The role of repeated rejections on job-search behavior: A lab-in-the-field experiment

Subjects are asked to imagine being on a short term employment contract which is due to expire in 10 weeks. The objective of the game is to be ``employed after this contract expires". Subjects can achieve this objective either by finding a new job by ``searching", or being offered a contract extension by ``focusing on their current job". To this end, subjects play a 10 round game (where each round represents one of the remaining ``weeks" on the contract) in which, before each round, they choose between two activities -- one activity is labelled as ``Search for a new job", and the other as ``Focus on your current job to try get a contract extension". Once they have selected an activity, they perform a real effort task which determines their probability of being made an offer. The relationship between effort and the probability of receiving an offer is identical for both activities, and is transparent and explained clearly to players: Players can complete between 0 and 30 tasks in each real effort activity, and their probability of receiving an offer is linearly related to effort as follows -- each completed task increases their chances of receiving an offer by one third of a percentage point. I.e. 15/30 completed tasks = 5% chance of being made an offer.

Subjects are thus aware that their chances of being made an employment offer are identical in the two activities, conditional on the effort they exert. If a subject is made an employment offer, she receives a fixed payment of R55 or R50, depending on the treatment condition (described in Experimental Design below). The crucial difference between the two activities is that in the ``Search" activity the player is informed immediately after each round whether she received an employment offer, while in the ``Focus on current job" activity the player is informed only after all 10 rounds have been completed whether or not she received an employment offer. Because round-by-round offer rates are low (on average about 5%), those who choose to search will be exposed to frequent rejection messages (up to 10 times per game), while those who choose not to search will be informed of the outcome only once per game.

Because the probability of being made an offer is identical across the activities, players' choices should not be influenced by their risk preferences, but only by their rejection sensitivity and the expected returns to each activity (whether R50 or R55). (Comprehension checks are used to ensure that players understand which activity has a higher potential payoff and that the probability of being made an offer is identical across the two activities.) Rational, expected value maximising, rejection insensitive players would only be motivated by expected returns and would always choose to search when the expected value of searching is higher than the expected value of not searching, and vice versa.

Variations of the game - and how they are designed to answer specific questions - are described in "Experimental Design" below.

The game can be viewed and played here: https://www.otreehub.com/projects/rzcapaciti-pilot/

2022-08-15

2022-09-26

1. The percentage of rounds in which players choose to search.
2. Time to switch away from searching

1. The percentage of rounds in which players choose to search.
In each game, players face several decisions of whether to search or not search. This outcome variable is simply the proportion of decisions made to search rather than not search over the full game. This variable does not differentiate when in the game decisions to search are made. Since the experiment requires players to play several games, each individual will have several observations for this outcome variable, each corresponding to a game.

2. Time to switch away from searching: A key question in this research is on the effect of rejection in discouraging job search. This discouragement plays out over time and as players experience repeated rejection. I will therefore use visual and statistical methods, such as survival analysis of recurrent event data, to study how discouragement plays out over time. Time is represented by the "rounds" in each game. Unlike the first outcome, which takes the average of search choices over a full game, for this outcome I consider not only differences across games but also how search behaviour evolves over the course of each game.

I collect data not only on the choices that players make, but also the effort they exert through the real effort task. This allows me to investigate whether discouragement through rejection affects not only search behaviour but also a) how rejection affects effort, and b) whether higher effort players self-select into search or not-search.

The different ``treatments" in the experiment are the five different variations of the base game described above (Intervention section). In each of these variations, different parameters are adjusted to attempt to tease out a) the effect of rejection in disincentivising job search, and b) how varying the type of feedback and/or information which is available to job seekers exacerbates/mitigates the disincentive effects of rejection. The five treatments are described below:

T1. Monetary incentive to not-searching.
In this variation of the base game, players stand to win 10\% more (R55 vs R50) if they ``keep their job" compared to if they ``find a new job" -- i.e. the expected monetary gains are greater if players choose not to search. Rational and expected-value maximising players will choose not to search in every round. The game is played over 10 rounds. The probability of being made a job offer in any given round is between 0\% and 10\%, depending on a players performance in the round (calibrated so the average player's probability of meing made an offer is approximately 5\%). If players choose to search, the rejection message is neutral: ``You were not offered a job".

T2. Monetary incentive to searching, neutral rejection messaging, 10 rounds
In this variation of the base game, players stand to win 10\% more (R55 vs R50) if they ``find a new job" compared to if they ``keep their job" -- i.e. the expected monetary gains are greater if players choose to search. Rational, rejection indifferent and expected-value maximising players will choose to search in every round. The game is played over 10 rounds. The probability of being made a job offer in any given round is between 0\% and 10\%, depending on a players performance in the round (calibrated so the average player's probability of being made an offer is approximately 2.5\%). If players choose to search, the rejection message is neutral: ``You were not offered a job".

T3. Monetary incentive to searching + doubled per-round rejection rate but same cumulative rejection rate as T2
This variation of the game is identical to T2 except that the game is played over 21 rounds rather than 10, and the probability of being made a job offer in any given round is between 0\% and 5\%, depending on a players performance in the round (calibrated so the average player's probability of being made an offer is approximately 2.5\%). By playing the game over more rounds, the cumulative probability of being made an offer in any given round remains the same for T2 (40.1\% @ 5\% per round) and T3 (41.2\% @ 2.5\% per round), but the number of rejections the average player will receive before being made an offer will double. Rejection sensitive players may be discouraged by this increased rejection frequency and switch into the non-search activity.

T4. Monetary incentive to searching + selective positive performance feedback
This variation of the game is identical to T2 except that top performing players (performance in effort task > median effort) who choose to search and are rejected are informed of their strong performance. That is, their rejection message is somewhat kind/encouraging.

T5. Monetary incentive to searching + information on peer outcomes
This variation of the game is identical to T2 except that when players choose to search, they are informed not only of their own outcomes, but also of the search outcomes of all other players in the game. For example, in a game played by 10 players, a player who searches will be informed whether she received a job offer and how many other players were made job offers.

The experiment will use a within-subjects design -- that is, all subjects will receive all treatments. Acknowledging that behaviour in any given round may be subject to framing and sequencing effects, the order in which rounds are administered will be randomised across sessions. In addition, results derived from the full sample using the within-subjects design will be verified by using first-period, cross-subject comparisons (i.e. dropping all observations from rounds after the first).

Randomisation will be undertaken using a random number generator in Stata or Excel.

The order in which treatments (variations of the base game) are administered will be at the experimental session level. Prior to each experimental session (with between 10 and 30 participants) I will randomise the order in which games are arranged.

Yes

10-30 experimental sessions

200-300 participants

200-300 participant observations (using a within-subject design)