choices between pairs of fictitious job offers.

UPDATE: see PAP

We will invite about 23.800 subjects to participate in the survey experiment. We hope to be able to collect data from at least 2400 subjects. Each subject will make 10 choices between pairs of fictitious job offers. UPDATE: see PAP

There are no treatment arms in our design, but the job offers' characteristics vary randomly in each individual choice experiment.

We do not have any baseline data and thus cannot provide minimum detectable effect sizes. However, assuming that our key variables are distributed similarly as the key variables in Maestas et al. (2018), we are confident that we will be able to detect relatively small effects. For instance, Maestas et al. (2018) estimate that a switch from a fast-paced to a relaxed job (holding all other job characteristics constant) is equivalent to a 4.4 percent wage increase. Similarly, they estimate that the option to telecommute (without a differentiation on how intensely this option may be used by the worker) is equivalent to a 4.1 percent wage increase. These estimates are based on a sample of 1,815 survey respondents and significant at the one percent level. Assuming that we will be able to collect data from at least 2,400 respondents, we expect to be able to identify wage premia for our key attributes (performance-dependent pay, option to negotiate further pay increase, and mobility requirements) of less than 4 percent. UPDATE: see PAP