TFP计算大全方法名称类型需要变量优缺点OP法半参对数化增加值lny,企业是否退出变量exit,对数化年龄lnage,对数化固定资本净值lnkop,对数化投资lninv,对数化劳动力lnl,国有优点:较好地克服内生性问题缺点:对数据要求较高,观测值遗失较为严重LP法半参对数化增加值lny,对数化固定资本净值lnkop,对数化劳动力lnl,对数化中间要素投入lnm优点:可以减轻OP法用投资作为代理变量引致的观测值遗失较为严重的问题缺点:没有考虑企业退出可能带来的内生性偏误问题ACF法非参数法对数化增加值lny,对数化固定资本净值lnkop,对数化劳动力lnl,对数化中间要素投入lnm优点:可以不需要知道初始的生产函数技术形式设定(因为第一步估计没有估计任何要素投入的弹性系数),以及克服OP、LP方法在第一步估计可能产生的多重共线性问题。缺点:没有考虑到企业退出行为带来的样本选择偏ACF法非参数法对数化增加值lny,对数化固定资本净值lnkop,对数化劳动力lnl,对数化投资lninv同上DLW法(加成率)非参数法核心变量:工业总产值、劳动、中间投入和资本的自然对数y、l、m、k,其他:soe是否出口虚拟变量优点:进一步放松了市场结构、需求结构等约束条件,并能在数据库缺少价格、产量数据的现实约束下,对市场势力进行有效估计;通过引入控制函数最大限度地消除了管理能力等不可观测因素所引起的内生性问题。缺点:采用产值数据替代产量数据包含了价格外生且固定不变的错误隐含假设codextsetfirmidyearsetseed1357opreglny,exit(exit)state(lnagelnkop)proxy(lninv)free(lnl)cvars(guoyou)vce(bootstrap,seed(1357)rep(5))gentfp_op=lny-_b[lnl]*lnl-_b[lnkop]*lnkoplabelvartfp_op"OP法估计的TFP"sumtfp_opsetseed1357levpetlny,free(lnl)proxy(lnm)capital(lnkop)valueaddedreps(5)i(firmid)t(year)level(90)capdrop*tfp_lppredicttfp_lp,omegareplacetfp_lp=ln(tfp_lp)sumtfp*,dacfestlny,free(lnl)proxy(lnm)state(lnkop)robustvaoveridpredictomega_acf,omegaacfestlny,free(lnl)proxy(lninv)state(lnkop)varobustinvestoveridpredictomega_acf,omegastage1:xi:regysoe*(l*m*k*)i.yearpredictphipredictepsilon,resstage2:求解最优化方程,得生产函数各参数,以求中间投入(l或m)的弹性matavoidGMM_DLW_TL(todo,betas,crit,g,H){PHI=st_data(.,("phi"))PHI_LAG=st_data(.,("phi_lag"))Z=st_data(.,("const","l_lag","k","m_lag","l_lag2","k2","m_lag2","l_lagm_lag","l_lagk","m_lagk","l_lagm_lagk"))X=st_data(.,("const","l","k","m","l2","k2","m2","lm","lk","km","lkm"))X_lag=st_data(.,("const","l_lag","k_lag","m_lag","l_lag2","k_lag2","m_lag2","l_lagm_lag","l_lagk_lag","k_lagm_lag","l_lagk_lagm_lag"))Y=st_data(.,("y"))C=st_data(.,("const"))OMEGA=PHI-X*betas'OMEGA_lag=PHI_LAG-X_lag*betas'OMEGA_lag_pol=(C,OMEGA_lag)g_b=invsym(OMEGA_lag_pol'OMEGA_lag_pol)*OMEGA_lag_pol'OMEGAXI=OMEGA-OMEGA_lag_pol*g_bcrit=(Z'XI)'(Z'XI)}注:soe*(l*m*k*)为e和其他各变量的交互项