【作者】 柳昌承；

【导师】 高闯；

【作者基本信息】 辽宁大学 ， 企业管理， 2008， 博士

【摘要】 自1972年以来,以美国为首次西方发达国家,对企业的社会绩效与财务绩效关系进行了大量的实证研究。虽然现有的研究显露出种种局限,没有取得具有一贯性的明确结论,但是作为倍受关注的“社会绩效”,正受到越来越多的研究者、企业经营者以及所有的利益相关者的关注。在韩国从1991年以来,有影响力的市民团体经济正义实践联合会下属的经济正义研究所每年都评价和发表“经济正义指数(KEJI Index)”,这一指数正成为评价企业社会绩效的指标。但是,依据该研究资料,只有少数研究者们揭示有限的研究结果,对公司高管的决策帮助不大。本研究采用自金融危机以后的2001年到2005年的期间,以110家韩国上市公司的共550个均衡面板数据,采用横截面和面板回归分析方法来比较、分析了韩国上市公司的财务绩效和社会绩效的关系。本研究利用2000年以后开始出现的这一领域的研究潮流,使用Pooled OLS、固定效果模型及随机效果模型等面板回归分析方法,观察了研究结果表现的差异,补充了缺少以韩国企业为对象的研究的不足。并且,为了减少由于省略变量而引起虚构相关的可能性,不仅观察了控制企业规模、杠杆率及产业的现有研究的单纯模型,而且还利用“顺向选择方式”采用大量控制变量来观察了系数的敏感性。而且,按年度进行的共5年的横截面回归分析的结果,在从中查看多数现存研究的一年期的横截面分析产生的结果是否可以一概而论。研究结果发现:第一,(1)从2002年以来韩国上市公司的社会绩效对企业价值产生U字形的曲线关系的影响,一部分支持“善意经营假设”。这意味着社会绩效达到一定水平的企业(把经济贡献度去掉之后的KEJI指数(65分满分)大于38.34-39.68),如果再提高社会绩效水平还能引起更大幅的企业价值的提高。(2)企业价值高的企业不附带高社会绩效,且在研究期间的韩国企业不支持“剩余资源假说”。这就表明韩国上市公司社会绩效的决定因素并不是企业价值,而是企业的规模或资本结构、是否属于财阀企业、所有持股率及独立董事比率等。第二,本研究比较、分析了面板回归分析和横截面回归分析产生的结果。通过F-检验(固定效果模型)和Lagrange乘数检验(随机效果模型)的结果显现了企业特有性质是存在的。因此,面板数据回归分析方法能比Pooled OLS或者横截面分析取得更有效的结果。这个事实暗示在以后的研究中利用面板数据分析的可行性。第三,关于模型设定误差,本文在社会绩效模型当中的Pooled OLS和几个横截面分析里发现了与McWilliams & Siegel(2000)主张过的一样的现象。但是,在面板回归分析里的企业价值和社会绩效变量则没有发现随着增加了控制变量其系数和显着性呈现大变化的现象。这说明本文采用的固定效果模型和随机效果模型结果的稳健性。第四,本研究分析了大多数现有研究通过1年的横截面分析得到的结果的一般性问题。分析结果表明,企业价值和社会绩效变量一贯地显示了不显着的结果,所以认为对其结果进行一般化的问题不大。但是,横截面回归分析与本研究的结果一样存在着提供不一致结果的可能性,从而又表明了使用面板数据方法的优越性。最后,关于插入到模型中的各控制变量,本研究发现大体上与理论及现有实证研究一致的结果。更多还原

【Abstract】 Since the empirical researches with the view to clarify the relationship between corporate social performance and corporate financial performance started in 1972, a number of researches have dealt with the firms in the developped countries. Although existing studies, with not a few limitations, have not gained a consistent result by which one cannot successfully see a clear picture, this issue, as one of the major issues of CSP studies, is gaining more and more attention from not only researchers, managers at fields but also all the stakeholders surrounding firms. In Korea, KEJI under CCEJ, one of the Korean influential civil groups, has been evaluating and pronouncing KEJI Index annually since 1991. However,only a few researchers in Korea have sporadically produced limited empirical results, which cannot be helpful enough for business managers to make CSP-related decisions.This research, by using 110 Korean listed firms’550 balanced panel data during 5 years from 2001 to 2005, analyzed the relationship between CSP and CFP through year-to-year OLS regression and panel regression such as fixed effect model and random effect model.This paper, according to the recent research tendency of using panel regression which was initiated since around 2000 in this field, has important meaning in compensating for the lack of research for Korean listed firms. In addition, this paper, in order to minimize the possibilities of“spurious correlation”caused by omitted variables, which had been pointed out as one of the limitations for empirical research in this field, not only examined the simple model including firm size, leverage ratio, and industry dummy variables made by existing researches, but also checked the sensitivity of results by other models with a variety of control variables through forward selection method. In addition, this paper, by analyzing the year-to-year cross sectional OLS regressions, examines the possibility of generalization the existing empirical papers’results provided by using 1 year data.The result found out:First, (1) Korean listed firms’CSP has significant quadratic curve relationship with CFP since 2002, which partially supports“good management hypothesis”. This means Korean firms with higher than certain level of CSP(according to KEJI index after removing economic performance, higher than 38.34-39.68), increase of its CSP level can be related to get better CFP. (2) The firms with high value don’t have to do with high CSP, which doesn’t support“slack resource hypothesis”for Korean firms during the research periods. The result means Korean firms’CSP is affected not by CFP but by other factors such as firm size, capital structure, chaebol incorporation, ownership structure, and ratio of outside directors.Second, this research proviedes the opportunity of comparing both results from cross-sectional regression and panel regression. F-test and Lagrange multiplier test prove there exist individual effects, which imply panel regression can get more efficient result than pooled OLS or cross-sectional regression. This implies the necessity of using panel regression method.Third, this research found out pooled OLS and some cross-sectional models of CSP model have mis-specified model problems McWilliams & Siegel(2000)had argued. However, no big changes haven’t been observed in the coefficients of CSP and CFP variables in the panel regression, which also proves robustness of panel regression.Fourth, this research also examines the generalization issue of the results provided by a lot of existing papers’cross-sectional regression. In this paper, firm value and CSP variables consistently provide insignificant result in all the year-to-year models. Therefore, this research concludes the generalization issue is not a big problem. However, cross-sectional regression has the possibility of providing biased result, which also tells the superiority of using panel regression method.Last, with regard to control variables in the models, the results generally are the same as theoretical and empirical results of other researches.更多还原