1. The be_founders database

1.1. Data

The database has rich and detailed information about a large sample of entrepreneurs who (co-)founded start-ups in Belgium. It contains information on three main areas:

1. Individual characteristics (name, gender, city, etc.)

2. Education history:

• Educational institution

• Study program

• Start/end date

3. Work experience:

• Employer / Start-up (co-)founded

• Job title

• Start/end date of the position

1.2. Sample

We built the sample in three steps. First, we obtained the list of all Belgian start-ups listed in Crunchbase, and retrieved their founders. Second, we collected information on the work and education history of the founders from the business directory LinkedIn. Third, we carefully parsed, cleaned and disambiguated the name-, work- and education-related fields (see Section 3, Methodology) in order to reduce the incidence of spelling variations and heterogeneity in entries referring to the same entity.

After extensive cleaning and disambiguation, the sample covers 774 entrepreneurs which (co-)founded 1211 companies. As well, the dataset includes 6526 work experience entries (employer/founded firm + job title) and 2295 education entries (educational institution + study program).

1.3. Database Structure

We organized the database into separate tables. The current version (January 2022) consists of 41 tables, whose names have the following format:

<N>_<xx>_<name>

where <N> is a letter that indicates the table’s family, <xx> a number and <name> is a unique string that reflects the table’s content. There are four families (with their respective leading letter in parentheses):

• Raw data (R)

• Founders’ information (F)

• Educational background (E)

• Work history (W)

• Ready-to-use (U)

• Other tables (T)

As usual, the tables contain id’s (one or more, local/foreign) that allow to link them to each other. See Section 4 (Tables) for a complete list of the tables and their contents.

All tables are located in the folder ‘./04_tables’, in csv format.

Note

be_founders will soon be available as a self-contained SQLite database engine.

1.4. Links to external databases

To disambiguate each field we used libraries and dictionaries of clean names from external databases (e.g., lists of company names, university names, job titles). Besides helping with the harmonization step, such lists allowed us to link our data to external data sources. Table 1.1 below illustrates the database’s key fields and the external source we matched them to.

Table 1.1 Field linking and mapping

Field	Parsed/ cleaned	Disamb.	Linked to: db name + (id)	Mapped into
Person name	Yes	Yes	Crunchbase (cb_person_uuid), LinkedIn (permalink)
School/ university name	Yes	Yes	Orbis (bvd_id, duo_bvd_id, guo_bvd_id), ETER (id), Carnegie classification (UNITID), Webometrics (string name) 1
Study program	Yes 2	Yes		ISCED levels and classification
Company name	Yes	Yes	Crunchbase (org_uuid), Orbis (bvd_id, duo_bvd_id, guo_bvd_id), Compustat (gvkey), CRSP (permno, permco, CUSIP)
Job title	Yes	Yes	O*NET (O*NET-SOC code)	Functional areas, 3 job ranks 4

Notes: ¹ Webometrics data does not provide a unique identifier; ² we cleaned the strings, but parsed only the study type (e.g. bsc, msc, phd, etc.); ³ e.g. business and marketing, R&D and engineering, etc.; ⁴ e.g. top management, sub-management, non-management