arlionn / technological-innovation-resource-allocation-and-growth-replication-kit

This repository provides the replication code and data for Kogan, L., Papanikolaou, D., Seru, A. and Stoffman, N., QJE 2017.

Replication Code and Data for Kogan, L., Papanikolaou, D., Seru, A. and Stoffman, N., QJE 2017

This package provides the replication code for the main results in Kogan, L., Papanikolaou, D., Seru, A. and Stoffman, N., 2017. Technological innovation, resource allocation, and growth. Quarterly Journal of Economics, 132(2), pp. 665-712. The paper is available at https://academic.oup.com/qje/article/132/2/665/3076284.

The folder ./Code contains all programs, while the folder ./Data includes all needed input data files.

Code

The folder ./Code includes code files of two types of programs: Stata and Matlab.

Code files using Stata:

Prerequisite:

In order for the code to run correctly, you need the following packages installed in Stata:

• winsorizeJ: Winsorize variables using breakpoints that vary by year. (A winsorizeJ.ado file is provided in the folder ./Code)

• cluster2: 2D clustered standard errors. (A cluster2.ado file is provided in the folder ./Code)

• ivreg2: IV regression with more features than STATA's ivreg. (Install the package using command "ssc install ivreg2")

• reghdfe: Regression with multiple levels of fixed effects. (Install the package using command "ssc install reghdfe")

• estout: Output Stata tables in tex format. (Install the package using command "ssc install estout")

File description:

• FilterReturnsCreateFirmMeasures.do: Creates the firm-level innovation measure using patent and stock return data

• PatentValueCites.do: Descriptive statistics for the patent-level measure, including robustness for alternative distributions (Table 1 in the paper and Table A.6 in the Online Appendix); relates the estimated patent values to forward citations (Table 2 in the paper and Table A.7 in the Online Appendix)

• CreateFirmSample.do: Creates the data for the firm-level regressions

• FirmSummaryStats.do: Descriptive statistics for the firm-level measure (Table 3 in the paper)

• FirmProfitsRegressionSM.do: Firm Profits and innovation (Panel a of Table 4 in the paper)

  FirmOutputRegressionSM.do: Firm Output and innovation (Panel b of Table 4 in the paper)

  FirmReallocationRegressionSM.do: Firm Capital/Labor and innovation (Panels c and d of Table 4 in the paper )

  FirmTFPRegressionSM.do: Firm TFP and innovation (Panel e of Table 4 in the paper)

• FirmProfitsRegressionCW.do: Firm Profits and innovation using citation-weighted patents (Panel a of Table 5 in the paper)

  FirmOutputRegressionCW.do: Firm Output and innovation using citation-weighted patents (Panel b of Table 5 in the paper)

  FirmReallocationRegressionCW.do: Firm Capital/Labor and innovation using citation-weighted patents (Panels c and d of Table 5 in the paper)

  FirmTFPRegressionCW.do: Firm TFP and innovation using citation-weighted patents (Panel e of Table 5 in the paper)

• FirmProfitsRegressionSMCW.do: Firm Profits and innovation using both our measure and citation-weighted patents (Panel a of Table 6 in the paper)

  FirmOutputRegressionSMCW.do: Firm Output and innovation using both our measure and citation-weighted patents (Panel b of Table 6 in the paper)

  FirmReallocationRegressionSMCW.do: Firm Capital/Labor and innovation using both our measure and citation-weighted patents (Panels c and d of Table 6 in the paper)

  FirmTFPRegressionSMCW.do: Firm TFP and innovation using both our measure and citation-weighted patents (Panel e of Table 6 in the paper)

• PatentValueScatterPlot.do: Plots the cross-sectional relation between forward patent citations and patent market value (Figure 2 in the paper)

• TimeSeriesPlots.do: Produces and plots the aggregate measures of innovation (Figure 4 in the paper)

• AggregateOutput.do: Runs the aggregate OLS regressions and VARs that relate our two innovation indices to output and TFP (The regression results for Figure 5 in the paper and Figure A.3 in the Online Appendix)

Code files using Matlab:

Prerequisite:

In order for the code to run correctly, you need to install the following program as well:

• matlab2tikz: Converts MATLAB®/Octave figures to TikZ/pgfplots figures for smooth integration into LaTeX. (Please go to https://github.com/matlab2tikz/matlab2tikz for detailed installation information)

File description:

• plot_OLS_responses.m: Take the output of AggregateOutput.do and creates Figure 5 in the paper

• csvimport.m: Define the function called in plot_OLS_responses.m which helps import the .csv files

• jbfill.m: Define the function called in plot_OLS_responses.m which helps plot the confidence intervals

Additional Notes:

• There is a minor discrepancy between the value of average acceptance probabilty used in the codes and reported on page 677 (published version of the paper). This minor difference is due to a typo in the published version.

• There are some minor discrepancies between the output that the code generates and the results reported in Table 5 (published version of the paper). These minor differences are due to typos in the published version.

• There are copyright restrictions and file size limitations in posting daily data from CRSP. Please contact the authors below to obtain the code and data needed to replicate Figures 1 and 3 in the paper that use this data.

Contact

Please contact Dimitris Papanikolaou ([email protected]) or Amit Seru ([email protected]) for any questions regarding the codes or data.

Please see the paper for more information on the codes and data. If you use these codes files or data, please CITE this paper as the source.