《中国工业经济》2023年第1期有1篇关于交错DID及异质性—稳健DID的前沿DID论文，具体分享如下：
论文原文及其附件代码自行前往《中国工业经济》官网下学习：
http://ciejournal.ajcass.org/Magazine/Show?id=84863
-
沈坤荣，林剑威，傅元海.网络基础设施建设、信息可得性与企业创新边界[J].中国工业经济.2023.1:57-75.
模型构建
-
其中，被解释变量 innov 同样表示衡量企业创新的相关指标，在此为企业新技术领域专利数 量加 1 后的对数值 -
broadband 为根据“宽带中国”试点政策构造的指示变量，城市 c 在入选 为试点城市的当年及之后年份取值为 1，否则为 0。 -
参考已有相关研究，X 为企业和城市层面的控 制变量，与正文相同，既包括人均 GDP、产业结构、交通基础设施水平等城市层面的变量，也包 括企业规模、研发投入占比、股权结构、净资产收益率、资产负债率等表征企业特征的变量。 -
同 时，本文还控制了省份—年份的交互固定效应以及行业—年份的交互固定效应，以控制来自省份 层面和行业层面的影响因素对结果的干扰。同样，αi 为企业固定效应， λt 为年份固定效应。εit为随机误差项。
估计结果
结果解释
完整命令
reghdfe patent_new did $control , absorb(id year#province year#sic_da) cluster(id)
(dropped 628 singleton observations)
(MWFE estimator converged in 13 iterations)
HDFE Linear regression Number of obs = 7,603
Absorbing 3 HDFE groups F( 11, 1603) = 1.84
Statistics robust to heteroskedasticity Prob > F = 0.0426
R-squared = 0.5847
Adj R-squared = 0.4157
Within R-sq. = 0.0047
Number of clusters (id) = 1,604 Root MSE = 0.7966
(Std. err. adjusted for 1,604 clusters in id)
------------------------------------------------------------------------------
| Robust
patent_new | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
did | .1298855 .0574112 2.26 0.024 .0172766 .2424945
gdp | -6.63e-07 8.19e-07 -0.81 0.418 -2.27e-06 9.43e-07
road | -.0061418 .0084765 -0.72 0.469 -.0227681 .0104844
highrail | .0020894 .0083714 0.25 0.803 -.0143306 .0185094
airport | -.1405601 .1627545 -0.86 0.388 -.459794 .1786739
rdspend | .004758 .0039467 1.21 0.228 -.0029833 .0124992
sharehold | .0666924 .3453499 0.19 0.847 -.6106923 .7440772
roa | -.0318536 .0285641 -1.12 0.265 -.0878805 .0241733
leverage | .0758731 .1598949 0.47 0.635 -.2377519 .3894981
size | .0830696 .0312542 2.66 0.008 .0217663 .144373
lnage | -.1029675 .0570214 -1.81 0.071 -.2148119 .0088769
_cons | .1044914 .6220923 0.17 0.867 -1.115708 1.324691
------------------------------------------------------------------------------
Absorbed degrees of freedom:
---------------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
-----------------+---------------------------------------|
id | 1604 1604 0 *|
year#province | 234 0 234 |
year#sic_da | 360 9 351 |
---------------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
. csdid patent_new $control , ivar(id) time(year) gvar(policy_year) method(dripw) agg(simple)
No never treated observations found. Using Not yet treated data
Panel is not balanced
Will use observations with Pair balanced (observed at t0 and t1)
......xx......xxxxxxxxxx
Difference-in-difference with Multiple Time Periods
Number of obs = 2,962
Outcome model : least squares
Treatment model: inverse probability
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
ATT | .3079178 .1794665 1.72 0.086 -.0438302 .6596657
------------------------------------------------------------------------------
Control: Not yet Treated
See Callaway and Sant'Anna (2021) for details.
014 最新爬取：全国2003-2021年万达广场分布数据（数据+python代码+可视化）
021 最全280个上市公司数字化转型指标2010-2021年
022 31省 & 274市政府数字化关注度2007-2021
023 31省 & 274地级市绿色发展注意力2007-2021