A Meta-Analysis of the Goal Setting-Performance Literature Thomas R. Chidester, The University of Texas at Austin W. Charles Grigshy, The Resource Group, Austin, Texas ABSTRACT Previous reviews of literature of Locke's model of goal setting have noted the percentage of confirming studies but have ignored questions of effect size and practical significance. A meta-analysis of the litera- ture was performed. The results support the two cen- tral themes of Locke's model: that hard goals lead to better performance than easy goals and that specific goals are superior to "do best" or no goals. However, several variables are found to moderate the strength of these relationships, for example, participative vs. assigned goal setting. The implications of the re- sults for motivational theory and industrial applica- tions are discussed. INTRODUCTION In reviewing the research literature on goal setting and performance, Locke, Shaw, Saari, and Latham (1981) conclude that two relationships are extremely reliable: (1) hard goals are superior to easy goals, and (2) spe- cific goals are superior to "do best" goals or no goals. Indeed, 84 percent and 96 percent (respective- ly) of the studies they reviewed supported the robust- ness of these goal setting effects. The Locke et al (1981) review, as well as an earlier review by Latham and Yukl (1975), describes previous studies, indicates the populations sampled and types of re- search designs, and notes the directions of the find- ings under review. However, the percentage statis- tics supporting their conclusions are derived from simple counts of confirming and disconfirming studies while ignoring other valuable data. Such a voting method has been a traditional tool for gaining con- fidence in the presence or absence of findings, yet it disregards variance effects due to different sample sizes, fails to identify statistically the magnitudes of moderator variable effects, and can lead to incor- rect conclusions about the reliability of findings (see Glass, 1978). The voting method approach to in- tegrating findings is no longer satisfactory because techniques are now available for using the statistics reported in a large number of research articles to estimate more reliably the magnitudes of relationships. A more accurate method for estimating the robustness of hypothesized relationships is based on calculating "effect size" statistics and correlations from de- scriptive and inferential statistics found in research publications (See Wimsatt, 1981 for a broad review of robustness analysis). This method of scientific in- quiry is referred to as meta-analysis (Glass, 1978; Hunter, Schmidt, and Jackson, 1982). Interest in goal setting effects on human performance has its origins almost fifty years ago (Mace, 1935). Of course, continued interest stems from both practi- cal and theoretical concerns. Practically speaking, information about the effects of goal setting on per- formance can easily be translated into action-specific steps for increased organizational productivity. The trend of widespread use of Management by Objectives programs is based on the fundamental assumption that if workers accept hard specific goals they will pro- duce more than if they have easy goals or no goals. Although some increased productivity should be ex- pected from goal setting, information appears lacking about how much increased productivity is likely to result from setting hard goals for workers who already have specific goals, and how much from setting specific goals were none existed before. Also in the practical vein, there seems to be an absence of re- views clearly establishing whether goals agreed upon through participation tend to be more effective than goals assigned by supervisors, and whether performance feedback or rewards given to workers tend to increase the success of goal setting. A participative approach to management and performance feedback systems may be factors crucial to the success of a goal setting pro- gram. Moreover, it would be valuable for organiza- tional planners to know the types of jobs where goal setting is likely to be least and most effective. For example, would those in autonomous or highly structured jobs respond more favorably? These are the kinds of practical considerations which need to be addressed in order to maximize the success of goal setting interventions in organizations. A close empirical examination of goal setting effects can make a theoretical contribution as well. The robustness of findings as evidenced across a large number of studies is essential to the advancement of any science (Wimsatt, 1981). Fundamental relation- ships and their magnitudes must be established in order to research more subtle processes, and to pull together a coherent theoretical structure. Locke et al (1981) suggest four underlying processes through which goals affect performance: (1) directing atten- tion and action, (2) mobilizing effort, (3) increasing persistence, and (4) motivating strategy development. More recently, self-evaluative and self-efficacy mechanisms have been hypothesized as the central processes underlying goal setting effects (Bandura and Cervone, 1983; Locke, Frederick, Lee, and Bobko, 1982). The identification of processes which reliably explain goal setting effects is not the purpose of this meta- analysis because such an inquiry seems premature at this time. Available information about the robustness of goal setting effects is currently rather limited. Information about the overall magnitudes of goal setting effects is needed in order to clarify the ex- tent that other variables come into play and to pro- vide researchers a firmer foundation to pursue the identification of underlying theoretical processes. The purposes of this meta-analysis are twofold: (1) to estimate quantitatively the performance incre- ments resulting from two types of goal setting; and (2) to explore the degree to which goal setting ef- fects are moderated by several factors, such as, performance feedback, reward, education, job autonomy, and participation in goal setting. 202