* Readme.txt
* PL Aggregate Productivity for the Japanese data.
* Hyeog Ug Kwon, Futoshi Narita, and Machiko Narita
* June 17, 2015

* See also do_all.do, which runs all the Stata codes in order.

[Stata Program Flow]
================================================================================
1. Data Setup

[Do files]
	KNN_data.do
	
[Input files]
	../matching code_industry.xlsx	(not included due to restricted access)
	../EF_def.xlsx			(not included due to restricted access)
	../JIP2011.xlsx			(available from the website of RIETI)
	../MC2010 by ind_jip2006.dta	(not included due to restricted access)

--------------------------------------------------------------------------------
	1-0-1. Prepare for industry code matching, using:
		"matching code_industry.xlsx"
	1-0-2. Prepare for electricity and fuel deflator, using:
		"EF_def.xlsx"
	1-0-3. Prepare JIP2011 data for merging industry-level data, using:
		"JIP2011.xlsx"
	1-1. Read a new data set from 
			../MC2010 by ind_jip2006.dta
		 Only for the variables needed.

	1-2. Set up the data.
		1-2-1. Create level variables.
		1-2-2. Drop the data with missing values.
		1-2-3. Drop plants with non-positive entries.
		1-2-4. Assign new id to firms who move between industries.
		1-2-5. Create new capital estimates.
		1-2-6. Create capital adjustment factors.
		1-2-7. Create lags and time differences.
		1-2-8. Create the flag variables such as 
			   flgDiffExist, flgEntry, and flgExit.
		1-2-9. Create the sum of nominal value added.

	Save the results as KNN_data.dta.


================================================================================
2. Production Function Estimation for Each Method.

[Do files without capital measurement error correction]
	KNN_OLST_GO.do		// OLS
	KNN_FET_GO.do		// Fixed Effect
	KNN_FDT_GO.do		// First Difference OLS
	
	KNN_WLPTM_GO.do		// WLP using intermediate material use as a proxy
	KNN_WLPTE_GO.do		// WLP using electricity use as a proxy
	KNN_WLPTF_GO.do		// WLP using fuel use as a proxy
	KNN_WLPTEF_GO.do	// WLP using both electricty and fuel use as proxies
	KNN_WLPTEF2_GO.do	// WLP using the sum of electricty and fuel use as a proxy
	
(Note) "GO" stands for "Gross Output" and "VA" stands for "Value-Added".
	 (Value-Added versions are not included)

[Do files with capital measurement error correction]
	KNN_WLPTM_CM_GO	// WLPTM with 2nd K constructed using material use share
	KNN_WLPTE_CM_GO	// WLPTE with 2nd K constructed using material use share
	KNN_WLPTF_CM_GO	// WLPTF with 2nd K constructed using material use share
	KNN_WLPTEF_CM_GO	// WLPTEF with 2nd K constructed using material use share
	KNN_WLPTEF2_CM_GO	// WLPTEF2 with 2nd K constructed using material use share

	KNN_WLPTM_CE_GO	// WLPTM with 2nd K constructed using electricity use share
	KNN_WLPTE_CE_GO	// WLPTE with 2nd K constructed using electricity use share
	KNN_WLPTF_CE_GO	// WLPTF with 2nd K constructed using electricity use share
	KNN_WLPTEF_CE_GO	// WLPTE with 2nd K constructed using electricity use share
	KNN_WLPTEF2_CE_GO	// WLPTEF2 with 2nd K constructed using electricity use share

	KNN_WLPTM_CF_GO	// WLPTM with 2nd K constructed using fuel use share
	KNN_WLPTE_CF_GO	// WLPTE with 2nd K constructed using fuel use share
	KNN_WLPTF_CF_GO	// WLPTF with 2nd K constructed using fuel use share
	KNN_WLPTEF_CF_GO	// WLPTEF with 2nd K constructed using fuel use share
	KNN_WLPTEF2_CF_GO	// WLPTEF2 with 2nd K constructed using fuel use share

	KNN_WLPTM_CEF_GO	// WLPTM with 2nd K constructed using the sum of electricity and fuel use
	KNN_WLPTE_CEF_GO	// WLPTE with 2nd K constructed using the sum of electricity and fuel use
	KNN_WLPTF_CEF_GO	// WLPTF with 2nd K constructed using the sum of electricity and fuel use
	KNN_WLPTEF_CEF_GO	// WLPTEF with 2nd K constructed using the sum of electricity and fuel use
	KNN_WLPTEF2_CEF_GO	// WLPTEF2 with 2nd K constructed using the sum of electricity and fuel use
	
(Note) "_C" means that we use capital measurement error correction, 
		which is followed by letters that indicate how we construct 
		another measure of capital stock.

[Do files for additional steps]
	KNN_rs_GO.do				// Calculate cost shares just for reference.
	KNN_WLPMIX_GO.do			// Choose an estimator for each industry, 
								//   including the ones with capital measurement error correction.
	KNN_WLPMIX_GO_FirmObs.do	// Count firm observations used in the estimation by industry.
	KNN_WLPNBD_GO.do			// Choose an estimator for each industry, 
								//   excluding the ones with capital measurement error correction.

[Do files for bootstrapped standard errors]
	Bootstrap\ind13\BS_ind13.do
	Bootstrap\ind14\BS_ind14.do
	Bootstrap\ind15\BS_ind15.do
	Bootstrap\ind18\BS_ind18.do
	Bootstrap\ind22\BS_ind22.do
	Bootstrap\ind23\BS_ind23.do
	Bootstrap\ind24\BS_ind24.do
	Bootstrap\ind25\BS_ind25.do
	Bootstrap\ind28\BS_ind28.do
	Bootstrap\ind30\BS_ind30.do

[Input files]
	KNN_data.dta (created in step 1)

--------------------------------------------------------------------------------
[For each do file for estimation, without capital measurement error correction]
	2-1. Read the data.

	2-2. Drop unnecessary variables, to save the memory.

	2-3. Estimate production functions. 

	Save the results as est_(Method)_GO.dta, 
		e.g., est_OLST_GO.dta.
		(The log files of KNN_OLST_GO.do, KNN_OLST_GO.do, and KNN_OLST_GO.do
		 are used to produce Table B.I)


[For each do file for estimation, with capital measurement error correction]
	2-1. Read the data.

	2-2. Prepare variables for the estimation. 
		2-2-1. Drop unnecessary variables, to save the memory.
		2-2-2. Construct khat, which has the same sample moments that 
				the true capital stock (kstar) should have.  

	2-3. Estimate production functions using khat as capital stock.

	Save the results as est_(Method)_GO.dta, 
		e.g., est_WLPTM_CM_GO_GO.dta.


[For do files for additional steps]
	2-4. Calculate cost shares for each industry 
		(Run KNN_rs_GO.do)

	Save results as est_RS_GO.dta.

	2-5. Choose estimates for each industry using measurement error correction
		(Run KNN_WLPMIX_GO.do)

	Save results as est_WLPMIX_GO.dta and est_WLPMIX_GO.xlsx.
		(Used to produce Table 3 and Table B.I)

		
	2-6. Count firms used in the estimation by industry
		(Run KNN_WLPMIX_GO_FirmObs.do)

	Save the results as KNN_WLPMIX_GO_FirmObs.txt.
		(Used to produce Table 3)

		
	2-7. Choose estimates for each industry without measurement error correction
		(Run KNN_WLPNBD_GO.do)

	Save the results as est_WLPNBD_GO.dta and est_WLPNBD_GO.xlsx.
		(Used to produce Table B.I)


[For each do file for bootstrapped standard errors]
	2-8. Compute bootstrapped standard errors for industries using measurement error correction
		(Run Bootstrap\indXX\BS_indXX.do for industry XX,
			e.g., for industry 13, run Bootstrap\ind13\BS_ind13.do)

	Save the results as BS_indXX_bs_std.txt for industry XX,
			e.g., for industry 13, it is Bootstrap\ind13\BS_ind13_bs_std.txt).
		(Used to produce Table 3)


================================================================================
3. Aggregation of Productivity Measures

[Do files]
	KNN_agg_GO_real.do

[Ado files]
	AggVarVec.ado	// Collapse aggregate variables in panel data 
					// as vectors named original names of the variables.
					// The variables should take the same value for all i 
					// for each time period (such as averages across plants).                             
					// This is used only to save memory use.

[Input files]
	../cpi.xlsx (not included but the data are available at the website of Bank of Japan)
	KNN_data.dta (created in step 1)
	est_WLPMIX_GO.dta (created in step 2)
	est_WLPNBD_GO.dta (created in step 2)
	est_OLST_GO.dta (created in step 2)
	est_FET_GO.dta (created in step 2)
	est_FDT_GO.dta (created in step 2)
	
--------------------------------------------------------------------------------
	3-0-1. Prepare for CPI for real labor cost share, using:
		"cpi.xlsx"

	3-1. Setup variables.
		3-1-1. Read KNN_data.dta.
		3-1-2. Drop unnecessary variables.
		3-1-3. Create the Domar weights.
		3-1-4. Create revenue shares of factor inputs for each plant.
		3-1-5. Read the production function estimates.    
		3-1-6. Create the estimated productivity measures.  

	3-2. Aggregate the estimated productivity measures.
		3-2-1. Prepare variables needed for PL's Aggregate Productivity Growth.
		3-2-2. Calculate BHC Aggregate Productivity Growth. 
		3-2-3. Calculate Entry and Exit terms for BHC.        
		3-2-4. Calculate Technical Efficiency Term.         
		3-2-5. Calculate Reallocation Effect Term for BHC.       
		3-2-6. Calculate Labor Reallocation Term.
		3-2-7. Calculate Capital Reallocation Term.
		3-2-8. Calculate Material Reallocation Term.
		3-2-9. Make Vecors of Aggregate Variables.     
		3-2-10. Calculate Factor-input Wedges.        

	3-3. Output the aggregated productivity measures. 
		3-3-1. Restore collapsed aggregated variables.         
		3-3-2. Drop the first-year observations.         
		3-3-3. Create Reallocation Effect and Net Entry.
		3-3-4. Check the reallocation decompositon.      
		3-3-5. Print rough output tables.               

	Save the results as PLAPG_GO.dta and PLAPG_GO.raw.
		(Used to produce Table 1, Table 4, Figure 1, Table A.I, Table B.I, and Table B.II)

	
================================================================================
4. Firm-level Data Setup

[Do files]
	KNN_Kikatsu.do

[Input files]
	../matching code_listed.xlsx	(not included due to restricted access)
	../BS2010 with ind_jip2006.dta	(not included due to restricted access)
	
--------------------------------------------------------------------------------
	4-1. Prepare for listed firm code matching, using:
		"matching code_listed.xlsx"

	4-2. Read a new data set from 
		../MC2010 by ind_jip2006.dta

	4-3. Adjustments for consistent data

	4-4. Calculate statistics for identifying Zombies

	4-5. Calculate other financial statistics 

	4-6. Setup multi-year restriction

	4-7. Create zombie indexes with multi-year restriction

	4-8. Produce tables

	Save the results as KNN_Kikatsu.dta
		(The log file is used to produce Table 5 and Table 6)

================================================================================
5. Merging Plant-leval and Firm-level Data

[Do files]
	KNN_SZ_data_GO_ind.do

[Input files]
	../cpi.xlsx (not included but the data are available at the website of Bank of Japan)
	KNN_data.dta (created in step 1)
	KNN_Kikatsu.dta (created in step 4)
	
--------------------------------------------------------------------------------
	5-1. Set up the plant-level data
		5-0-1. Prepare for CPI for real labor cost share, using:
				"cpi.xlsx"
		5-1-1. Read KNN_data.dta
		5-1-2. Drop unnecessary variables.
		5-1-3. Read the production function estimates

	5-2. Produce tables for Table 7
	
	5-3. Merge firm-level data and plant-level data 

	5-4. Keep the data only for 1989 and 1999 

	Save the results as KNN_SZ_data_GO.dta
		(The log file is used to produce Table 5 and Table 7)

================================================================================
6. Propensity-score Matching Estimation

[Do files]
	KNN_TE_GO_ind.do

[Input files]
	KNN_SZ_data_GO.dta (created in step 5)
	
--------------------------------------------------------------------------------
	6-1. Read KNN_SZ_data_GO.dta

	6-2. Calculate the propensity-score matching estimates

	Save the results as KNN_SZ_data_TE.dta
		(The log file is used to produce Table 8 and Table C.I)

================================================================================
7. Model Cariblation and Counter-factual Exercise

[Do files]
	KNN_entry_exit.do
	KNN_SZP_GO_ind.do
	
[Input files]
	KNN_SZ_data_TE.dta (created in step 6)
	
--------------------------------------------------------------------------------
	7-0. Calculate plants' exit rate using KNN_entry_exit.do
			The results need to be typed in KNN_SZP_GO_ind.do (parameter lambda).
		
	7-1. Read KNN_SZ_data_TE.dta

	7-2. Calibrate the model parameters

	7-3. Conduct a counter-factual exercise
		7-3-1. Calculate alternative "tau"s.
		7-3-2. Adjust "tau"s so that the exercise is cost-neutral.

	7-4. Calculate Aggregate Productivity Growth (APG) under the counter-factual scenario
		7-4-1. Calculate APG without production function estimates
		7-4-2. Calculate the technical efficiency term
		7-4-3. Calculate the reallocation terms.         
		7-4-4. Create reallocation effects and net entry.
		7-4-5. Check the reallocation decompositon.      
		7-4-6. Calculate APG without entry and exit.                
		7-4-7. Show the results in annualized rates.                

	Save the results as KNN_SZP_GO_ind.dta
		(The log file is used to produce Table 9 and Table 10,
		 by choosing different zombie indexes)

================================================================================
[List of files to produce Tables and Figure 1]

The list is in the order of the appearance in the text.

--------------------------------------------------------------------------------
[Main Text]

Table 1: Growth in the manufacturing sector and the whole economy 
	PLAPG_GO.raw
	JIP2011.xlsx

Table 2: Summary statistics (full sample)
	KNN_SZ_data_GO_ind_adjK14.txt

Table 3: Production function estimates
	est_WLPMIX_GO.xlsx	
	KNN_WLPMIX_GO_FirmObs.txt
	Bootstrap\indXX\BS_indXX_bs_std.txt for industry XX, 
		e.g., for industry 13, it is Bootstrap\ind13\BS_ind13_bs_std.txt).

Table 4: Decomposition of APG in Japan in the 1980s and 1990s 
	PLAPG_GO.raw

Figure 1: The reallocation effect (RE) and its decomposition 
		(in percent of aggregate value added)
	PLAPG_GO.raw

Table 5: Shares of zombie firms and plants (percent, in 1999) 
	KNN_Kikatsu.txt
	KNN_SZ_data_GO_ind_adjK14.txt

Table 6: Financial indicators for zombie and non-zombie firms
	KNN_Kikatsu.txt

Table 7: Resource reallocation of zombie and non-zombie plants
	KNN_SZ_data_GO_ind_adjK14.txt

Table 8: Propensity-score matching estimation of the effect of zombie lending 
	KNN_TE_GO_ind_adjK14.txt

Table 9: Macroeconomic costs of zombie lending 
		(in percent of aggregate value added, from 1989 to 1999, annualized) 
	KNN_SZP_GO_ind_adjK14_flg_z12c3.txt

Table 10: Macroeconomic costs of zombie lending under different definitions 
			of zombie lending (in percent of aggregate value added, 
			from 1989 to 1999, annualized) 
	KNN_SZP_GO_ind_adjK14_flg_z12.txt	(by choosing flg_z12 as the zombie index)
	KNN_SZP_GO_ind_adjK14_flg_z12c2.txt	(by choosing flg_z12c2 as the zombie index)
	KNN_SZP_GO_ind_adjK14_flg_z12c3.txt	(by choosing flg_z12c3 as the zombie index)
	KNN_SZP_GO_ind_adjK14_flg_z12c4.txt	(by choosing flg_z12c4 as the zombie index)
	KNN_SZP_GO_ind_adjK14_flg_z15.txt	(by choosing flg_z15 as the zombie index)


[Appendix A to E]

Table B.I: Production Function Estimates by Estimator 
	est_WLPMIX_GO.xlsx	// the WLP estimator with measurement error correction
	est_WLPNDB_GO.xlsx	// the WLP estimator without measurement error correction
	KNN_OLST_GO.xlsx	// the OLS estimator 
	KNN_FET_GO.xlsx		// the FE estimator 
	KNN_FDT_GO.xlsx		// the FD estimator 

Table B.II: Reallocation effects using different production function estimates
	PLAPG_GO.raw

Table C.I: Share of zombie plants by propensity score
	KNN_TE_GO_ind_adjK14.txt

Table E.I: Calibrated parameters 
	KNN_SZP_GO_ind_adjK14_flg_z12c3.txt
	JIP2011.xlsx


[Online Appendix A to C]

Table A.I: Reallocation terms of APG using PL's and BHC's approach
	PLAPG_GO.raw

================================================================================
[List of miscellaneous files]

	List_est_WLPMIX_GO.xlsx	// List which estimator was used for the WLP estimator 
				//   with measurement error correction.
	List_est_WLPNDB_GO.xlsx	// List which estimator was used for the WLP estimator 
				//   without measurement error correction.

================================================================================