A Conversation with 590 Nascent Entrepreneurs

by Jeffrey R. Campbell and Mariacristina DeNardi

published in Annals of Finance 2009

This file (which should be named cdn2.zip) contains the data and source code necessary for the paper's replication.

Original Data

The data used for this paper comes from the first Panel Study of Entrepreneurial Dynamics (or PSED I). A PSED home page at Clemson University describes their collection and structure. That site distributes an SPSS formatted data set, which we converted to Stata format for our work. Our paper uses questions from only the first wave. We used the questionnaire from that interview to guide us through the available information. We also found the questionnaire from the initial screening interview to be helpful.

As a precaution against the Clemson website someday being taken down, this replication file contains copeis of both questionnaires in the /data directory. That directory also contains the data set in Stata format.

There currently exists a second PSED home page at the University of Michigan. This includes codebooks and data from the second PSED, which we do not use in this paper. However, we were unable to locate the PSED I data set there.

Required Software

Stata SE, Version 10.0 or higher with the listtex module installed We use listtex to directly export Stata results into LaTeX tables and macros.
Matlab, Release 2008b or higher
pdflatex with the packages

natbib

Specifies author/date reference style.

hyperrref and hypernat
Enables hyperlinks in the .pdf file.

setspace
Makes small changes to the default spacing. We do not use any of this package's options, but the paper looks a little better with it.

pgf/tikz

Used to include .pdf files created by Matlab and to carry out simple mathematical calculations of numbers in the text.

pgfplots
Enables the \pgfmathprintnumber command for formatting numbers included in the text.

rotating
Enables the \sidewaystable command.

bbding
Makes the bbding dingbat font available. Used to place an envelope next to the corresponding author's name.

todonotes
Creates the boxed margin and inline notes with details and source information.(Requires pgf/tikz

versions
Controls whether or not the source and detail notes appear in the compiled .pdf file.
bibtex
gnu make

Stata does most of the calculations, and we use Matlab only to create the figures. (In retrospect, it would have been better to make these within LaTeX using the pgf and pgfplots packages.) The paper's text is written in LaTeX, which we process with pdflatex and bibtex. We bring the whole enterprise together with a set of makefiles.

Stata and Matlab are both commercial programs, available from their respective vendors. Any self-respecting research institution should have licenses for their use. The last three programs are all part of the standard Linux installation.

We include todonotes.sty in the replication file because it is not yet part of the standard TeXLive distribution of TeX and its friends. If you do not have some of the the other required LaTeX packages, we suggest that you or your system administrator download and install the latest TeXLive distribution (for Unix and Mac) or MiKTeX distribution (for Windows). (We have tested the programs with the 2008 release of TeXLive.)

Since we only use gnu make to automate the paper's construction, the paper can be reproduced by running the individual files in sequence by hand. On our Linux machine, the commands to start Stata and Matlab are stata-se and matlab. If these are different on your machine, then you will have to change each directory's makefile appropriately before using make.

Running the programs under Unix

These instructions assume that the relevant Unix machine's hard drive already has a copy of cdn2.zip. If this is not the case, Microsoft Windows users can install WinSCP and use it to transfer the file. To begin, log into the machine with your favorite client and start a terrminal/ssh session. Change your working directory to the directory containing the replication file, and then issue these commands.

mkdir cdn2 (Creates an empty directory for the replication work)
mv cdn2.zip cdn2 (Moves cdn2.zip into the new directory)
cd cdn2 (Changes the working directory to the new directory)
unzip cdn2.zip (Extracts the files for replication.)
make

The last step starts gnu make, which in turn runs all of the Stata and Matlab programs and then processes the results with pdflatex. Upon completion (which should take very little time), the current directory will contain the completed paper, cdn2.pdf.

If something goes wrong in one of the first four steps, then you probably do not have permission to write in the current directory. Contact your system administrator or a local Unix expert for help. An error in the fifth step probably indicates that the machine does not have the required software. If necessary, type Control-C to abort the replication. Then get help from a local expert.

Running the programs under Microsoft Windows

Windows does not come with any flavor of make, so replicating the paper under Windows requires running each of the programs by hand. After unzipping into a suitable working directory,

Run psed.do in the data directory.
Run all of the .do files in the Tables and Figures directories. The order in which these are run is irrelevant.
Run all of the Matlab .m programs in the Figures directory. Again, these can be run in any order.
Open a command prompt window and process the paper once with the command pdflatex cdn2. This should generate lots of complaints about missing references and cross references.
Create the paper's bibliography with the command bibtex cdn2
Repeat step 4 three more times. This corrects the paper's references, cross references, and hyperlinks.

Auditing the paper's results

To ensure accuracy, we typed no quantitative result into the paper by hand. Instead, every number is generated by a Stata program and written to a text file. LaTeX reads these files and places the numbers in the appropriate places. If the number is spelled out in the text (as at the beginning of a sentence), then the text contains a margin note with its automatically generated counterpart. This scheme guarantees that there exists an audit trail for every result. To follow it, you can take one of two approaches.

Most numbers reported in the text are associate with a table or figure. Each of these has a note indicateing the Stata and Matlab programs used in its creation. You can examine those programs directly.
Since all numbers are automatically generated, you can open the LaTeX source file cdn2.tex and find the code generating the number. Inline numbers are created from LaTeX macros (e.g. \amacro. The first part of the macro indicates which Stata file generated it. (So \obsXX was generated by obs.do.)

Manifest

annals-cover.jpg: Image of Annals of Probability cover used above and in the text to encourage readers to access the published version.
cdn.bib: BibTeX bibliography file.
cdn2.tex: LaTeX file with paper's text.
makefile: Project's master makefile.
todonotes.sty: LaTeX file for todonotes package.
/data/erc_q1.pdf: Questionnaire for the PSED I's first wave interview.
/data/erc_sc.pdf: Questionnaire for PSED I's screening interview.
/data/ercw14s.dta: Stata version of PSED I data set.
/data/makefile: Makefile for /data subdirectory.
/data/psed.do: Stata program for initial data manipulations.
/Figures/demofig_fe.do: Stata program for creating results presented in the "Women" panel of Figure 1.
/Figures/demofig_me.m: Matlab program for creating the "Women" panel of Figure 1.
/Figures/demofig_ma.do: Stata program for creating results presented in the "Men" panel of Figure 1.
/Figures/demofig_ma.m: Matlab program for creating the "Men" panel of Figure 1.
/Figures/hcfig_fe.do: Stata program for creating the results presented in the "Women" panel of Figure 2.
/Figures/hcfig_fe.m: Matlab program for creating the "Women" panel of Figure 2.
/Figures/hcfig_ma.do: Stata program for creating the results presented in the "Men" panel of Figure 2.
/Figures/hcfig_ma.m: Matlab program for creating the "Men"panel of Figure 2.
/Figures/makefile: Makefile for /Figures subdirectory.
/Figures/moneyfig_fe.do: Stata program for creating the results presented in the "Women" panel of Figure 4.
/Figures/moneyfig_fe.m: Matlab program for creating the "Women" panel of Figure 4.
/Figures/moneyfig_ma.do: Stata program for creating the results presented in the "Men" panel of Figure 4.
/Figures/moneyfig_ma.m: Matlab program for creating the "Men" panel of Figure 4.
/Figures/parentfig_fe.do: Stata program for creating the results presented in the "Women" panel of Figure 3.
/Figures/parentfig_fe.m: Matlab program for creating the "Women" panel of Figure 3.
/Figures/parentfig_ma.do: Stata program for creating the results presented in the "Men" panel of Figure 3.
/Figures/paentfig_ma.m: Matlab program for creating the "Men" panel of Figure 3.
/Tables/allhours.do: Stata program for creating bottom panel of Table 9.
/Tables/allinvest.do: Stata program for creating Table 10.
/Tables/anticipatedSize.do: Stata program for creating Table 6.
/Tables/conception.do: Stata program for creating Table 8.
/Tables/externalFundsPartner.do: Stata program for creating Table 12.
/Tables/externalFundsSolo.do: Stata program for creating Table 11.
/Tables/female.do: Stata program for calculating female oversample sizes and gender breakdown reported in the text.
/Tables/fundsNeeded.do: Stata program for creating Table 14.
/Tables/incomeresponse.do: Stata program for calculating response counts to income and wealth questions reported in the text.
/Tables/industry.do: Stata program for creating Table 2.
/Tables/LegalForm.do: Stata program for creating Table 4.
/Tables/makefile.do: Makefile for /Tables subdirectory.
/Tables/obs.do: Stata program for creating Table 1.
/Tables/partners.do: Stata program for creating Table 5.
/Tables/representative.do: Stata program for calculating gender composition, educational attainment, and sample count for the representative NE sample reported in the text.
/Tables/respondenthours.do: Stata program for creating top panel of Table 9.
/Tables/sponsorship.do: Stata program for creating Table 3.
/Tables/stage.do: Stata program for creating Table 13.
/Tables/stageTwo.do: Stata program for tabulating responses to questions about business accomplishments discussed near Table 13.
/Tables/TimeUse.do: Stata program for creating Table 7.
/Tables/wealth.do: Stata program for calculating mean and median wealth for the comparison group, as reported in the text.

Last modified on June 17, 2010