摘要
资源错配在现实中普遍存在。本文指出,当存在资源错配时,以Olley and Pakes(1996)为代表的代理变量方法和Gandhi et al.(2020)提出的结合静态投入要素份额回归的结构估计方法,均不能得到生产函数参数的一致估计量。因为这两种方法均需要先将生产率与随机误差项分离开,才能准确估计生产率的演化方程,其前提假设是生产率是研究者面对的、影响企业要素投入的唯一不可观测因素。这一假设在资源错配情况下无法成立,因为企业面临的扭曲税(Hsieh and Klenow,2009)也是不可观测且会影响企业的要素投入。为了解决这一问题,本文提出了一种新的企业生产函数估计方法。该方法的核心思想是直接将包含生产率和随机误差项的混合项进行多项式展开,使用工具变量方法(2SLS)解决生产率中混入随机误差项带来的生产率演化方程的估计偏误,进而利用企业要素投入和生产率演化的结构信息构造正交条件,用GMM方法得到生产函数参数的一致估计量。该方法由于无需将生产率与随机误差项分离,因此不需要假定生产率是研究者面对的影响企业要素投入的唯一不可观测因素。蒙特卡洛模拟表明,该方法在资源错配情况下能够得到生产函数参数的一致估计量,且与既有方法相比无需额外的数据信息,有较强的实用性。
Increasing total factor productivity(TFP)and improving resource allocation efficiency are two important engines driving China's high-quality economic development. The estimation of firm production function is crucial for understanding the evolution of TFP and resource allocation efficiency.While existing literature has found that resource misallocation is prevalent in the economy,this article points out that production function estimation is affected by the presence of resource misallocation. In particular, the proxy variable method represented by Olley and Pakes(1996) and the structural estimation method proposed by Gandhi et al.(2020) which combines static input factor share regression cannot obtain consistent estimates of production function parameters. This is because both methods require separating productivity from the random error term before estimating the evolution of productivity. The premise assumption is that productivity is the only unobservable factor that affects firms' factor inputs. This assumption cannot hold in the presence of resource misallocation,because the distorted tax(Hsieh and Klenow,2009)faced by firms is also unobservable and affects firms' inputs.To tackle the challenge of production function estimation with resource misallocation,this article proposes a new method for estimating firm production function. The main idea is to directly expand the mixed term containing the productivity and the random error term into a polynomial and use instrumental variable method(2SLS)to solve the estimation bias caused by mixing the random error term and productivity. This article then uses GMM method to construct orthogonal conditions with structural information on firm factor inputs and productivity evolution to obtain consistent estimates of production function parameters. Since this method does not require separating productivity from the random error term,it does not need to assume that productivity is the only unobservable factor that affects firms' inputs. In addition,compared with the dynamic
作者
李世刚
黄洪钜
莫家伟
LI Shi-gang;HUANG Hong-ju;MO Jia-wei(International School of Business and Finance,Sun Yat-sen University;School of Economics,Peking University)
出处
《中国工业经济》
北大核心
2023年第12期117-134,共18页
China Industrial Economics
基金
国家自然科学基金面上项目“地方政府间人才竞争对经济发展的影响研究:空间一般均衡的视角”(批准号72373168)
国家自然科学基金青年项目“进口产品类型、关税结构与企业生产率研究”(批准号72103001)
教育部人文社会科学基金青年项目“中国制造业企业产品质量估计:方法与应用研究”(批准号18YJC790089)。
关键词
资源错配
企业生产函数
全要素生产率
结构估计
resource misallocation
firm production function
total factor productivity
structural estimation 1
