Detecting causal relationships between work motivation and job performance: a meta-analytic review of cross-lagged studies

Job performance is a core construct in applied psychology with implications for organizational outcomes and individual wellbeing. Prior meta-analyses established that motivation correlates with performance, but the causal direction remains unclear. The study addresses whether work motivation causes job performance, whether job performance causes work motivation, whether the relationship is reciprocal, or whether they are causally unrelated. To overcome limitations of single primary studies and cross-sectional evidence, the authors use a meta-analytic approach aggregating longitudinal cross-lagged panel data to test competing causal hypotheses. The purpose is to provide robust evidence on causality to inform theory (e.g., SDT, JD-R) and practice (HR management).

The paper reviews job performance as comprising task performance and organizational citizenship behavior (OCB), noting their conceptual distinctions and correlations. Performance is differentiated from efficiency and productivity. Work motivation is distinguished from personality, goal pursuit, and job attitudes, and is conceptualized across frameworks such as Self-Determination Theory (SDT) and the Job Demands-Resources (JD-R) model. Engagement is highlighted as a commonly used indicator of motivation in the work context. Four competing hypotheses are articulated: (1) motivation causes performance, supported by SDT and JD-R and evidence of medium correlations between motivation (including engagement) and performance; (2) performance causes motivation, potentially via feedback and rewards; (3) a reciprocal model; and (4) no causal relation when accounting for methodological biases. Potential moderators are proposed for exploration: type of performance measure (task vs OCB), type of motivation measure (work engagement vs other motivation constructs), and time lag between waves (1–6 vs 7–12 months).

Design: Meta-analytic structural equation modeling (MASEM) on cross-lagged panel studies measuring work motivation and job performance over at least two waves. Two MASEM steps were followed: (1) construct a meta-analytic correlation matrix; (2) conduct path analysis on the pooled matrix. Literature search: Web of Science and Google Scholar searches (January 2023) for English-language studies from 2000–2022 using terms for motivation (motivation or engagement), performance (performance, job performance, task performance, organizational citizenship behavior), and longitudinal/cross-lagged designs. No restrictions on source types. Inclusion criteria: Workplace/organizational samples; provision of full two-wave correlation matrices including six correlations (two synchronous, two cross-lagged, two stability correlations). Student/athlete samples excluded. Screening and coding: After deduplication, 120 longitudinal articles were reviewed; 11 studies met criteria, yielding 84 correlations (total n = 4389). Coded data included bibliographic info, sample characteristics, time lag, effect sizes, and reliability (Cronbach’s alpha). Intercoder agreement reached 100% after discussion. Performance measures: 8 self-reported, 2 leader-reported, 1 objective (appraisal results). Most samples from companies (k = 10) across diverse industries; 9 from Europe and 2 from East Asia. Bias assessment: Publication bias tested via Trim-and-Fill and Egger’s regression using the metafor R package. Minimal asymmetry detected; Egger’s tests non-significant across variables. Meta-analysis procedures: Hunter–Schmidt methods used to aggregate effect sizes, correcting for measurement error using reported reliabilities and for sampling error with random-effects models (psychmeta R package). Produced uncorrected r and corrected true-score correlations ρ for all six correlations (overall and subgroup matrices by performance measure, motivation measure, and time lag). Path analysis conducted on the pooled matrices to estimate cross-lagged effects (Motivation T1 → Performance T2 controlling Performance T1; Performance T1 → Motivation T2 controlling Motivation T1) and stability paths. Subgroup comparisons used z-tests for moderation (performance measure: task vs OCB; motivation measure: engagement vs other; time lag: 1–6 vs 7–12 months).

Data set: 11 studies, 84 correlations, total n = 4389. Publication bias was not serious (Trim-and-Fill imputed 1 effect for P1–P2; change −0.02; Egger’s p > .05 across all variables). Pooled correlations (overall; shown as r, ρ): M1–M2 = 0.73, 0.80; P1–P2 = 0.49, 0.54; M1–P1 = 0.31, 0.34; M2–P2 = 0.34, 0.37; M1–P2 = 0.28, 0.31; P1–M2 = 0.23, 0.26. Path analysis (overall): - Motivation T1 → Performance T2 (controlling Performance T1): β = 0.143, p < .001 (significant positive effect). - Performance T1 → Motivation T2 (controlling Motivation T1): β = −0.014, ns. - Stability paths: Motivation T1 → Motivation T2: β = 0.805, p < .01; Performance T1 → Performance T2: β = 0.491, p < .01. Synchronous associations: M1–P1 β = 0.340, p < .01; M2–P2 β = 0.128, p < .01. Moderator analyses: The positive cross-lagged path from Motivation T1 to Performance T2 remained significant across subgroups by performance measure (task performance β = 0.129; OCB β = 0.085), motivation measure (engagement β = 0.134; other motivation β = 0.101), and time lag (1–6 months β = 0.121; 7–12 months β = 0.153). The reverse path (Performance T1 → Motivation T2) was non-significant or negative across subgroups (e.g., task performance β = −0.016; OCB β = −0.052; engagement β = −0.014; other motivation β = −0.071; 1–6 months β = −0.014; 7–12 months β = −0.028). Z-tests showed no significant moderation by measurement type or time lag. Hypotheses: Supported H1 (motivation causes subsequent performance). Rejected H2 (performance causes motivation), H3 (reciprocal), and H4 (causally unrelated).

Findings indicate that work motivation causally predicts later job performance, while prior performance does not causally predict later motivation once stability and synchronous associations are controlled. This addresses the central research question by favoring a motivation-causes-performance model over reciprocal or reverse-causality alternatives. The magnitude of pooled correlations ranges from medium to large, with higher same-time associations likely reflecting common method variance. The null reverse effect may reflect indirect pathways (e.g., performance enhancing basic psychological needs or job resources which then elevate motivation), which were not testable within the present cross-lagged meta-analytic framework. The results align with SDT and JD-R theory, as well as experimental evidence linking motivation to performance, and were robust across performance and motivation measures (task performance, OCB; engagement, other motivation constructs) and across 1–12 month time lags. Practically, interventions that enhance motivation (e.g., resource provision, supportive leadership, performance management and compensation practices that foster autonomous motivation and engagement) are likely to improve performance over time; organizations may also strengthen feedback mechanisms and resource allocation to sustain motivation, given that observed performance alone did not predict later motivation in these data.

This meta-analytic review using longitudinal cross-lagged data provides the first pooled causal evidence that work motivation predicts subsequent job performance, whereas the reverse effect is not supported. Reciprocal and causally unrelated models were not supported. Results were robust across performance and motivation measurement types and across time lags between 1 and 12 months. Contributions include clarifying the causal direction underpinning the motivation–performance linkage and reinforcing SDT and JD-R frameworks with longitudinal evidence. Future research should test longer time frames, incorporate objective performance indicators, examine cultural contexts, employ instrumental variable approaches to bolster causal inference, and broaden motivation measures (e.g., intrinsic/extrinsic components) to refine understanding of motivational dynamics.

- Measurement method: Predominant reliance on self-reports for motivation and performance raises potential common method bias, particularly for same-time associations. More objective or multi-source performance indicators are needed. - Causal inference limits: Cross-lagged panel designs allow tentative causal conclusions but cannot exclude all alternative explanations; instrumental variable or experimental/quasi-experimental approaches could strengthen causal claims. - Heterogeneity: MASEM generalizability can be limited with heterogeneous correlation matrices; replication with more homogeneous samples is encouraged. - Sampling geography: Majority of primary studies were European; cultural factors (e.g., individualism) may shape motivation–performance dynamics, warranting cross-cultural tests. - Scope of measures and lags: Limited availability of certain motivation metrics (e.g., extrinsic motivation) and time lags confined to 1–12 months; future work should extend constructs and temporal windows to assess potential non-linear or delayed effects.