What it is
The Niederle-Vesterlund (NV) method measures willingness to compete — the preference for being evaluated against others rather than on an absolute basis — using a real-effort task and a choice between two payment schemes (Niederle & Vesterlund, 2007). In the key round, respondents choose whether their output will be paid at a fixed rate per unit (piece rate) or only if they outperform every other member of their group (tournament). Because the same respondents first complete the task under both schemes and separately report risk preferences, differences in tournament entry reveal differences in competitive preference rather than differences in ability, expected earnings, or risk aversion.
The core finding from Niederle and Vesterlund (2007) is that men enter tournaments at roughly twice the rate of women, even after controlling for performance — and that both groups deviate from the optimal entry rate in opposite directions. Men with below-median performance enter at rates that are too high; women with above-median performance enter at rates that are too low.
When to use it
Use the NV method when the research question concerns competitive preferences directly, or when competition preference is a plausible mechanism or moderator. The method is well-suited for:
- Gender gap research: Testing whether gender gaps in earnings, occupational choice, or leadership attainment are partly explained by differential willingness to compete rather than ability differences alone.
- Programme evaluation: Measuring whether an intervention — mentoring, soft skills training, confidence-building — shifts competitive preferences as a mechanism. Buser, Niederle and Oosterbeek (2014) show that competitiveness measured at age 15 predicts gender-differentiated track choices in Dutch secondary schools, establishing competitive preference as a consequential outcome.
- Labour market and field settings: Studies of entry into competitive job markets, promotion decisions, and educational streaming. Berge et al. (2015) apply the method among small firm owners in Tanzania, demonstrating feasibility in low-income field contexts.
- Incentive structure studies: Evaluating how changing payment schemes — from piece rate to tournament — affects effort, selection, and productivity.
The method is not appropriate when respondents cannot be grouped for tournament comparisons, when the real-effort task has large pre-existing gender skill gaps, or when real monetary stakes cannot be paid.
How it works
The standard protocol has four rounds. All rounds use the same real-effort task — typically mental addition, counting, letter-encoding, or a slider task. The task must be gender-neutral in ability and completable within a short fixed window (two to five minutes).
Round 1 — Piece rate (baseline performance). Respondents complete the task for a fixed time window. Each correct unit earns a fixed piece-rate payment. Performance here measures individual ability, uncontaminated by competitive incentives.
Round 2 — Tournament. Respondents complete the task again. Payment is now winner-takes-all within a group of N respondents (standard: N = 4). The winner — the respondent with the highest output — receives N times the per-unit piece rate for each correct unit. For a group of four, each correct unit earns four times the piece-rate payment; all others earn zero. When two respondents produce identical output, a random tiebreak determines the winner. This rule must be pre-specified and communicated to respondents before the session begins.
Round 3 — Entry choice (the key measure). Respondents complete the task a third time and, before starting, choose whether to be paid at the piece rate or under the tournament scheme. This choice is the measure of competitive preference.
Round 4 — Retroactive tournament entry. Respondents can choose to re-submit their Round 1 performance under the tournament — no new output is generated. This separates the performance decision from the entry decision: because the respondent’s output is already fixed and known, Round 4 entry measures competitive preference unconfounded by effort incentives or uncertainty about one’s own output.
Payment is made for one randomly selected round at the end of the session. The main analysis regresses Round 3 entry (binary: 1 = chose tournament) on gender, controlling for Round 1 performance, Round 2 performance, and a separately elicited risk aversion measure. The residual gender coefficient is the clean estimate of the gender gap in competitive preference. A separate decomposition regression adds the elicited rank belief (see “Overconfidence elicitation” in Key Decisions) to decompose the gap into a component explained by differential beliefs and a residual preference component.
Key decisions
Task selection. Pilot the task before the main survey and check for gender performance differences in Round 1. If a gap exists, the entry decision in Round 3 partly reflects rational responses to different expected earnings rather than pure preference. Avoid tasks that draw on culturally gendered skills.
Group size. Standard is four. Smaller groups reduce stakes and statistical power; larger groups introduce strategic complexity about the likely ability distribution of the group.
Payment calibration. The tournament prize must be economically meaningful. A useful heuristic: set the tournament prize at one to two hours of local unskilled wage. Piece-rate payment should be set so that the expected value of the tournament equals the piece-rate payment for a respondent of median ability, ensuring entry is not driven purely by expected-value maximisation.
Field adaptations. When respondents cannot be simultaneously present for grouping, use a reference distribution design: respondents are told they will be compared to four randomly selected participants from a previous session or pilot study. The reference group performance distribution is determined in advance. Respondents must be told clearly that the comparison group is drawn from real participants to preserve incentive compatibility. Disclose summary statistics of the reference group’s performance (mean and range of correct answers) so respondents can form a meaningful expectation of winning — without this, the tournament entry decision is noise rather than a preference measure.
Overconfidence elicitation. After Round 1, ask respondents to estimate their rank in a group of four. This allows testing whether entry decisions are justified by actual performance and whether overconfidence differs by gender. Including this measure allows decomposing the gender gap in entry into preference vs. belief components.
Caveats & common mistakes
Omitting risk aversion measurement. This is the most common error. Entering a tournament is risky — losers receive nothing. If women are more risk-averse than men, lower female tournament entry may reflect risk aversion rather than competitive preference. The two mechanisms have very different policy implications. Include a separate risk elicitation (e.g., Holt-Laury MPL) and include it as a covariate in all main regressions.
Using hypothetical stakes. The method is only valid with real incentives. Hypothetical tournament entry is confounded by social desirability bias and experimenter demand effects. If budget constraints prevent paying all respondents, use a random-payment design: pay one randomly selected round for real.
Ignoring the learning effect between rounds. Performance typically improves from Round 1 to Round 2. If learning is differential by gender, Round 2 performance will not fully separate ability from learning. Include both Round 1 and Round 2 performance as separate controls rather than averaging them.
Confusing the entry decision with the effort decision. Round 3 entry reflects both competitive preference and any effect of incentives on effort. Round 4 retroactive entry controls for this: because no new effort is possible, it is a cleaner preference measure. Report both Round 3 and Round 4 entry rates in all analyses.
Small within-group samples. Random group assignment with a small total sample can produce groups that are all high- or all low-performers, inflating or deflating entry rates. Pre-stratifying groups on Round 1 performance produces more stable incentives across groups.
Analysis Guide
import pandas as pd
import numpy as np
import statsmodels.formula.api as smf
# r1_output = correct units, Round 1 (piece rate)
# r2_output = correct units, Round 2 (tournament)
# r3_entry = 1 if chose tournament in Round 3, 0 = piece rate
# r4_entry = 1 if chose retroactive tournament in Round 4
# female = 1 if female
# switch_row = Holt-Laury switching row (higher = more risk averse)
# see the Risk Preference — Holt-Laury MPL guide for construction
# 1. Summary statistics by gender — entry rates around 73% (men) and 35% (women)
# in the original NV (2007) sample; deviations suggest a less competitive sample
df.groupby('female')[['r3_entry', 'r4_entry', 'r1_output', 'r2_output']].agg(
['mean', 'std', 'count'])
# 2. Main regression: tournament entry on gender, performance, risk aversion
# LPM with HC1 robust SEs is standard in development economics; for the
# original NV (2007) probit specification use smf.probit(...) instead
smf.ols('r3_entry ~ female + r1_output + r2_output + switch_row',
data=df).fit(cov_type='HC1').summary()
# Probit (original NV 2007 specification) — uncomment if needed
# smf.probit('r3_entry ~ female + r1_output + r2_output + switch_row',
# data=df).fit(cov_type='HC1').summary()
# 3. Overconfidence check: entry by above/below median performance x gender
# The interaction tests whether men and women with the same performance
# level enter at different rates — isolates preference from ability
df['above_median'] = (df['r1_output'] > df['r1_output'].median()).astype(int)
smf.ols('r3_entry ~ female * above_median + switch_row',
data=df).fit(cov_type='HC1').summary()Reading the output
- In the original Niederle and Vesterlund (2007) sample, Round 3 tournament entry rates were approximately 73% for men and 35% for women. If your sample shows a gender gap of similar magnitude after controlling for Round 1 performance and risk aversion, you are replicating the core result. A smaller gap suggests either a less competitive sample or that your controls are absorbing more variation.
- The coefficient on
femalein the main regression is the gender gap in tournament entry after conditioning on performance (Round 1 and Round 2 output) and risk aversion (switch_row). If this coefficient remains large and significant, the gap cannot be explained by ability or risk aversion alone — it reflects a preference for competition. - In the overconfidence check, a positive coefficient on
female#above_medianmeans that above-median women enter tournaments at a higher rate than the gender coefficient alone would predict — they are adjusting for ability. The interaction tests whether men and women with the same performance level enter at different rates. r4_entry(retroactive entry) is a cleaner preference measure thanr3_entrybecause no new effort is involved. If the gender gap inr4_entryis similar to the gap inr3_entry, the difference is not driven by effort incentives. If the gap inr4_entryis smaller, some of the Round 3 gap may reflect differential effort responses to competitive incentives rather than pure preference.
References
Berge, L. I. O., Bjorvatn, K., Pires, A. J. G., & Tungodden, B. (2015). Competitive in the city: A field study of gender differences in competition in Tanzania. Journal of Economic Behavior & Organization, 119, 337–345. https://doi.org/10.1016/j.jebo.2015.08.013
Gneezy, U., Niederle, M., & Rustichini, A. (2003). Performance in competitive environments: Gender differences. Quarterly Journal of Economics, 118(3), 1049–1074. https://doi.org/10.1162/00335530360698496
Buser, T., Niederle, M., & Oosterbeek, H. (2014). Gender, competitiveness, and career choices. Quarterly Journal of Economics, 129(3), 1409–1447. https://doi.org/10.1093/qje/qju009
Niederle, M., & Vesterlund, L. (2007). Do women shy away from competition? Do men compete too much? Quarterly Journal of Economics, 122(3), 1067–1101. https://doi.org/10.1162/qjec.122.3.1067 | NBER Working Paper 11474
Niederle, M., & Vesterlund, L. (2010). Explaining the gender gap in math test scores: The role of competition. Journal of Economic Perspectives, 24(2), 129–144. https://doi.org/10.1257/jep.24.2.129
Niederle, M., & Vesterlund, L. (2011). Gender and competition. Annual Review of Economics, 3(1), 601–630. https://doi.org/10.1146/annurev-economics-111809-125122