The Fall and Rise of the "Gender Gap" in World Class Running
Draft of background document
December 12, 1996, length = 10,000 words
By Steve Sailer, with contributions from Dr. Stephen Seiler's
What are we investigating?
This paper focuses on questions raised by the "gender gap" in running speed:
This last question would seem to be both innately interesting and potentially crucial to the science of physiology. For if the popular hypothesis that there is no biological difference in running capacity between the sexes is true, then standard physiological models will need to be substantially revised.
As so often happens, attempting to disentangle the influences of nature and nurture raises difficult methodological issues. One of the most promising approaches to better understanding whether the gender gap is innate is to track its historical trends. If the gender gap has been rapidly narrowing in recent years, then it might seem prudent to withhold judgment on whether it will continue to exist in the long run. In contrast, if the gender gap has not shrunk lately, despite all the recent encouragement given to women's athletics, this would imply that it's likely the gender gap has a biological component. Finally, if conventional wisdom on the subject was wrong, and the gender gap was actually widening at present, that would raise fascinating implications.
An earlier attempt to examine long term gender gap trends among elite runners "generated a stampede of interest from the media" in 1992. Physiologists Susan Ward and Brian Whipp compared the average rate of improvement in world records for men and women over the previous seven decades and projected that if these trends continued, the men's and women's marathon world record would be equal by 1998 and women recordholders would catch up with men at other distances early in the 21st century. The American public appears to agree: a pre-Olympic U.S. News/Bozell poll of 1,000 adults in May, 1996 found that "66 percent of Americans believe the day is coming when top female athletes will beat top males at the highest competitive levels."
We think Ward and Whipp's use of historical trends is valid, although we will propose a more reliable methodology than focusing solely on world records, which tend to be unreliably rare events.
Who are we examining?
We are not investigating the difference between average men and women, but between world-class elite athletes, men and women at their peaks in terms of conditioning and desire.
Why is running a good source of data on sexual differences?
First, it's important to note that gender gaps differ dramatically between sports. Women have won Olympic gold medals in open competition against men in sports like yachting, equestrian, and shooting. On the other hand, the gender gap in the shot-put is close to 100%. So, any analysis of gender gaps is specific to the sport being studied.
However, running is a particularly intriguing because it might be the closest to a universal sport. For example, more countries enter Athletics (i.e., track & field) in the Olympics than any other sport, and more nations win medals in Athletics than any other sport. (Within Athletics, more nations compete in the running events than in the more specialized field events). Thus, international track and field can be conceived of as a vast laboratory experiment in understanding physiological differences between the sexes.
By examining elite performances in the Olympics and World Championships, we can obtain data on human performance stressed to the maximum. Olympic athletes undergo years of training to reach their utmost potential. No controlled laboratory experiment could elicit this kind of cooperation from its subjects.
Finally, compared to many other sports, quantifying running performance is fairly straightforward. Performance is measured on only a single dimension: time. This makes it easy to mathematically compare performances between sexes.
How do we calculate the gender gap?
We are calculating the gender gap in running in either of two absolutely equivalent fashions:
(Woman's Time / Man's Time) - 100%
or
(Man's Speed / Woman's Speed) - 100%
Thus, if the fastest time in Women's 100 meter finals is 11.00 seconds, and the fastest time in the men's 100 meter finals in 10.00 seconds, the gender gap between the winners is 11.00 / 10.00 - 100% = 10.0%. Alternately, the winning man averaged a speed of 10.00 meters per second, while the winning woman averaged 9.09 meters per second: 10.00 meters per second / 9.09 meters per second - 100% = 10.0%.
Where do we find data?
We've primarily focused on the most prestigious track meets: namely, the quadrennial Olympic Games and the International Amateur Athletic Federation World Championships, which are now held every other year. In some ways, these are superior to laboratory experiments for analyzing physiology. For example, few laboratory research projects could assemble subjects from so many countries (e.g., the 100 meter dash alone at the 1992 Barcelona Olympics brought together competitors from 66 nations for the men's event and 41 nations for the women's event).
We've assembled data from multiple sources.
Which competitions to use:
Whipp and Ward's analysis of the gender gap focused on world records, but there are some obvious weaknesses. Most notably, the number of data points is tiny. Second, world records sometimes are at least as much the result of exceptional environmental conditions as exception performances. For example, Pietro Mennea's world record in the men's 200 meters, which endured from 1979 until Michael Johnson broke it in 1996 was a classic "Beamonesque" record. It was a full quarter of a second (or 1.25%) better than any other time Mennea ever ran. This was because it was set in Mexico City's 7,300 ft. altitude with a tailwind 90% of the legal limit. Making admittedly arguable adjustments for altitude and wind (see below for the tables we use for adjusting for wind and altitude), Mennea's run would appear to be the equivalent of 20.00 seconds at sea level with no wind. In contrast, Carl Lewis' 19.80 into a headwind at the 1984 Los Angeles Olympics would appear to be the equivalent of a 19.52 under Mennea's conditions. Yet, it was Mennea's efforts that appeared on the record books for over 16 years, making it appear, according to Ward and Whipp's methodology, that during that decade and a half that men were not getting faster at the 200 meters.
Even though we are dubious about relying too heavily on world records for investigating the gender gap, we have, however, checked the outdoor world record for 10 comparable running events at the end of the 1980's and following the 1996 season. Contrary to Whipp and Ward's forecast, by this methodology the gender gap has not shrunk, but has actually grown. Seven of the 10 men's world records that were on the books at the beginning of this decade have been shattered, versus only three of the 10 women's world records. The average gender gap among the 10 world records has grown from 10.2% to 10.9% today.
In fact, two of the three new women's records were set by Chinese runners at an extremely controversial meet in Beijing in September, 1993, when a host of astonishing, never replicated times were recorded by various Chinese women running for the mysterious Coach Ma. If we exclude the Chinese records in the 1500 and the 10,000 meter runs, as some track experts suggest, only one women's mark has been bettered since the last decade: the 5000 meters, and that by only 0.1%. Excluding the two dubious Chinese women's records, the gender gap for world records has grown from 10.2% to 11.2%.
End of End of 1996 Within
the Sex
Eightie Change
s
Male Fem WR Gender Male Fem WR Gender Male Female
WR Gap WR
Gap
100 9.92 10.49 5.7% 9.84 10.49 6.6% -0.8% --
4x100 37.83 41.37 9.4% 37.40 41.37 10.6% -1.1% --
200 19.72 21.34 8.2% 19.32 21.34 10.5% -2.0% --
400 43.29 47.60 10.0% 43.29 47.60 10.0% -- --
4x400 02:56.2 03:15.2 10.8% 02:54.3 03:15.2 12.0% -1.1% --
800 01:41.7 01:53.3 11.4% 01:41.7 01:53.3 11.4% -- --
1500 03:29.5 03:52.5 11.0% 03:27.4 03:50.5 11.1% -1.0% -0.9% Chinese
5000 12:58.4 14:37.3 12.7% 12:44.4 14:36.5 14.7% -1.8% -0.1%
10000 27:08.2 30:13.7 11.4% 26:38.1 29:31.8 10.9% -1.9% -2.3% Chinese
Marathon 2:06:50 2:21:06 11.2% 2:06:50 2:21:06 11.2% -- --
Bold
= New
Record
Average 10.2% ------ 10.9% -1.0% -0.3%
------
---->
Avg 10.2% 11.2% -1.0% -0.01%
w/out 2
Chinese
1993
marks
Boxed
records
were
set by
athletes
from
European
or
Asian
communis
t
countrie
s with
a
history
of
systemat
ically
doping
athletes
.
By the way, men's records in these 10 events have been broken or equaled on 23 separate occasions during the 1990's. In contrast, a new record has been set only once by women in the 1990's, aside from 5 record-smashings that occurred at that one Beijing track meet.
Still, interesting as this is, we'd caution against leaping to the conclusion from this that the conventional wisdom about the gender gap is wrong: any analysis of world records is freighted with too much ambiguity; conditions simply vary to much from one setting to another to put much confidence in the results. Instead, we prefer to compare results obtained by men and women in the same track meet, running on the same track.
[Update in 12/97 -- for the National Review article of 12/31/97 we devised a new methodology regarding world records to get around the paucity of datapoints when you just look at world records at a particular point in time: we counted all the "record-settings" (defined as record-breaking efforts plus record-tying efforts). This gives us substantially more data points (e.g., almost 89 for the Seventies alone). The results, however, lead to the exact same conclusion as all the other methodologies: the gender gap is widening in the Nineties.]
We've assembled data from the Olympics and the five International Amateur Athletic Federation World Championships. The prestige of the Olympics as the world's most important track meet is hardly challenged. The World Championships, which began in Helsinki in 1983, are unquestionably the second leading track meet. In fact, since the World Championships have never been damaged by extensive boycotting, the strength of competition in the 1983 and 1987 meets were certainly stronger than in the US-boycotted 1980 Olympics, or the East European boycotted-1984 Olympics, and may have been stronger than the African-boycotted 1976 Games. In other words, the first two World Championships may well have been the most representative competitions between the 1972 and 1988 Olympics. We've included all five World Championships (1983, 1987, 1991, 1993, and 1995).
5. When?
Women's track was first offered at the 1928 Olympics. Deciding when to begin tracking the gender gap in the Olympics is fairly arbitrary. To avoid the large gap caused by the cancellation of the 1940 and 1944 Games, we decided to restrict our analysis to the post-war era. Further, the 1948 London Games were conducted in the shadow of the war, with depleted entries, austere conditions, and terrible weather. So, we began with the highly successful 1952 Helsinki Games (which featured the first appearance of the Soviet Union). In summary, we have 17 championships:
1950's: 2 Olympics
1960's: 3 Olympics
1970: 2 Olympics
1980's: 3 Olympics, 2 World Championships
1990's: 2 Olympics, 3 World Championships
How many results from each race?
We've amassed the times for Top 6 finishers in each event.
It's easiest just to compare the gold medalists, and easiest to remember the winners' times. However, this diminishes the number of data points, reducing confidence. So, we recommend using either the Top 3 (i.e., medalists) or the Top 6 finishers.
Why the Top 6 rather than the Top 8? First, before 1964, the Olympics only allowed 6 finalists in the shorter races. Second, it's troublesome to take all the starters, since they don't all always finish (especially in the 4 x 100 meter relay where dropped batons and disqualifications for illegal passes are common) or always give a full effort all the way to the finish line. For example, of the 8 starters in the famous 1988 Seoul Olympic's men's 100 meters, Ben Johnson's stunning time of 9.79 seconds was later disqualified for drug use, and Raymond Steward limped home in a 12.26 seconds, leaving only six reasonably representative times. Fortunately, in the small number of races from 1952-1960, when only six were allowed to start each race, there are no extravagantly bad times 6th place times. Since 1964, only one race stands out where the 6th place finisher's time is arguably unrepresentative: in the 1993 World Championships' women's 800 meters, three runners collided and fell. Liu Li scrambled back to her feet to finish in 6th place in 2:04.45, a full 4.88 seconds behind the fifth place finisher. So, in general, Top 6 appears to be the broadest net that can reasonably be cast.
The good news is that it doesn't seem to particularly matter whether you look at just the gold medalists, all three medalists, or the top 6 finishers. The same general historical gender gap trends emerge.
One sticky issue is what to do about the times achieved by individuals caught using performance-enhancing drugs. We've decided to follow the quite lenient IAAF lead. We do not include the very small number of times (most notably Ben Johnson's 9.79 in the 1988 Olympic 100m) disqualified by post race urinalysis. However, the IAAF seldom retroactively disqualifies old performances when a runner is later caught cheating. We do not use the only performance the IAAF has retroactively banned: Benoid's 9.83 at the 1987 World Championship, but that's the only one. We even include Johnson's 10.22 bronze medal performance at the 1984 Olympics, even though he later gave sworn testimony that he had been using steroids since 1981, because the IAAF doesn't seem to have a problem with it.
Obviously, there are some other spectacular times recorded by individuals never caught by doping tests but that serious track fans tend to be dubious about. We've left all those in, but we will return to the topic of doping.
Which events?
A. We are only looking at the gender gap in running, so we excluded Field events.
B. We want only comparable races where men and women compete under exactly the same conditions, so that we can calculate the gender gap: Thus, we excluded all hurdle races:
1. Women do not run the Steeplechase.
2. The High Hurdles are different distances for men and women (110 meters vs. 100 meters) and the hurdles are different heights.
3. The women's hurdles in the 400 meters are 6 inches lower than the men's hurdles.
C. There are now 10 races where men and women compete under identical rules. The following table indicates at which Olympics or World Championship were women first allowed to compete at a particular event:
Event First Year for Women # of Times (since '52)
100m 1928 17
4x100m 1928 17
200m 1948 17
400m 1964 14
4x400m 1972 12
800m 1960 (but first tried in 15
1928)
1500m 1972 12
5000m 1995 WC (replacing 3000m, 2
run from 1983-1993)
10000m 1987 WC 7
Marathon 1983 WC 9
Thus, beginning in 1952, we have observations from 122 events in which both men and women competed. Counting the top 6 finishers for both men and women in each event, we thus have 1,464 individual data points. However, we are missing results for 12 data points: for the 4th through 6th place finishers in both sexes 1987 relays. Also, In the 1960 Olympics, in both the men's and women's 4x100, there were only five official finishers. So, in total we have 1,450 data points.
A major methodological question in measuring trends in the size of the gender gap over time is which events to aggregate over time. For example, it's tempting to average all the gender gaps from 1952 through 1996.
The problem with this approach is that the size of the gender gap appears to be correlated with the length of the race in some rough fashion. The 100m consistently has the smallest gender gap, while the 5000m and 10,000m frequently have notably larger gender gaps. For example, the 100m typically has by far the smallest gender gap, with the 200m and the 4x100m relay being also below average.
100 8.7% 4x100 11.1% 200 10.2% 400 11.9% 4x400 12.6% 800 12.3% 1500 11.3% 5000 13.5% 10000 13.1% marathon 13.1%
Now, the problem with aggregating the gender gaps of all the events is that over time, the average length of comparable events has gotten longer. For example, in 1952 and 1956 the only events for which we can calculate gender gaps were the 100m, 4x100m, and 200m. Since these will tend to have lower gender gaps than the longer events introduced for women later, merely aggregating all the comparable events will tend to artificially increase the gender gap over time.
One workaround is to choose a subset of events that both sexes have competed in over a subset of the time. One obvious group are the seven races (including the two relays) from 100m through the 1500m. We can calculate gender gaps for each of these events for the last 12 championships. That's 84 comparable events composed of the top 6 finishers of each sex, which equals slightly over 1000 individual performances. (Other subsets are possible. Most show generally the same trends.) Here is what we've seen:
The gender gap began dropping sharply after the 1968 Olympics (not shown on this chart because women competed in only 5 flat races then) and bottomed out in 1980 when the US boycotted the Moscow Games. In 1984, the Communist boycott caused a temporary increase in the gender gap. The mid-Nineties (1993, 1995, and 1996) are consistently higher than the non-boycotted years in the Eighties (1983, 1987, 1988).
The shrinking of the gender gap in the 1970's was largely the product of women getting consistently faster while men's times stagnated. Beginning in the early Eighties, men's times have improved slowly and slightly, but with some degree of consistency, reaching an all-time best in 1996. In contrast, women's times have been slower in the mid-Nineties than in the previous decade and a half. Women's times were better in 1987, 1988, and 1992 than in 1996. So the growth of the gender gap in mid-Nineties is caused largely by women getting slower.
For anyone who is not a dedicated, quantitatively-oriented track fan, these results are probably quite unexpected. We've come to assume that runners will consistently get faster, yet women have gotten slower. And we've been assured so many times by the press that gap between men and women in all fields of achievement is narrowing that it comes as something of a shock to see it widen in even one.
The key question to unravel is: Why, despite unprecedented encouragement for female sports in industrialized societies at present, is the gender gap in the mid-90's greater than in the 1980's?
To summarize the gender gap trends, we'll look at the seven events (100m to 1500m, including relays) that have been contested by men and women since 1972. And, we'll leave out the boycotted 1980 and 1984 Olympics, and the transitional years 1991 and 1992:
1972, 1976 11.7% 1983, 1987, 1988 10.5% 1993, 1995, 1996 11.6%
Two other intriguing facts might be related to our primary question.:
1. Eastern and Central European women, especially Germans, used to dominate world championships, but not their menfolk.
To illustrate this point, let's look again at just at the stable seven events from 100m to 1500m that both men and women have competed in since 1972. We'll assign 6 points for a first place finish and 1 for a 6th place finish, and divide by total points at stake to obtain a measure of dominance.
2 The relative performances of Eastern and Central European, especially Germans, have declined dramatically for both sexes since the Eighties.
Factors affecting performance results --
To explore what's happened to the gender gap over time, it's important to understand a wide variety of factors (e.g., wind, altitude, track surface, better nutrition, illegal drugs, or changing social attitudes) that can influence the times achieved in a particular race. We'll consider each influence on performance, as we search for explanations for these unexpected trends in the gender gap and in women's performances.
Factors can
or both.
Whether the factors change the gender gap or not, their duration and predictability is also important to comprehend.
With three of these factors (wind, altitude, and hand-timing) we've found reasonable ways to quantitatively adjust the official results to eliminate much of the bias. Future researchers might find reliable methods for adjusting away the impact of some other factors like heat and humidity.
Factor's
Impact on
Results
Impacted Impacted
one sex both
more? sexes?
Factor Impacted Duration & Impacted Duration & Adjusted Which
one sex predictabili both predictabil for in Events
more? ty of sexes? ity of Study? Impacted
impact? impact? ?
Hand-Timing Trivially Long term Yes; Long term Yes Primaril
hurt (52-64) and understate (52-64) y
women's random d times and random Sprints
times noise noise
Altitude Unknown 68 & Yes 68 & Yes Favored
(slightly) (slightly) sprints
72 72
Wind Unknown Fairly Yes Fairly Yes for 100m,
random random 64-96, 200m
some of
52-60
Other Unknown Unknown Yes Heat No Heat
weather impacts hurts
differences single distance
years, s
rain more
random
Track Unknown Unknown Yes 91, 96 No Hard
Hardness hard, but track
no data on favors
most years sprints,
hurts
distance
s
Race Unknown Unknown Yes Fairly No Distance
Tactics random s,
mostly
800 and
1500
Changes in Doubtful Not much Slight No E.g.,
Scheduling long term 400m
positive
trend
New Events Dilutes 1960-1992 No None No Longer
female races
talent,
but only
equalizes
with male
Improved Doubtful Unknown Yes Long term No All
Techniques trend events
Participati Yes - Long term Yes Long term No Distance
on - societies trends - trend s
Population can't push African
Growth both high dominance
fertility in 1990's
and elite especially
female
sports
Participati Unknown Unknown Yes Highly No All
on-Injuries random, events
but
unknown if
any long
term trend
Participati Yes - Single Yes Single No Various
on-Boycotts different Years: (76 Years events
nations & 80 hurt in
have men), (84 various
different hurt women) years
strengths
Participati Yes Long term Yes General No See
on-Social trends long term discussi
pressures growth in on
opportunity
Decline of Yes-various Complex Yes - Long term No Most
Amateurism impacts long term pros trend -- events
trends older no
than amateurism
amateurs in East,
died in
West in
80's
Participati Yes Long term Yes Long term No See
on-Money trends trends discussi
on
Doping vs. Yes - Doping Yes Yes - No All
Testing Steroids advances we'll events,
more helped discuss in but
effective female detail shorter
on women times in events
70's, more
testing
hurt them
in 90's
Hand-timing vs. electronic timing -- Until 1968, the Olympics used manual stopwatches to time races to the tenth of a second. It was widely recognized that it took the timekeepers generally about 0.1 to 0.2 seconds to react to starter's gun. Therefore, reported times generally understated than actual times. This factor impacted both sexes up through 1964. Furthermore, as well as this long term trend, there was a substantial amount of unpredictable random error. For example, the official men's 100 meter hand-timed results in the 1952 Olympics appear to be about 0.4 seconds (3.6%) less than the actual results, as judged from the movie of the event. Finally, there might have been a tiny bias toward increasing the gender gap. For example, if a man ran the 100 in an actual time of 10.15 seconds, while a woman ran it in 11.15 seconds, and the hand times were 0.15 seconds less for each (10.0 vs 11.0), then the gender gap would 9.85% with electronic timing versus 10.0% with hand-timing. Of course, longer races would be affected even less.
Fortunately, from the original documentary movies, Bob Sparks has calculated the actual times to the 1/100th of a second from movies for almost all events from 1952 onward. We've used these times wherever possible.
Altitude -- This impacts both sexes. It is predictably limited to championships conducted at high altitude. The Mexico City Olympics of 1968 were contested at 7347 feet and the 1972 Munich Games at 1700 feet. No other championships have been held at over 1000 feet. The impact of the rarefied air was extremely evident in Mexico City, with world records being broken by the men in all five of the anaerobic flat races and relays from 100m through 400m, with the 400m and 4x400m records lasting into late 1980's. On the other hand, the aerobic distance races were, as expected, far off record pace. In contrast, altitude's impact on the times of the Munich Games was much more subtle and the time adjustments are minor.
Oddly enough, while the IAAF will not accept as a world record any 100m or 200m times aided by a tailwind over 2.0 meters per second, it blithely ignores the impact of altitude. We, however, do not ignore altitude, and so we've adjusted the 1968 and 1972 times according to the tables shown for "Wind" below. The adjustments appear quite reasonable -- after altitude adjustment, the 1968 results no longer appear nearly superhuman. Performances at Mexico City were impressive but, without the aid of altitude, they fit in nicely with the long term trend of the era toward better times.
All other championships have been held at altitudes of no more than 1000 feet. We have not attempted to adjust those because the impact would be trivial (e.g., about 0.01 seconds or 0.1% in the 100m at 1000 foot altitudes).
High altitude certainly hurts long distance runners, but since in 1968 women weren't allowed to compete at over 800m, this effect will not impact our gender comparisons.
Wind -- Wind has undeniable impact on 100m and 200m times. We've adjusted it according to the following tables provided posted by Patrick Hoffman on his website
The track and field "Big Red Book" has some data compiled by P.N. Heidenstrom of New Zealand from data from Nobel Laureate A.V. Hill in the 1930's to calculate the effect from wind of a runner of average size who can run a 100m in 10.10 in still air. It sometimes can help (or hinder) in a 200 but it is not used. In longer races theoretically it might balance out, but the nuisance factor usually makes it harder...For the 100 meter a rule of thumb is 0.1mps of wind has an effect of 0.01 on the overall time. (Heidenstrom believes head wind has more of an effect
But the same table is used.
wind effect on time meters per
second .1 .01
.2 .02
.03
.03
.04
.05
.06
.07
.08
1.0 .08
.09
.10
.11
.12
.12
.13
.14
.15
.15
2.0 .16
.17
.17
.18
.19
.20
.20
.21
.22
.22
3.0 .23
.23
.24
.25
.25
.26
.26
.27
.28
.28
4.0 .29
.29
.30
.30
.31
.31
.32
.32
.33
.33
5.0 .34
5.5 .36
6.0 .38
6.5 .40
7.0 .41
7.5 .43
8.0 .44
8.5 .45
9.0 .45
9.5 .46
10.0 .46 (22 miles per hour wind)
to convert mps to miles per hour multiply by 2.23693
Altitude factors..... Higher altitude ... less air resistance .. good news
less oxygen bad news
800 meters is break even good/vs bad
over 800 slower times at higher altitude estimates 2200m (7200ft) adds .10 sec in 100, .20 secs in 200 , .35 in 400
200m(700ft) .01 sec (100m)
520m(1700) .03 " (Munich)
(4500) .06 "(Provo Utah)
For distance runners it depends how much they have trained at Altitude.
Horizontal jumpers may gain 2% at 2200m (6 inches for 26 feet)
Wind and Altitude.. not as much as both effects added; e.g., wind adds .10 at 2200m (.10) the benefit is only .16 not .20 seconds
Hoffman does not attempt to provide adjustments for the 200m. The complicating problem is that the first half of the race is run around the curve, and so the impact of the wind depends not just on speed but direction. (In fact, wind impact will differ depending on which lane the runner is in, since, due to the staggered start, a runner in the inside lane will be running away from the finish line for a longer period than a runner in the outside lane.) However, the impact of the wind during the second half of the race on the straight-away is very similar to its impact during the 100m, so we've simply used the same adjustments as for the 100m. This probably understates the impact of the wind on the entirety of the race, but at least it does no harm.
Wind readings came from multiple sources. We have accurate wind readings for all men's and women's 100m and 200m races from 1964 on. Unfortunately, "[i]n the early times, wind readings were not always recorded, hence the problems with 1952-1960 for a few events." Before then, we sometimes have the wind reading for one sex but not the other for a particular event. We decided not to wind adjust any event where we didn't have the wind reading for both sexes' race, on the general data analysis principle of first, do no harm. Out of the 68 total sprints, the following are the eight unadjusted sprints (both sexes):
1952 100m 200m
1956 200m
1960 200m
We would particularly like to have the wind readings to be able to adjust the very slow men's 1952 100m (won by the obscure Lindy Remigino in 10.79 seconds) and the very slow women's 1960 200m (won easily by the great Wilma Rudolph in 24.13 seconds, a full 0.8 seconds slower than her opening heat).
Oddly enough, the wind has favored male sprinters more than female sprinters in each of the five championships of this decade. So, by adjusting for the wind, we've avoided overstating the gender gap in recent years.
Other weather differences -- These include temperature, humidity, rain, air pollution, etc. They affect both sexes. Some of these weather factors are predictable (e.g., the cool weather expected in Sydney in 2000 should benefit distance runners at the expense of sprinters), while others, like summer thundershowers, are proverbially erratic. So, this catch-all category is both Predictable to some extent and highly Random.
Does, say, temperature impact the sexes differently? We don't know. Fortunately, our general methodology works to minimize the gender-biasing impact of weather fluctuations. First, by relying only on results from championship meets, our methodology dramatically narrows the range of possible weather conditions. For example, all races in Atlanta in 1996 were held under temperatures that ranged between Warm and Hot, and humidities that varied between Sticky and Steambath. Further, at least in recent Olympics, in the seven comparable events from 100m to 1500m, the women generally run about 20 minutes before the men. Thus, with the occasional exception of a sudden thundershowers, the weather conditions will tend to be very similar for both sexes. In the three distance events, the need to hold the two marathons on separate Sunday mornings leads to the 5000m and 10000m being flip-flopped in the schedule. So, these three events could be contested under somewhat different weather conditions. Still, the general climate remains the same.
It's certainly true that some random weather factors can cause biases in the gender gaps for an individual race. For example, as well as running into a headwind, the women's 200m in 1960 was conducted in a downpour. This contributed to an enormous gender gap (calculated by comparing the average time of the top 6 finishers in each race) of 18.6%, the largest seen in the 122 events included in the database. However, despite the notoriously predictable unpredictability of the weather, in the long run, we would expect the Law of Large Numbers to even out the weather conditions under which men and women run. And we certainly have a fairly Large Number of observations in our database.
Interestingly, it's even possible that equivalent weather conditions would impact the sexes differently. (For example, since women, with their smaller average weight, tend to dissipate heat faster than men, the ideal temperature for women marathoners might be a few degrees different than for men.) We encourage future researchers to study these complicated issues. We have not, however, adjusted any of the results for any weather factors other than wind.
Track Surface --- A hard track speeds up sprinters but beats up distance runners. This affects both sexes. Track hardness is quantifiable, so its impact will be fairly predictable, with little degree of randomness. A major championship will therefore typically be biased toward either the sprinters or the distance runners. We're not aware of any evidence that it has differential impact on the sexes, although that's possible to some degree.
The 1991 Tokyo World Championships featured an extremely hard track, which contributed to the remarkable feat of six men breaking 10.00 seconds in the celebrated men's 100m final and the famous long jump duel where Carl Lewis averaged 29' but lost when Mike Powell broke Bob Beamon's seemingly untouchable record. After that meet, standards were adopted limiting the hardness of the track. The 1996 Atlanta track was designed to be just under the limit for maximum hardness.
Since track hardness must now be measured, future analysts will have the raw materials for devising quantitative adjustments for track hardness, but we didn't attempt to adjust here.
One thing to keep in mind is that while average times at these big meets have only improved marginally since 1972, world records continue to improve. This might be because we've reached the point where a particular championship can only be optimized for one type of record. Whereas the 1932 Los Angeles Olympics featured relatively excellent times at a variety of distances due, in part, to conditions that were far above average for the era, modern Olympic organizers must choose which events to highlight. Once the 1996 Olympics were slotted for Atlanta in the middle of summer, it was clear that world records in distance races were unlikely due to the climate. Furthermore, Atlanta's organizers knew that Americans were far more of a record threat in the sprints than the distance events. So, Atlanta's organizers logically picked just about the hardest legally allowable track, and they were duly rewarded with two sprint World Records, most notably American Michael Johnson's epochal 19.32 in the 200 meters. In contrast, Ethiopia's Haile Gebresellaise, the world record holder at the time in both the 5000m and the 10000m dropped out of the 5000m, claiming his legs were too beaten up in winning the 10000m to attempt the double.
Race Tactics -- At up through 400 meters, racers typically run as fast as they can, paying little attention to the competition. At longer races, however, runners can choose among various tactics. Some runners prefer a race that's fast all the way through, while others with a strong finishing kick want a slow initial pace. Which strategy dominates in a particular race is a complex problem in group dynamics far beyond the scope of this paper. It's widely believed that Olympic finals tend toward slow paces culminating in a final mass sprint, especially in the showcase 1500 meter event. Possibly runners don't want to risk embarrassing themselves by setting too fast an early pace and then, with the gold medal in their grasps and the whole world watching, dying agonizingly on the homestretch of the bell lap. Anyway, whatever the reason, times in the 1500m have not progressed as steadily as in some of the other races. For example, Herb Elliott won the 1960 Rome 1500m in 3:35.6, but 32 years later in Barcelona, Fermin Cacho took the gold with a sluggish 3:40.12.
This factor appears to be concentrated more in the 800 and 1500 meters than in the longer races. Within those medium-distance races, it appears to be fairly random in occurrence. Thus, we've made no effort to adjust for it, since we wouldn't have a high degree of faith in any scheme of adjustment.
Even if random, however, race tactics can impact the gender gap in a single year. For example, in 1992 the slow men's 1500m, combined with the quick women's 1500m, narrowed slightly the overall gender gap for the entire Olympics. Still, over enough time and events, the impact of race tactics should even out.
There is however, a logical possibility that beginning with the arrival of the Kenyans on the world stage in 1968, men might be systematically (if only slightly) more inclined than women to races with a fast early pace. Kenyan men have been far more prevalent in Olympic finals than Kenyan women (we have 57 Kenyan male top 6 finishers in our database versus only 5 Kenyan females), and Kenyans tend to prefer a fast pace from the opening gun. In contrast, Europeans and North American whites have often been stronger kickers. As Kip Keino, whose upset of the great kicker Jim Ryun in the 1968 1500m announced to the world the arrival of Kenya as a track superpower, explained:
Our style of running was not like theirs. When I ran, I liked to 'front-run.' European athletes, on the other hand, used to space themselves with the aim of saving themselves for a final lap sprint finish.
We encourage future researchers to look further into the complex question of race tactics.
Scheduling -- More intelligent scheduling of preliminary races at the Olympics probably has contributed slightly to the improvement in times over the years. For example, from 1920 through 1956, the 400m finals were run the same day as the semifinals. Beginning in 1960, the finals have been run the following day. The large improvement in the male gold medalist's time from 1956 to 1960 (from 46.85 to 45.07) can probably be attributed in part to the extra rest the new schedule allows. (Women didn't run the 400m until 1964, so we aren't including these two men's 400's in our database, anyway).
For runners attempting more than one event, the details of scheduling can be all important. Recall Michael Johnson's enormously publicized and ultimately successful campaign to have the order of events at the 1996 Olympics revamped to aid his quest for gold medals in both the 200m and 400m. On the other hand, there is little evidence that scheduling could predictably afford one sex an advantage over the other. (For example, the changes made to accommodate Michael Johnson also allowed Marie-Jose Perec to win the women's 200m and 400m). In the Olympics, in the seven races up through 1500m, the women typically precede the men by about 20 minutes. This tends to equalize tiredness, temperature, crowd size, and other factors. In the three longer races, women and men run on different days, which allows more random differences to creep in (e.g., different temperatures), but it seems unlikely that this biases the results in a systematic fashion toward one sex or the other.
Keep in mind that Olympics and World Championships have massive numbers of entrants requiring heats in every race except the marathons. This works to the advantage of some runners (e.g., Michael Johnson says that running a 200m semifinal in 20.27 seconds is the best warm-up imaginable), while other runners are at their best on the Grand Prix circuit, where they only have to run one race a week (e.g., over Merlene Ottey's amazingly long career she has tended to win far more often in one-race invitationals that when she has to run heats). However, we're not aware of evidence that there are gender differences in sensitivity to the number of heats, or that there have been trends over time. Similarly, stage fright undoubtedly has some impact on our results, but there's little evidence that one sex differs from the other in this regard.
One other minor point can be dealt with here: these Championships limit the number of entrants from one country to 3. This can have a detrimental effect on the times recorded, most notably in the steeplechase (not in our database) where Kenya routinely has 9 of the top 10 steeplechasers in the world but can only bring 3 to the Olympics. Overall, this rule negatively impacts performances, but probably only slightly. Whether it has been growing or declining in significance, or whether it has differential impact on the sexes is uncertain. The redrawing of borders with the collapse of communism has had some impact: East Germany and West Germany went from 6 legal entrants to 3, but the breakup of the Soviet Union had the opposite effect. So, the overall effect was probably relatively sex neutral.
New events for women -- Although the number of running events for men has been stable in the Olympics since the 1920's, there has been a steady growth in the number of flat races for women. Conceivably, this could have diluted female talent in any one race. For example, the women's 3000 meters was added to the Olympic roster in 1984. The top American runner, Mary Decker Slaney, decided not to run in the 1500 meters so she could concentrate on winning the new event. (As it turned out, she collided with Zola Budd and was knocked out of the 3000 meter final).
While this would tend to widen the gender gap, the addition of events for women only levels the playing field. Men have had to choose between 10 flat races (as well as the 3000 meter steeplechase) for decades. For example, in Atlanta in 1996, after winning the 10000 meters, Ethiopia's Haile Gebresellaise announced that his legs were too beaten up by Atlanta's hard track to attempt the 5000 meters. This opened the door for Burundi's Venuste Niyongabo, who had been considered the top challenger to Algeria's Noureddienne Morceli to switch from the 1500 to the 5000, which he won. And Morcelli won the 1500 in only a moderate time due to a lack of challengers (especially after Hicham El-Gerrouj fell). This plethora of events clearly diluted competition among men, and now women face the same roster.
Techniques -- Improvements in running techniques, training, nutrition, shoes, track surfaces, etc. have indisputably improved times over the decades for both sexes.
It's unknown whether new techniques and technology have impacted the gender gap, but it's likely. In general, the more technologically advanced the country, the less sports was restricted to men-only. Among track superpowers, East Germany, an enthusiastic supporter of women's athletics (to put it mildly), may well have been the leader for most of the 1970's and 1980's in advancing technical expertise. "While U.S. and Soviet scientists were busy conquering space and designing weapons systems, the East Germans were studying athletes -- taking blood tests and muscle biopsies, checking oxygen levels, and testing for nutritional needs." At the other end of the spectrum, Kenya and Morocco, the producer of enormous numbers of great male runners but few great female ones yet, have conducted little cutting edge research. (The new powerhouse, Burundi, has been too preoccupied with genocidal strife to conduct any.) So, technology-based innovation is likely to narrow the gender gap because it's likely to occur in more sexually equalitarian nations. On the other hand, this will merely narrow the gender gap closer to whatever it's natural level will be ... unless the innovations aid one sex more than the other. In fact, of course, one of the innovations that has turned out to be extremely important, anabolic-androgenic steroids, or artificial male hormones, has a greater positive impact on women's performances.
Participation -- The most complex factor we call "participation." In other words, who shows up at the starting line?
Participation: Population growth -- All else being equal, the more people on Earth, the more really fast runners there will be. Since the population increases yearly, this factor Predictably Biases results toward getting better over time.
Interestingly, population growth tends to widen the gender gap. This is because a society can effectively encourage women to have many babies or to compete in sports at the world class level, but not both. Nations that choose the former will grow much faster in population than those choosing the latter.
One reason (among many) for the growth in Kenyan dominance of distance running and the decline of traditional European powers is that the ratio of Kenyans to Europeans has changed radically over recent decades. Until very recently, Kenya had one of the highest birth rates in the world: in the early Eighties Kenya's birthrate was 50.5 per 1000 people, just trailing Rwanda, #1 at 51.0. In contrast, women's track superpower, East Germany's birthrate was only 14.5, and West Germany (another strong force in women's track) had the lowest birthrate in the world among sizable nations at 9.5. Societies like Kenya or Morocco (early Eighties birthrate of 46.8) that encourage men to run and women to stay home and have lots of babies have come to dominate men's distance running. In contrast, countries like many in Europe that encourage women to run are shrinking in their share of the world population.
Participation: Injuries/Illness -- Injuries and illness can have a striking impact on who wins a particular race, but by observing the Top 6 finishers we've mitigated this factor somewhat. Injuries and illness are highly random and unpredictable. We are not aware of evidence that one sex suffers more poor health than the other. Nor do we know whether there are long term trends affecting the influence of health. We have made no attempt to adjust for health.
Participation: Boycotts -- Boycotts have had huge impacts on some Olympic Games. The most important were the US-lead boycott of the 1980 Moscow games (which kept many top male runners home) and the tit-for-tat Communist boycott of the 1984 Los Angeles games (which sidelined many top female runners). The third most important was the 1976 African boycott, which probably narrowed the gender gap since there were few female African runners at that time. There have also been minor boycotts in 1956 and 1988. Besides single year boycotts, Communist China kept itself out of the Olympics until 1984, while the International Olympic Committee banned South Africa until 1992.
Boycotts undoubtedly reduce competition for both sexes, but the most striking impact has been on the gender gap. The American boycott of 1980 reduced the gender gap that year to the narrowest ever recorded, while the communist boycott of 1984 drove the gender gap up to heights not seen again until the next decade.
Participation: Decline of amateurism - Baron de Coubertin's ideological animus against professional athletes was often evaded, but the first appearance of the Soviet Union at the Olympics in 1952 signified the first massive challenge to the amateur policy. Soviet athletes were obviously full time professionals, but the IOC responded by largely ignoring that fact.
Amateurism declined slowly in the West, but by 1984 it was obviously moribund. Professionalization undoubtedly has contributed to improved times for a variety of reasons. For example, stars like Carl Lewis and Merlene Ottey now stick around longer, continuing to compete at a high level into their later 30s.
The impact of all this on the gender gap is difficult to sort out. Professionalization in the U.S. may have widened the gender gap slightly because US male sprinters are now less likely to abandon track for professional football than in the past. For example, in 1968 Wyomia Tyus successfully defended her women's 100m crown, but her male counterpart from 1964, 100m champ Bob Hayes, was unable to attempt to defend his title because he was starring in the National Football League by then. Other world record setters who gave up track for the NFL include Jim Hines, Renaldo Nehemiah, Ron Brown, and O.J. Simpson (whatever happened to him?).
In contrast, female runners like Wyomia Tyus have not been tempted by as many lucrative alternative sports. The small number of high-paying sports for women (e.g., tennis, golf, or figure-skating) tend to require radically different talents than track does, so they don't compete with running for athletic talent the way football and track compete for very fast males. Thus, women runners have fewer other options.
However, one growing professional sport for women could someday sap talent from women's running: basketball. Jackie-Joyner Kersee, for example, is playing in a new U.S. based women's league.
Participation: Prize Money -- The Olympics have never offered cash prizes to winners, although many national Olympic committees independently offer reward to medalists. In contrast, on the Grand Prix circuit or at, say, the Boston Marathon, runners expect to be directly rewarded. Thus, it's possible that competitors might be offering less than their best efforts at the Olympics these days. There is little evidence for this, however: the prestige of winning an Olympic medal (and the multitudinous ways to cash in on it) continue to elicit outstanding performances.
[The World Championships began offering a certain amount of prize money in 1997, although it appeared that a number of top athletes were more focused on the invitationals immediately following the WC, where both conditions and, possibly, rewards were better. Times at the 1997 WC were generally unimpressive, but the weeks following the Athens meet saw one of the great record-breaking binges in male track history.]
The marathon might be a different story. The marathon is much more grueling than any other running event, and runners realize they have only a limited number of chances in their lifetime to give their maximum effort. The long string of unspectactular marathon times and poor performances by stars might stem from lack of monetary motivation. On the other hand, marathon times have stagnated since the 1980's in all events, paid or unpaid, so it's unclear exactly what's going on with the event.
In general, you can't get Michael Jordan-rich running these days. For example, the top American runner Michael Johnson will not make the 1996 Forbes Top 40 list of highest compensated athletes. While prizes like the $100,000 first prize in the Boston Marathon are less than overwhelmingly exciting to jaded Westerners, that still represents a huge amount of money in Africa. This is another reason for the dominance of African runners. And since African societies tend to discriminate against women, the moderate size of the purses tends to widen the gender gap.
Participation: Social Prejudice -- Social prejudice against women running is widespread in numerous countries. The pressures faced by Kenyan women have been well-documented. This causes an imbalance in teams from East Africa (with the recent exception of Ethiopia where the government trains women runners) and Northwest Africa. Undoubtedly, these kind of social factors exist to some degree everywhere, but their power in polygamous countries is almost certainly greatest.
In one other country, there appeared a huge imbalance in favor of men: surprisingly enough, it was Brazil. Brazilians appear 30 times in the database since 1952 and they are all men. This is rather odd since most of the countries with which Brazil shares one or more similarities, whether of language, ethnic origin, climate, etc, nations like Portugal, Nigeria, Cuba, and Jamaica, show much balance between the sexes. We can safely predict that the sporting press will not report on the reasons for the poor performance of Brazilian women in track until a Brazilian woman champion emerges. At that point, we can expect to see numerous articles detailing whatever it is that has held back Brazilian women so far. Similarly, Hassiba Boulmerka's triumphs at 1500 meters in 1991 and 1992 generated much coverage of how Muslim fundamentalists in Algeria militate against women athletes. Without Boulmerka's victories, however, it's unlikely we would have heard much about it.
On the other hand, some countries have had far more women reaching the top 6 spots than men. While it would be nice to think that this was because of a lack of prejudice against women in these countries, it's worth keeping in mind that most have been communist police states. With the downfall of the East German dictatorship, for example, came the downfall of women's running in that land.
There are undoubtedly many other factors that influence the gender gap and long term performance trends within each sexes. We encourage further research. Still it does not appear to us that any or all of the aforementioned fully explain the general trend seen in the gender gap over the last 25 years. There is, however, one other factor:
Doping and Testing -- The simplest, most power explanation for the growth of the gender gap is that in the Nineties runners can no longer use artificial male hormones with quite the impunity they could in the past. This also explains why women are running slower; why the East German women used to dominate, but not the East German men; and why Germany is no longer a track power.
It's impossible to discuss the use of illegal drugs in running without almost instantly treading on extremely controversial grounds. For example, Carl Lewis was widely criticized in the press during the run-up to the 1988 Olympics for implying that rival Ben Johnson's enormous increase in musculature, personality change from shy to surly, and yellow-tinged eyes suggested that he was taking steroids. After all, it was widely written, Ben had passed every drug test he'd ever taken, and Lewis must be a cad to dare impugn a great champion's honor. Then, after Ben flunked his urine test following his astounding, not-yet-equaled 9.79 in the 1988 Olympic 100m, some of Ben's die-hard supporters reversed course 180 degrees and argued that absolutely everybody takes steroids and the only thing Ben did different was get caught.
Getting the straight story on doping in sports is extremely difficult. Nobody knows everything. Those insiders who do know a lot aren't telling, unless they get caught, at which point they have every incentive to claim everybody does it.
Fortunately, for our purposes, we don't have to know where the percentage of drug cheaters at the world class level has been 5% or 20% or 90%. Three much less controversial propositions can suffice for helping us understand trends in the gender gap:
1. The communist regimes of Eastern Europe engaged in extensive and effective doping during the Seventies and Eighties. "Indeed, getting around the rules has become an official government mission in some countries. Under East Germany's notorious State Plan 14.25, more than 1,000 scientists, trainers and physicians spent much of the 1980's developing better ways to drug the nation's athletes." The Soviet Union likewise engaged in cheating on an impressively industrial-scale cheating. "Zmena, an official publication of the Communist Party of the Soviet Union, did admit in March 1989 that the Soviets indeed had a $2.5 million testing facility floating 60 kilometers from Seoul on board the ship Michail Shalokhov during the 1988 Olympic Summer Games. Evidently, some athletes were barred from participating in the Olympics [by Soviet coaches] because they would have failed after-competition tests. Those who tested positive were kept from moving into the Olympic Village and competing for the Soviet Union; injury or illness were the reason given for the athletes' withdrawals."
This is by no means to say that non-communists didn't cheat, too, just that the Eastern Europeans were better at it. Robert Voy, the former Chief Medical Officer for the United States Olympic Committee, wrote:
The Eastern Europeans, however, are using drugs more effectively than we are. They have the science of performance enhancement mastered. Their athletes aren't using black market drugs or working under the guidance of steroid gurus. The Eastern Europeans are working with experts.
Any knowledgeable track fan can point to several highly unlikely marks set by extremely pumped-up looking North American runners, but the point is that steroid abuse in the West has been more of an underground, entrepreneurial, hit or miss business, with little of the Prussian efficiency of the East German system.
2. There has been an on-going arms race between dopers and testers. At present, the laboratories would probably catch athlete using steroids as heavily as was common in the Seventies and Eighties. By 1996, "Today's tests are highly sensitive to designer steroids such as Anadrol and Dianabol, which differ chemically from anything the body produces on its own." Random drug testing, introduced in 1989 after the Ben Johnson humiliation, has made a difference, too. Finally, Eastern Europe is a much more open place today than before the fall of the Berlin Wall, so industrial-scale doping on the old East German model can't be kept hidden very well.
For example, consider the fates of Jarmila Kraticholivova and Katrin Krabbe. The extremely muscular Jarmila won the 400m and the 800m at the 1983 WC, and her 800m record still stands, largely unthreatened. One of her rivals, Rosalyn Bryant said, "I'm still not envious of the 'Wonder Woman' of Czechoslovakia. I could have chosen the same way, but I didn't want to change my body, given to me by God, into a new shaper. ... Five years ago she was a normal woman. Now she is all muscles and runs World Records." Runner Gaby Bussmann called her, flatly, "a man." Jarmila replied, "One day, if she produces performances like mine, she will have to have sacrificed some of her good looks. In athletics, one has to decide how much to sacrifice. The women of the West don't work as hard as we do." However, she was never caught by the drug tests of her day.
In contrast, Katrin Krabbe, a product of the old East German training system who the 100m and 200m at the 1991WC, was described by Track & Field News as "sleek" and "beautiful" and pointedly compared to the "masculine" Jarmila. Even before she won both the 100m and 200m in the 1991 World Championships, Katrin had signed a million dollars in modeling and product endorsement contracts. Yet, with the new random drug tests and without the communist regime to shield her, she was disqualified the following year because of attempted tampering with her urine sample.
Similarly, the shot-put and the discuss are universally acknowledged to be the track and field events most driven by steroid abuse. Recent performances are down substantially over the records set in the 1980's.
On the other hand, the light steroid abuser might well continue today to slip through. Furthermore, there are newer drugs like Human Growth Hormone that are not yet tested for.
3. The effect on women of the traditional drugs of choice -- artificial male hormones or anabolic/androgenic steroids -- is more dramatic, since men have so much more of the natural stuff to begin with. On the other hand, the newer drugs like Human Growth Hormone that have become popular as replacements for steroids don't seem to benefit women a lot more than men, and don't seem to be quite as effective as good old anabolic-androgenic steroids.
So, we don't really need to know that much about the magnitude of doping today to understand what has happened to the gender gap. It's possible that more runners are clean of all drugs today. Or, maybe runners are using lighter cycles of steroids to avoid detection. Or maybe they've switched to less effective, but less detectable alternative drugs, which benefit women less disproportionately. Or, most likely, some combination of all three.
Put it all together, and it appears to us that the single most important explanation of the fall and rise of the gender gap since the end of the Sixties has not been the benign one of more liberal social attitudes toward women's sports, Instead, the true story may be more sinister: the rise and fall in the use of masculinizing drugs.
# # #
Back to Steve Sailer Home Page