Motivation and Contribution in Open Source Software: A Meta-Analytic Investigation

Introduction

• Findings on motivation are mixed: studies report conflicting relationships between motivations such as enjoyment, pay, status, and contribution effort.Reciprocity and own use have been reported as positively, negatively, or not related to contribution.Source: Introduction · p. 2
• Contribution effort and continuance intention are distinct outcomes, and studies also operationalize reputation and monetary rewards differently.Effort is work actually done; continuance intention is a stated plan to keep contributing.Source: Introduction · p. 2
• The meta-analysis pools correlations for eight antecedents with effort and continuance intention, while testing motivation ordering, reputation-system moderation, and crowding predictions.It focuses on OSS and uses related settings for comparison.Source: Introduction · p. 3
• Each antecedent relates at least as strongly to continuance intention as to effort; intrinsic motivation is not stronger than extrinsic motivation for effort, but has a larger intention–effort gap.Reputation relates to effort mainly where a reputation system exists, and the evidence shows no sign that money crowds out enjoyment.Source: Introduction · p. 3

Theoretical Background and Hypotheses

• Open collaboration creates shared goods available to contributors and noncontributors; the framework groups eight antecedents as intrinsic, extrinsic, or neither.It draws on self-determination theory’s distinction between inherent satisfaction and separable outcomes.Source: Theoretical Background and Hypotheses · p. 3

Construct Definitions and Measurement

• Measures are assigned by item content rather than labels; effort includes hours, self-rated effort, or trace-recorded work, while continuance intention is the stated plan to continue.Platform-displayed reputation indicators are reserved for a robustness check, and ideology is excluded because few studies report it and definitions differ.Source: Construct Definitions and Measurement · pp. 4–5

Motivations, Contribution Effort, and Continuance Intention

• The authors hypothesize that all eight antecedents will relate positively to both contribution effort and continuance intention.The antecedents include enjoyment, reciprocity and own use, reputation, career and human capital, monetary rewards, self-efficacy and expertise, community identification, and altruism.Source: Motivations, Contribution Effort, and… · p. 7

Intrinsic versus Extrinsic Motivation

• Intrinsic motivation comprises enjoyment, community identification, and altruism; extrinsic motivation comprises reputation, career and human capital, and monetary rewards.The main comparison predicts a stronger intrinsic relationship with effort; reciprocity and own use are excluded because classifications disagree, while self-efficacy and expertise is not a motivation.Source: Intrinsic versus Extrinsic Motivation · p. 7

The Intention-Behavior Gap

• The intention–behavior gap is the difference between a motivation’s links to continuance intention and effort; the authors predict a larger gap for intrinsic motives.They cite resource constraints such as time and the possibility that respondents overstate socially desirable motives alongside plans to continue.Source: The Intention-Behavior Gap · pp. 7–8

Moderators

• The moderator hypotheses predict stronger effort relationships in OSS than other settings, changing enjoyment and reward links over time, and stronger reputation and career links where reputation systems exist.The setting test is exploratory because the search was designed for OSS.Source: Moderators · p. 8

Motivation Crowding

• The crowding hypotheses predict weaker enjoyment–effort links where monetary rewards are present, a negative enjoyment–reward relationship, and greater cross-study heterogeneity for rewards than enjoyment.Source: Motivation Crowding · p. 9

Literature Search

• The search covered seven scholarly databases plus ProQuest dissertations and SSRN, using three AND-combined term blocks and no date or document-type limit.It retrieved 20,959 records; after duplicate and language exclusions, 13,836 remained.Source: Literature Search · p. 9

Screening, Eligibility, and Coding

• After screening, 41 studies were included; the main analysis uses 178 zero-order correlations from 29 independent samples in 27 studies (N = 9,632).The samples include 18 from OSS and 11 from other settings; effort was measured in 20 samples.Source: Screening, Eligibility, and Coding · p. 10
• Two authors coded each effect and sample, independently coding all 41 studies and resolving the remaining 11 of 247 effects after discussion.Agreement was high across coded variables, with correlations within .01 for 97.0% of effects.Source: Screening, Eligibility, and Coding · pp. 11–12

Analysis

• Correlations were Fisher-z transformed and pooled with random-effects models using inverse-variance weights; estimates were transformed back to correlations.The analysis did not correct for measurement error, so estimates are described as conservative.Source: Analysis · p. 12
• Each sample contributes one effect per relationship, and pooling requires at least three samples; results include confidence and credibility intervals and heterogeneity statistics.The expanded robustness analysis adds partial correlations from samples reporting regression results.Source: Analysis · p. 12
• The tests compare pooled intrinsic and extrinsic relationships, intention–effort gaps, moderator subgroups, and enjoyment–reward correlations.Moderator tests use subgroup analyses or meta-regression, while crowding heterogeneity is assessed by comparing τ².Source: Analysis · p. 12
• Robustness checks separate self-reported from trace-based effort, vary samples and measure types, and assess publication bias with published-versus-unpublished comparisons and two tests.Egger’s test and trim-and-fill are used for relationships with at least 10 samples.Source: Analysis · pp. 12–13

Motivations, Contribution Effort, and Continuance Intention

• All eight antecedents were positively related to intention, while all except reciprocity and own use were positively related to effort; every antecedent related at least as strongly to intention as to effort.Correlations with effort ranged from r = .11 to r = .28, while those with intention ranged from r = .22 to r = .53; I² exceeded 80% for 14 of 16 relationships.Source: Motivations, Contribution Effort, and… · p. 13, Table 3 · p. 14

Panel B. Contrasts

• r = .21 for intrinsic versus r = .28 for extrinsic motivation with effort (Δz = −.07, 95% CI [−.24, .09], p = .369), so H9 was unsupported.The within-sample difference was also similar in 14 samples reporting both groups (Δz = −.06, p = .205).Source: Intrinsic versus Extrinsic Motivation… · p. 14, Table 4 · p. 15

Intrinsic versus Extrinsic Motivation and the Intention-Behavior Gap

• The intention-effort gap was larger for intrinsic than extrinsic motivation (Δz = .32 versus .04; difference Δz = .28, 95% CI [.02, .54], p = .033).Only three samples reported both outcomes for both groups, and dropping one flagged altruism-intention value reduced the difference to Δz = .22 (p = .078).Source: Intrinsic versus Extrinsic Motivation… · p. 14, Table 4 · p. 15

Moderators and Motivation Crowding

• Reputation related to effort on platforms with a reputation system (r = .35), but not without one (r = .00; Δz = .37, p < .001).The result held after excluding the sample with the highest correlations and after excluding citizen science samples.Source: Moderators and Motivation Crowding · p. 15, Table 5 · p. 16
• Monetary rewards and enjoyment were positively related (r = .21, 95% CI [.07, .34], k = 11), with no observed undermining effect.The reward-effort relationship varied less across studies than the enjoyment-effort relationship (τ² = .027 versus .055); the enjoyment-by-reward comparison was not a valid C1 test because three of four no-reward samples came from one article.Source: Moderators and Motivation Crowding · pp. 15–16, Table 5 · p. 16

Robustness and Publication Bias

• The H9 effort comparison held across robustness checks, but the larger intrinsic intention-effort gap lost significance in several restricted analyses and after dropping the flagged sample.No sign of publication bias was found, despite significant Egger tests for reputation and the extrinsic group with effort.Source: Robustness and Publication Bias · p. 16, Table 6 · p. 17

Discussion

• The synthesis finds that every antecedent relates at least as strongly to continuance intention as to contribution effort; intrinsic motivation is not stronger than extrinsic motivation for effort, but has a larger intention–effort gap.The authors describe the review as the first quantitative synthesis of OSS and related open collaboration settings.Source: Discussion · p. 17

Intrinsic and Extrinsic Motivation

• Intrinsic motivation’s apparent effort disadvantage may reflect the contrast used: autonomous motivation includes internalized extrinsic goals, while effort measures may miss motivation’s relationship to contribution quality.The authors recommend distinguishing autonomous from controlled motivation and studying contribution quality alongside amount.Source: Intrinsic and Extrinsic Motivation · pp. 18–19

From Intention to Effort

• The intention–effort gap is larger for intrinsic motives, especially altruism and community identification; intrinsic motivation related less to platform-recorded than self-reported effort, while extrinsic motivation did not.Because the gap test rests mainly on comparisons across samples, the authors say it should be confirmed within samples.Source: From Intention to Effort · p. 19

Reputation and Platform Design

• Reputation related to effort mainly on platforms with a reputation system, where contributions become visible through points, ranks, or badges.The authors caution that few platforms in their data lacked a reputation system, so the moderation should be confirmed, for example with field experiments.Source: Reputation and Platform Design · pp. 19–20

Monetary Rewards and Motivation Crowding

• Monetary rewards were positively related to effort, more strongly for actual payment than desire for money; the analysis found no sign that rewards crowd out enjoyment.The authors suggest pay may provide time for contribution, but cross-sectional correlations cannot rule out crowding within individuals.Source: Monetary Rewards and Motivation Crowding · p. 20

Reciprocity, Career, and Ability

• Reciprocity and own use was the only antecedent not reliably related to effort, while self-efficacy and expertise related positively to both effort and intention.Contributors may stop once their own need is met; the authors suggest separating private need from obligation to give back.Source: Reciprocity, Career, and Ability · p. 20

Implications for Practice

• Platform owners should not treat intention surveys as forecasts of contribution, especially when intentions rest on enjoyment or belonging; maintainers can lower time and skill demands with small, clear tasks.The authors also note that firms can pay OSS developers without evidence here that pay displaces enjoyment.Source: Implications for Practice · pp. 20–21

Limitations and Future Research

• The evidence is bounded by an OSS-focused, English-language search, heterogeneous measures, incomplete pooling, and mostly cross-sectional studies; contribution quality and actual departures were not examined.The authors call for measures of motivation before contribution and field experiments varying rewards or recognition to strengthen causal tests.Source: Limitations and Future Research · p. 21

• Motivation explains stated intention better than effort, and the gap is widest for intrinsic motives.Source: Conclusion · p. 21
• Intrinsic motivation was not more strongly related to effort than extrinsic motivation, and money did not crowd out enjoyment.Source: Conclusion · p. 21
• Reputation mattered mainly where platforms made contributions visible; the authors identify learning how to turn willingness into work as the next step.Source: Conclusion · p. 21

• Search records, coded data, and the R analysis script are available at Mendeley Data under DOI 10.17632/kdkxpgtgr9.1.Source: Declarations · p. 21
• The authors used Claude Opus 5.5 to improve grammar and writing quality, not to design the study, code studies, analyze data, or produce results.The authors say they reviewed and edited the AI-assisted text and take responsibility for the manuscript.Source: Declarations · p. 22
• The authors declare no conflict of interest and report receiving no funding for the work.Source: Declarations · p. 22

Appendices

• Appendix D reports expanded analyses and robustness checks, Appendix E reports moderator tests, and the full coding sheet and analysis script are available at Mendeley Data.Source: Appendices · p. 27

Appendix A. Examples of Excluded Reports by Eligibility Criterion

• The appendices provide examples of reports excluded under eligibility criteria, while Figure 2 reports exclusions by criterion.Source: Appendix A · p. 27, Table A1 · p. 27

Appendix B. Included Studies and Samples

• Table B1 lists 41 included studies and 46 independent samples, with one sample per row; the appendices also identify studies whose reports were merged.Appendix B’s notes define sample, design, antecedent, outcome, analysis, reward, reputation-system, and publication-status codes.Source: Appendix B · pp. 28, 32, Table B1 · p. 28

Appendix C. Correlations by Sample

• Appendix C reports sample-level correlations for effort and intention, with partial correlations added only when a sample lacks a zero-order correlation for that relationship.The appendix also lists platform reputation indicators used in a robustness check.Source: Appendix C · pp. 33, 36, Table C1 · p. 33, Table C2 · p. 34
• Brown (2016) contributed the flagged .89 correlation between altruism and intention that was dropped in a sensitivity check.Source: Appendix C · p. 35
• Lakhani and Wolf (2003) reported a value of .04 for reciprocity and own use predicting effort, based on a sample of N = 630.Source: Appendix C · p. 36
• Racero et al. (2022) reported a .50 value for the relationship between enjoyment-based intrinsic motivation and intention, with N = 231.Source: Appendix C · p. 36

Appendix D. Expanded Analysis and Robustness Checks

• Only monetary rewards’ relationship with continuance intention loses significance in the expanded analysis; H9 is nonsignificant in every specification.Table D1 compares the main and expanded analyses, while Table D2 reports H9 and H10 across alternative specifications.Source: Appendix D · p. 38, Table D1 · p. 38, Table D2 · p. 39
• H10’s difference between the intrinsic and extrinsic intention–effort gaps stays positive across specifications, but is nonsignificant in several checks.Nonsignificance occurs for published studies only, surveys only, self-reported effort only, without the flagged correlation, or after dropping one of three samples.Source: Appendix D · p. 38, Table D3 · p. 40
• The expanded analysis adds partial correlations to the main analysis, and the robustness tables test alternative specifications and samples dropped one at a time.Source: Appendix D · p. 39, Table D2 · p. 39, Table D3 · p. 40

Appendix E. Moderator Results by Antecedent

• The OSS-versus-other-setting comparisons cover antecedents and outcome groups, but are exploratory because studies in other settings form the comparison set.Source: Appendix E · p. 43, Table E1 · p. 44
• Publication-year slopes estimate change in Fisher’s z per decade, centered at 2015, using random-effects meta-regression with the Knapp–Hartung adjustment.Source: Appendix E · p. 45, Table E2 · p. 45
• The publication-year meta-regression was run only for relationships with at least four available samples.Source: Appendix E · p. 45