Maybe the World Cup Wasn’t the Best Example

In our recent New York Times column, we talked about what makes people good at what they do.

As one example, we conjectured (based on some academic work done by others) that people born in the early months of the year would be overrepresented on World Cup rosters. The underlying theory is that in 1997, FIFA made January 1 the cutoff for determining ages in all international soccer competitions. If this rule had an important impact in determining who made the national youth soccer teams, then these early selection rules would play out to more long run success at the highest levels of soccer. The academic evidence is that these national teams are overwhelmingly made up of players born early in the calendar year, even on the age 21 and under teams, where a few months of physical development isn’t likely to make a big difference. A commenter on our blog, Bill Loyd, has done some hard work to gather data and argues that for past World Cups and for a few of the 2006 squads that he found, he doesn’t see the pattern we predict.

Why might this be the case? For the earlier World Cups, it might not be very surprising that no pattern is there because the FIFA rule didn’t come in until 1997. More fundamentally, the FIFA selection rules and the rules that different countries use for play within the county differ.

For instance, as many readers have emailed us, in the U.S., the age cutoffs tend to be in the summer. In Germany, the within country age cutoff is August 1. Thus, in soccer there are two different competitive pressures at work: one pushing towards more players born in the early months and the other towards more players in the later months. Much of the study of birth-date timing focuses on the cutoff rules within countries, virtually all of them finding important effects.

In light of this difference between FIFA and country rules, the example we gave of the World Cup might not have been the best one, even though the age effect is very strong in the national youth squads that feed many World Cup teams.

This shouldn’t distract from the important fact that the evidence in the literature overwhelming supports the basic point — that across many activities, you can identify long-term effects of essentially arbitrary age cutoffs early in life.

Perhaps a better example than the World Cup would have been the N.H.L. Here is one graph that I found on the web of the birth month of NHL hockey players versus Canadians and Americans more generally:

The black-and-white dots are the NHL players, who are much more likely to be born in January and February and much less likely to be born September-December. This is the sort of pattern that appears over and over in these sorts of studies.

Some other readers have offered a clever, very Freakonomics-y alternative explanation for these age patterns: the parents are lying about their child’s birthday. If the parents want the kid to be a star, they take an older kid and change his date of birth to make him eligible to play with younger children. While I don’t think this is actually the primary reason for what people find in these studies, is definitely worth thinking about.

TAGS:

Leave A Comment

Comments are moderated and generally will be posted if they are on-topic and not abusive.

 

COMMENTS: 118

View All Comments »
  1. zbicyclist says:

    Beach wrote: >

    Thanks! That was quite interesting (earlier surnames like A, B, have better success in economics, a field where joint authors are typically listed alphabetically).

    Luckily I’m not an economist and ZBicyclist isn’t my real name.

    Thumb up 0 Thumb down 0
  2. zbicyclist says:

    Beach wrote: >

    Thanks! That was quite interesting (earlier surnames like A, B, have better success in economics, a field where joint authors are typically listed alphabetically).

    Luckily I’m not an economist and ZBicyclist isn’t my real name.

    Thumb up 0 Thumb down 0
  3. syhu says:

    I would like to provide one more piece of evidence on this discussion using birth month data from baseball players in Taiwan.

    In Taiwan, the school year begins from September. All kids born after September the 1st will not be able to enroll the elementary school until the next academic year. The theory would suggest that the birth month of Taiwan’s atheletes is more concentraed in September, October, and November.

    The following is the distribution of the birth month of Taiwan’s professional baseball players (6 teams in total). You can see that the number of players born at the first three month (September, October, and November) of the school year is 49, while the number from the last three month (June, July, August) is only 26 players.

    Birth month / Number of players
    Jan. 12
    Feb. 14
    Mar. 20
    Apr. 7
    May 9
    June 8
    July 8
    Aug. 10
    Sept. 15
    Oct. 22
    Nov. 12
    Dec. 11

    Thumb up 0 Thumb down 0
  4. syhu says:

    I would like to provide one more piece of evidence on this discussion using birth month data from baseball players in Taiwan.

    In Taiwan, the school year begins from September. All kids born after September the 1st will not be able to enroll the elementary school until the next academic year. The theory would suggest that the birth month of Taiwan’s atheletes is more concentraed in September, October, and November.

    The following is the distribution of the birth month of Taiwan’s professional baseball players (6 teams in total). You can see that the number of players born at the first three month (September, October, and November) of the school year is 49, while the number from the last three month (June, July, August) is only 26 players.

    Birth month / Number of players
    Jan. 12
    Feb. 14
    Mar. 20
    Apr. 7
    May 9
    June 8
    July 8
    Aug. 10
    Sept. 15
    Oct. 22
    Nov. 12
    Dec. 11

    Thumb up 0 Thumb down 0
  5. dkane says:

    Allow me to be the first to connect this discussion to the full World Cup data for 2006. You can get the raw data from the official site. You can find a spreadsheet with the raw data plus my month calculations here.

    Comments:

    1) There is some minimal evidence for a birth-month effect. 74 World Cup players have birthdays in January, about 12 more than the 62 out of 736 that we might expect by chance. November (54) and December (56) are less common that we might expect.

    2) Splitting the sample into first part of the year (379) versus second part (357) is also consistent with the birth-month claims (ignoring different age cut-offs in different countries. But the percentage breaks (51.5% versus 48.5%) are hardly impressive. In an e-mail dialogue (before looking at this data), I had offered to bet Dr. Levitt $500 to the charity of his choice versus a signed copy of Freakonomics that fewer than 52% of World Cup players would be born in the first half of the year. He (wisely) declined to bet.

    3) Dr. Levitt writes:

    In Germany, the within country age cutoff is August 1. Thus, in soccer there are two different competitive pressures at work: one pushing towards more players born in the early months and the other towards more players in the later months.

    Interesting theory. Here is the data for the German team:

    Jan 1
    Feb 2
    Mar 0
    Apr 2
    May 3
    Jun 3
    Jul 0
    Aug 2
    Sep 2
    Oct 1
    Nov 7
    Dec 0

    23 players is too small a sample for much of anything, but there is no reason here to think that an August 1 cut-off in Germany affects anything.

    4) I need to do some more analysis, but my guess is that the very small effect is caused by a phenomenon similar to hockey. I would wager that some younger players who barely make the team are included because they star on youth teams in part because they are older than their peers. I bet that if we restricted attention to the 331 starting players, the effect would disappear completely.

    Thumb up 0 Thumb down 0
  6. dkane says:

    Allow me to be the first to connect this discussion to the full World Cup data for 2006. You can get the raw data from the official site. You can find a spreadsheet with the raw data plus my month calculations here.

    Comments:

    1) There is some minimal evidence for a birth-month effect. 74 World Cup players have birthdays in January, about 12 more than the 62 out of 736 that we might expect by chance. November (54) and December (56) are less common that we might expect.

    2) Splitting the sample into first part of the year (379) versus second part (357) is also consistent with the birth-month claims (ignoring different age cut-offs in different countries. But the percentage breaks (51.5% versus 48.5%) are hardly impressive. In an e-mail dialogue (before looking at this data), I had offered to bet Dr. Levitt $500 to the charity of his choice versus a signed copy of Freakonomics that fewer than 52% of World Cup players would be born in the first half of the year. He (wisely) declined to bet.

    3) Dr. Levitt writes:

    In Germany, the within country age cutoff is August 1. Thus, in soccer there are two different competitive pressures at work: one pushing towards more players born in the early months and the other towards more players in the later months.

    Interesting theory. Here is the data for the German team:

    Jan 1
    Feb 2
    Mar 0
    Apr 2
    May 3
    Jun 3
    Jul 0
    Aug 2
    Sep 2
    Oct 1
    Nov 7
    Dec 0

    23 players is too small a sample for much of anything, but there is no reason here to think that an August 1 cut-off in Germany affects anything.

    4) I need to do some more analysis, but my guess is that the very small effect is caused by a phenomenon similar to hockey. I would wager that some younger players who barely make the team are included because they star on youth teams in part because they are older than their peers. I bet that if we restricted attention to the 331 starting players, the effect would disappear completely.

    Thumb up 0 Thumb down 0
  7. dkane says:

    Not that anyone is still reading this thread, but a quick calculation reveals that neither the 10% of observations for January nor the 51.5% of observations for the first half of the year is statistically significant.

    There is no significant evidence that birth month affected one’s chances of playing in the 2006 World Cup.

    Will the New York Times be publishing a correction anytime soon?

    Thumb up 0 Thumb down 0
  8. dkane says:

    Not that anyone is still reading this thread, but a quick calculation reveals that neither the 10% of observations for January nor the 51.5% of observations for the first half of the year is statistically significant.

    There is no significant evidence that birth month affected one’s chances of playing in the 2006 World Cup.

    Will the New York Times be publishing a correction anytime soon?

    Thumb up 0 Thumb down 0