China’s emerging labor market was buffeted by changes in demand and supply and institutional changes in the last two decades. Using the Chinese Urban Household Survey data from 1989 to 2009, our study shows that the market responded with substantial changes in the structure of wages and in employment and types of jobs that workers obtained that mirrors the adjustments found in labor markets in advanced economies. However, the one place where the Chinese labor market appears to diverge from the labor markets in advanced countries is the rapid convergence in earnings and occupational positions of cohorts who entered the job market under more or less favorable conditions. On this dimension, China’s labor market seems more flexible than those in other countries. Three related factors may explain this pattern: (1) the rapid growth of China’s economy; (2) the high rate of employee turnover; (3) the relative weakness of internal labor markets in China. Bottom line, the Chinese labor market has responded about as well as one could expect to the changes in the demand and supply factors and institutional shocks in this critical period in Chinese economic history.
(1) 说实话,我是被题目中的“Really big changes”吸引过来的,我想看看对于"Really big changes", "big changes" 和"marginal changes"的处理有什么不同,不过好像没发现。
(2)首先看到的是Abstract,第一句话使我联想到Katz and Murphy, QJE., 1992那篇经典,我以为这篇是那篇在中国背景下的引用,结果不太一样。
(3) 以下说正事。我觉得本文中以下两处实证结果之间可能存在着不一致,它们分别是Table 2, Panel A, Cols. 1, 2和Appendix A。
首先我们看第一处的证据,因为作者对模型设定说明的不太清楚,我只能凭感觉认为作者意在估计收敛速度beta (speed of convergence),如果我的理解正确,那么我们可以用70-法则大概确定半衰期,也就是用0.7/0.012 (col.1)和0.7/0.014 (col.2)。可以发现它们的值相当大,这说明实际上从初始状态y(0)向稳态y*的过渡实际相当缓慢。我这里的理解是否正确,请大家检查。
如果我以上的理解有误,那么我们采纳作者的说法,认为从初始状态向稳态的过渡很迅速,只需要大致3~5年,那么我认为此处的证据与Appendix A之间就存在一些出入,理由如下:如果我们横向观察各Entry year dummy系数的估计值,可以发现基本上其数值随进入市场后的年限的上升而下降。我注意到,在约2000年之前,此下降的速度较快,但通常在第五年的时候系数仍然是显著的;而在2000年之后,该系数随进入市场时间增加而下降的速度明显放缓,通常第五年的系数仅仅比刚进入市场时的系数略小。以上事实是否说明:至少在约2000年之后,Entry year dummies,或者说进入市场当年的市场景气条件的影响有很强的持续性。如果此解释正确,那么上面提到的两处证据之间就存在着不一致。
当然,作者可以说,Entry year dummies在2000年之后的缓慢下降并不意味着进入市场时的景气状况对后续工资有持久影响,而是一种Cohort effect,例如,这一代在未成年时期的营养状况、所受到的教育,在进入市场之前已经被决定,因此比较稳定很正常。但是我认为,这样的解释有其内在不一致性:如果认为2000年后entry year dummies的缓慢下降主要应该有Cohort effects的稳定来解释,那么如何对2000年之前该系数的快速下降做出统一的解释?
(4) 最后,表达我自己的一种希望,作者有非常难得的数据,代表性好,且时间跨度较长(这点很重要),如果作者能够报告一些在中国背景下劳动力如何分组, linear synthesis以及要素替代弹性的结果就好了,毕竟美国文献中常用的H=college graduate or college graduate +0.5 some college, L=high school graduate...不一定适用于中国,但是这些重要的参数对于我们这些拿不到这些数据,又想要做研究的人很重要。其实这对文章的引用率也绝对有帮助,想想多少人是为了那个神奇的数字,sigma (H,L)=1.4而引用Katz and Murphy, 1992的。
(i) Table 2, Panel A, Col. 2中的回归设定和结果:
因变量:logY(t=5)/Y(t=0), where Y: earnings, t=0: 进入市场时点,t=5: 进入市场5年以后
自变量:表述不是很清楚,作者解释如下:"...starting... earnings are relative earnings, indicating to what extent the earnings exceed the time trend, equal to the gap between actual starting... and the predicted ones from time trend“,据此,我猜测应该是Y(0)/trend(0)
回归设定:线性,包含常数项
回归结果:logY(t=5)/Y(t=0)=1.277-0.014*Y(0)/trend(0),所有参数均在a=1%下显著。
(ii) 作者对此的解释:"... the results show substantial "regression to the mean" in the form of sizable negative coefficients on starting ... earnings. If the starting salary was lower in a year by 10% from the predicted trend, the growth rate in the first 3-5 years would be 14% percent (原文如此) higher than it otherwise would have been, implying an approximately 3 year catch up period...." 结论是:"This is a must faster catch-up than is found in the US and other advanced countries."