Does relative age affect fame? Ask Wikipedia

ABSTRACT

We analyze whether age relative to school classmates affects the likelihood of becoming famous. We measure such likelihood as the ratio of Wikipedia entries to births, by state and date of birth, among people born in 1969–1988 in the US. Using a reduced-form Regression Discontinuity Design, we find evidence that men born after the Kindergarten cutoff date (i.e., relatively older) are roughly between 5 and 10 percent more likely to become famous, by Wikipedia standards, in comparison to those born before the cutoff (i.e., relatively younger). We don’t find evidence of a similar effect for women.

KEYWORDS:

• relative age
• school-starting age
• Wikipedia
• fame

JEL Classification:

• I21
• J11

Acknowledgement

We are grateful for the valuable comments of two anonymous reviewers and the excellent research assistance of Jun Ho Choi. All errors and omissions are our responsibility.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 A partial summary can be found in Peña (2017).

2 That is the approach taken in the study of professional hockey players (Barnsley and Thompson 1988) made famous by Malcolm Gladwell in Outliers.

3 The kind of publication bias alluded here has been documented by Franco, Malhotra, and Simonovits (2014) in a more general context.

4 Wikipedia has been used to study economic phenomena, such as the determinants of content growth (Aaltonen and Seiler 2016), incentives to contribute (Chen and Yeckehzaare 2020), and the effects of content on economic variables (Hinnosaar et al. 2019).

5 Alternatively, we could fix the distribution of performance and assume relative age produces gains of different magnitudes depending on the baseline level of performance. The insights would be the same.

6 Our purpose is to show that the relationship between the threshold $a$ and the odds ratio can be increasing, decreasing, or constant. That property doesn’t depend on the choice of the magnitude of $\mathrm{\Delta }$. The power-law and exponential distributions clearly illustrate this point. In the case of the normal distribution, we must assume a magnitude of $\mathrm{\Delta }$ in order to compute numerically the odds ratio, but the specific value chosen is inconsequential.

7 Source: https://en.wikipedia.org/wiki/Wikipedia:About (last accessed Oct. 30, 2020).

8 On Wikipedia, ‘notability is a test used by editors to decide whether a given topic warrants its own article. For people, the person who is the topic of a biographical article should be ‘worthy of notice’ or ‘note’—that is, ‘remarkable’ or ‘significant, interesting, or unusual enough to deserve attention or to be recorded’ within Wikipedia as a written account of that person’s life. ‘Notable’ in the sense of being ‘famous’ or ‘popular’—although not irrelevant—is secondary.” Source: https://en.wikipedia.org/wiki/Wikipedia:Notability_(people) (last accessed Oct. 30, 2020).

9 Some of those differences or biases have been documented by Reagle and Rhue (2011), Greenstein and Zhu (2012, 2018), and Graham et al. (2014).

10 The download was done on Oct. 18, 2020. We used the Python package ‘qwikidata’ to collect the information from Wikidata through SPARQL queries. For individuals whose exact date of birth is unknown, Wikidata queries conducted this way return January 1 as their birth date. To correctly categorize those cases as missing the birth date we used the Python packages ‘requests’ and ‘bs4.’ We web-scraped the Wikidata webpage of each person whose birth date was originally retrieved as January 1. We then verified whether, according to the Wikidata webpage, the month and day of birth were actually January 1.

11 Source: http://www2.nber.org/data/vital-statistics-natality-data.html (last accessed Oct. 30, 2020).

12 Forty-two states have births of each gender in all 7,305 days. The state with the fewest days with births is Wyoming, which has 7,261 days (99.4%) with female births and 7,280 days (99.7%) with male births.

13 We found only one person (a woman born in Alaska in 1973) with a Wikipedia entry whose date and state of birth were not matched by the birth data. This is likely due to the random sampling of birth certificates in the earlier years of the period covered by the sample.

14 The missing information in Bedard and Dhuey (2012) for the school years 1989–1994 and for Alaska and Hawaii was complemented with information from Bedard and Dhuey (2007).

15 A similar approach was taken by Peña (2019) using the same birthdate data combined with prison records to analyze relative age effects on incarceration rates in Florida.

16 Educational attainment and marital status of the mother are observed only on the more recent years of the sample. Using that information would imply a reduction the statistical power of our analysis.

17 The small magnitudes we find are roughly in line with the results of Dickert-Conlin and Elder (2010), who use birth certificates from 1999–2004 to analyze intentional timing and parental sorting around the cutoff date for school eligibility in the US, and find no evidence of such behaviors.

18 For instance, in the unweighted regressions California and Wyoming are given the same importance on each date. In contrast, in the weighted regressions their importance is proportional to the number of births in each state on that particular date.

19 For evidence on this type of behavior in the social sciences see Franco, Malhotra, and Simonovits (2014).