본 연구는 우리나라 주요 공공기관의 경영실적평가 자료를 이용하여 성과평가제도안에서 목표설정이 어떻게 이루어지고 있는지를 다양한 실증모델을 통해 분석하였다. 첫째, 피평가자의 차기 목표달성확률이 당기에 목표를 달성했는지 여부에 따라 차이가 있음을 발견하였는데, 이는 피평가자의 성과목표를 설정할 때 과거성과를 완전히 반영하지 않는 것을 의미한다. 둘째, 목표설정 시 톱니현상(ratchet principle)의 존재 여부에 대한 분석 결과, 전체표본에서는 차기목표가 당기성과를 기준으로 설정되었으나, 당기의 목표달성 여부에 따라 당기성과가 차기목표에 반영되는 정도는 다르지 않았다. 목표설정방법을 구분한 분석에서는 추세치, 목표부여 및 목표대실적에서 모두 톱니현상이 발견되었는데, 특히 목표부여 지표에서는 당기 목표미달 시 차기목표가 증가하는 현상이 발견되었고 목표대실적 지표에서는 하방탄력적 톱니현상이 나타났다. 셋째, 차기목표의 결정요인을 분석한 결과 목표달성여부, 그리고 목표초과달성도가 차기목표 설정에 중요하게 고려되고 있으나, 목표향상도는 중요하게 고려되지 않는 것으로 나타났다. 목표달성여부만 고려할 경우에는 성과개선도가 차기 목표결정의 중요한 요인으로 나타나지만, 목표초과달성도를 추가로 고려한 경우에는 성과개선도의 중요성은 사라지는 것으로 나타났다. 본 연구에서 목표설정과정에 피평가자가 참여하는 정도에 따라 톱니현상이 다르게 나타나는데, 이는 정보우위에 있는 피평가자가 목표설정에 보다 깊이 참여할수록 차기목표가 달성하기 쉽도록 설정될 가능성이 크다는 것을 시사한다. 또한 공공기관의 경영평가제도에서 지속적 성과개선 보다는 일시적 성과개선을 유도하거나 정부정책 및 거시경제 상황을 반영하는 목표설정이 일어날 가능성을 시사한다.

How to set targets is a crucial factor in the successful implementation of performance-evaluation and compensation plans. The primary use of targets is to coordinate the allocation of resources and to evaluate performance (Ittner and Larcker 2001; Leone and Rock 2002). Targets are also a key component of management control systems; the way firms set targets, the way they revise them over time, and the ease with which ratees achieve those targets affect their incentives to increase firm value (Indjejikian and Nanda 2002). More challenging targets lead higher performance by motivating employee to be more productive (Stedry 1960; Stedry and Kay 1966; Rockness 1977; Chow 1983). The first-best strategy to maximize firm value may be to set a target that approximates employees` true capability. Though target-setting is important and pervasive in performance-evaluation system, evidence concerning target-setting practice is scant because researchers rarely have access to such inherently firm-specific and often confidential internal information (Ittner and Larcker 2001). This paper investigates how targets are set by using a variety of empirical models coupled with performance evaluation data of the Korean Government-owned enterprises (GOEs). First, we follow the simple framework developed by Indjejikian and Nanda (2002), which assume that a firm`s executive works for two periods and the firm rewards the executive based on a measure of his performance evaluated relative to a target. This framework also assume that a target depends only past targets and past performance. Thus, under the framework, if the firm does not adjust the target for the executive`s past performance, then target at any period is timeless and does not depend on prior period`s performance. We provide empirical evidence that ratees` chance of achieving their targets depend on whether the ratees` performance exceeded their targets in the previous year. This findings suggest that raters do not adjust targets to fully reflect ratees` past performance. Second, we examine whether the ratchet principle is implemented in the performance evaluation system of Korean GOEs. The ratchet principle is the well-known tendency of raters to use current performance as a criterion for determining future targets (Weitzman 1980). Given the uncertainty about the true performance capability of ratees, raters regard the performance in any period as the revealed minimal level of their capabilities in the next period (Berliner 1976). Therefore, in a principalagent setting, the ratchet principle is adopted if contracts explicitly or implicitly reflect past performance information because performance information reduces the variance with which future performance is measured (Milgrom and Roberts 1992). We find a symmetric ratchet principle for performance measures of which targets are revised by a formula, and a super-variable ratchet principle for performance measures of which targets are set by ratees` participation. Surprisingly, targets for the next year increases rather than decreases when ratees miss targets at any period for performance measures of which targets are endowed by raters. Finally, we investigate the determinants of next period`s targets in order to establish that firms design targets to motivate their executives optimally. Leone and Rock (2002) consider two determinants, whether or not and how much ratees` beat (or meet) targets. Besides them, we additionally examine how much performance at any period increase compared to prior period`s performance and how much targets at any period change compared to prior period`s targets. We find that whether or not and how much ratees` beat (or meet) targets are significant determinants of target-setting, while change in target is insignificant. In addition, change in performance has an influential effect on the next period`s target-setting when the fact whether ratees` beat (or meet) targets or not is considered. However, we find that change in performance is no longer significant when the fact how much ratees` beat(or meet) targets is additionally included in the regression model. Our findings that ratcheting pattern depends on the degree of ratee`s participation in the target-setting process suggest that the more ratees participate in targetsetting, the more easily achievable targets seem to be set. They also imply that the current GOEs` target-setting mechanism may induce temporarily improved performance rather than persistent innovation in performance.