# Supplementary archive

Supplementary material for **"Entrepreneurship and Behavioral Characteristics"**, a
chapter in *Experiments in Entrepreneurship, Volume 1*, edited by Scott Shane
(Edward Elgar, 2027).

David J. Cooper (University of Iowa) · Saulo Dubard Barbosa (Emlyon Business School) ·
Krista J. Saral (University of North Carolina at Charlotte) ·
Marie Claire Villeval (CNRS, GATE, University of Lyon)

**Start with `index.html`.** Unzip the archive and open that file in any browser. Nothing
is loaded from the internet and no software is required to read it.

---

## What is here

| | |
|---|---|
| `OPEN ME — index.html.txt` | A one-page signpost for anyone who unzips this and is not sure where to start. |
| `index.html` | **Landing page.** Every measure collected outside the laboratory games, with group means and both classification coefficients, filterable. Links to the methods page. |
| `methods.html` | **Methods appendix.** Design history, instruments, session log, power, and known limitations, measure by measure. |
| `methods-appendix.md` | The same methods text as plain markdown. |
| `chapter_analysis.do` | **Stata. Reproduces every number in the chapter.** |
| `data/` | The three de-identified datasets the do-file reads. |
| `instructions/` | What subjects actually saw, screen by screen. |
| `programs/` | The z-Tree and oTree source that generated those screens. |

## Reproducing the results

Unzip the archive, open Stata in this folder, and run:

```stata
do chapter_analysis.do
```

It writes `chapter_analysis.log`. Nothing else needs setting. The `DA` global at the top
already points at `data/`; change it only if you move the datasets somewhere else.

Written for **Stata 18 SE**. It uses `dprobit`, which still ships with Stata 18.

⚠ **All three datasets are needed, and no single one reproduces the chapter.**
`ent_djc_3.dta` is the cross section, `ent_djc_4.dta` the 2018 follow-up, and
`ent_final2.dta` supplies the Phase 3 online-experiment measures only. `ent_djc_3.dta` is
dated 26 December 2017 and predates the second wave's online experiment, so it carries 76
observations on those measures where 95 exist. The `p3_fill` program in the do-file merges
them in.

The eyes test needs more than a merge. `numcorrect_eyes` differs between `ent_djc_3.dta`
and `ent_final2.dta` for **59 of the 76** subjects present in both, while every other
online variable agrees exactly. `ent_final2.dta` takes its scores from the raw oTree
export, and for the 47 French-language subjects these were re-scored item by item against
the answer key in the oTree source and match. The origin of the `ent_djc_3.dta` values is
not known, so eyes scores are taken from `ent_final2.dta` throughout.

### Which block produces which table

The do-file keeps several specifications per outcome, because the analysis went through
four rounds. **Tables 3, 4 and 5 come from the matrix blocks**, which print as `NT`, `T4`
and `T5` in the log. The loose `dprobit` and `tobit` lines earlier in each section are
earlier takes, kept for transparency, and they do **not** match the printed tables.

## The data

| File | Observations | Variables |
|---|---|---|
| `ent_djc_3.dta` | 131 | 58 |
| `ent_djc_4.dta` | 131 | 62 |
| `ent_final2.dta` | 131 | 54 |

The analysis sample is **122**. The do-file gets there by keeping subjects with a
non-missing classification and dropping one record, flagged as `repeat_2017`, belonging to
a subject who took part in both waves. Trust runs on 102 because session 1 used a
different instrument.

### De-identification

These files are **not** the originals. They were built by whitelist, not by blacklist:
only the 62 variables the analysis actually reads were kept, out of 514 in the source
files. Everything identifying therefore disappeared by construction rather than by
remembering to name it.

Removed, among 452 dropped variables: IP addresses, oTree participant codes and session
codes, z-Tree session timestamps, exact survey start and end times, nationality (17
values, 11 held by a single subject), the Emlyon programme code, and every free-text
field, some of it first-person and detailed.

Two further changes. `subject_id` was replaced by a fresh sequential id under a seeded
permutation, and the crosswalk is **not** in this archive and will not be published,
because the original ids map to paper participant records that carry names. `session` was
collapsed to 1 for session 1 and 2 for every other session, which preserves every result
since the analysis only ever tests `session != 1`.

**Kept deliberately:** `age` is exact rather than banded, because it is a continuous
control in Tables 3, 4 and 5 and banding it would change the published coefficients. Ten
age values are unique to one subject. `female`, `wave2`, and the coarse education and
income indicators are all controls and all coarse. On 122 subjects, age and gender without
nationality, programme, session date or free text is a weak re-identification route. It is
a judgement call and it is recorded here so it can be revisited.

## Instructions and surveys

`instructions/` holds the screens subjects saw, in order, and the survey instruments. All
are the English versions; the sessions were run in French and English and the French
instruments are available on request.

| File | What it is |
|---|---|
| `lab-instruction-screens-english.docx` | The laboratory session screen by screen, with the experimenter's running notes and the z-Tree program each block ran |
| `online-experiment-screens-english.docx` | The online experiment, all five incentivized tasks |
| `survey-I-intake-english.pdf` | Phase 1 intake survey, wave 1 first sitting. Self-efficacy, entrepreneurial intent, job-choice risk, demographics |
| `survey-II-english.pdf` | Phase 1, wave 1 second sitting. Quick risk and patience, Big Five (TIPI), grit, entrepreneurial history |
| `survey-followup-2018-english.pdf` | The 2018 follow-up, two years after the original session |

Wave 2 (sessions 7 and 8, May 2017) took a single combined instrument administered in
French only, which is why it is not listed here. Its coverage is documented in
`methods.html`; six items from Survey I were dropped from it and have no 2017
observations.

### The experimental programs

The **z-Tree treatment files** for the five laboratory games and the **oTree application
source** for the five online tasks are not in this deposit, to keep it small and because
parts of them cannot be redistributed (below). They include the specification written for
the developer, with the exact parameter values for every task.

**They are available on request. Write to Krista Saral, `ksaral@charlotte.edu`.**

### ⚠ Third-party instruments are not included

Two of the instruments are copyrighted and we have no right to redistribute them, so they
have been removed from everything in this archive. Everything else is ours.

| Instrument | Where it was removed from | Source |
|---|---|---|
| **Raven's Progressive Matrices**, the market entry skill task | 2 screens in the lab instructions, and the 13 item images in the z-Tree program | Raven, Raven & Court (1998) |
| **Reading the Mind in the Eyes**, online Task 5 | 8 screens in the online instructions, the 37 item images in the oTree app, and the item and answer-key PDFs | Baron-Cohen et al. (2001) |
| **Cognitive Style Index** (Allinson & Hayes), the 38 `tf` items | the survey instruments | Allinson & Hayes (1996), licensed |

Each removed screen is replaced in the document by a bracketed note saying what was there.
The copies of the programs sent on request have the same material removed, so the oTree
`eyes_mind` app and the z-Tree `IQ2.ztt` will not run without obtaining the instruments
from their publishers and restoring the filenames the code expects. All other tasks run as
they are.

## Citing this archive

> Cooper, D. J., Dubard Barbosa, S., Saral, K. J., & Villeval, M. C. (2026).
> *Entrepreneurship and Behavioral Characteristics: supplementary measures and analysis.*
> Available on the website of Krista J. Saral, University of North Carolina at Charlotte.

Licensed **CC BY 4.0**, covering our own text, code, data and tables. The third-party
instruments described above are not included and are not covered by that license.