
The data used in the study are from China's Annual Survey of Above-scale Industrial Firms. The sample
selection and cleaning procedure is explained in the data section of the paper. From the raw data obtained,
we follow the cleanning procedure in:

	Brandt, Loren, Johannes Von Biesebroeck, and Yifan Zhang. 2012. "Creative Ac-
	counting or Creative Destruction: Firm-level Productivity Growth in Chinese Manufac-
	turing." Journal of Development Economics, 97: 339-351.

Details and cleaning code are provided in authors' website at https://feb.kuleuven.be/public/n07057/China/. 
The definition of variables are documented in code-book for the survey as well as clearly noted in our program files.

This text file provides instruction to run our computer programs. Our code files are contained within folder "code_llp".
We recommend keeping the same folder structure within this folder to minimize replication efforts.

Within this folder, there are 8 main do-files. "0_llp_project.do" is a master do-file that run through all these 8 do-files
to produce main results of the paper. Each of the do-file are self-explanatory and has detailed documentation within them.

On top of these main do-files, we have 4 additional folders. 

- Two first folders: "code_boot_model_1" and "code_boot_model_2" run bootstrap code to compute standard errors for our main
results in Table 2. Within these folder, "call_boot_llp_1.do" and "call_boot_llp_2" are the master do-files.

- The next folder: "code_boot_tech_cic" contain programs to reproduce our results in Table 5, using alternative definition
of high-tech industries, based on China's "High-tech Industry Statistical Classification Catalog" (Guo Tong Zi [2002] No.33).
The folder has similar structure to those in the first two folders.

- The last folder: "code_ols_gmm" contain programs to estimate benchmark models in Table A1 of the appendix. Again, these files
have similar structure and are self-explanatory.



  





