The Price of Work: Understanding Wage Inequality by Estimating Task Prices

在过去的几十年里，在整个发达国家，工资中位数相对于收入分配中较高部分的工资大幅下降。与此同时，中等收入工作岗位的数量稳步下降。这引发了一场激烈的辩论，即程序化技术的兴起是否对这些趋势负有责任。程序化技术可以自动化大量中等收入工作，但不会有很多低收入或高收入工作。例行化预测，如果工人的技能构成不变，观察到的中位数工资和中等收入工作的工资不仅应该比最高工资下降，而且应该比最低工资下降。然而，这种情况只发生在美国，而在欧洲和加拿大，低收入工作和低端工资分配的工资甚至进一步下降。因此，中等收入(和低收入)工人日益增加的问题是由普遍的技术变化推动的，还是由于不同国家的政策决定造成的，因此是可逆的，这仍然是一个悬而未决的问题。回答这个问题的关键是任务价格，也就是为不同工作单位的技能支付的价格。不同工作和收入分配不同部分的工资变化由于构成效应而变得模糊，因为不同技能的工人随着时间的推移在不同的工作和工资分配的分位数之间流动。例如，随着工业机器人和自动化生产流程的兴起，制造业的许多生产工作岗位已经消失。从经验上看，尽管自动化之前生产工人所做任务的价格有所下降，但观察到的制造业工作的平均工资可能有所上升，因为剩下的工人技能更高。因此，为了评估技术变革对工资的影响，人们需要将任务价格从组成效应中分离出来。拟议的项目开发了一种估计任务价格变化的新方法。这种方法是基于我最近的一个理论结果，该结果将工人的工资增长和他们的工作分类与不断变化的任务价格联系在一起。关于工人和公司的纵向数据，现在广泛可用，使这一结果的经验性实施成为可能。总而言之，新的理论见解和数据将导致对现有任务价格估计方法的实质性改进。新的任务价格估算方法还有几个额外的和相关的应用。例如，在许多国家，金融部门收入飙升的驱动因素是什么？为什么低技能服务性工作的工资下降，而就业增加？与男性相比，女性的收入增加是因为她们专门从事需求增加的工作吗？估算任务价格将为这些和其他重要问题提供新的答案。

Over the last decades, median wages have dropped substantially relative to wages in the upper parts of the income distribution throughout the developed world. At the same time, the number of middle-income jobs declined steadily. This has generated an intense debate about whether the rise of routinization technology, which automates a lot of middle-income but not many low- or high-income jobs, is responsible for these trends. Routinization predicts that, if the skill composition of workers does not change, observed median wages and wages in middle-income jobs should not only fall compared to wages at the top, but also at the bottom. However, this only happened in the United States, while wages in low-income jobs and in the lower end of the wage distribution dropped even further in Europe and in Canada. It thus remains an open question whether middle-income (and low-income) workers' rising troubles have been driven by pervasive changes in technology or whether they are due to policy decisions in the different countries, and therefore reversible. The key to answering this question are task prices, that is, prices that are paid for a unit of skill across jobs. Observed wage changes in different jobs and in different parts of the income distribution are blurred by composition effects, as workers of different skills move across jobs and quantiles of the wage distribution over time. For example, many production jobs in the manufacturing sector have disappeared with the rise of industrial robots and automated production processes. Empirically, while the price of the tasks that production workers did before automation has dropped, observed average wages in manufacturing jobs may have increased because the remaining workers are more highly skilled. Therefore, in order to evaluate the effects of technological change on wages, one needs to disentangle task prices from composition effects. The proposed project develops a new method for estimating changes in task prices. This method is based on a recent theoretical result of mine that links workers' wage growth and their sorting into jobs to changing task prices. Longitudinal data on workers and firms, which are by now widely available, enable the empirical implementation of this result. Together, the new theoretical insight and the data will lead to a substantial improvement upon existing approaches of estimating task prices. There are several additional and related applications of the new estimation method for task prices. For example, what are the drivers of the surging earnings in the financial sector in many countries? Why did wages in low-skill service jobs drop while employment increased? Are women's earnings rising compared to men's earnings because they are specialized in jobs for which demand has increased? Estimating the task prices will provide new answers to these and to other important questions.