The Olympic Winter Games: Participation and Performance Gerard Kuper and Elmer Sterken1 Department of Economics University of Groningen The Netherlands

Abstract We analyse the decision to participate and performance at the Modern Olympic Winter Games at the country level. We use an unbalanced panel of 69 countries over all 18 editions of the Winter Games since 1924. The main focus of the paper is on the impact of economic, geographic and demographic determinants of Olympic participation and success. We present separate results for events before and after the Second World War as well as the most recent editions. Since the cross-sectional variation and time variation differ and we are interested in the role of ‘fixed’ factors in explaining cross-sectional differences we present results for event-specific and country-specific intercepts. Moreover, we present forecasts for the Salt Lake City 2002 Winter Games participation and medal tally.

1

We thank Mr Anthony T. Bijkerk, Secretary-General of the International Society of Olympic Historians, for his kind provision of participation data. Gerard Kuper: [email protected], Elmer Sterken: [email protected]

1

Introduction

The Modern Olympic Games attract nowadays an immense amount of attention days. Due to the increase of labour productivity, leisure has become more attractive. Both active and passive sports are popular leisure activities. The economics of leisure therefore cannot neglect the role of sports. In particular the organization of both Summer and Winter Games is a huge but profitable investment project. There are several aspects of the Olympics that make it special in the world of sports. First, its four-year cycle makes the event rather scarce. Second, winning an Olympic title is found to be more valuable for an athlete (at least in the majority of the sports events) than whatever other title. Third, the Olympic Games attract the interest of all sports fans both as spectators and media watchers in the same couple of weeks. And fourth, more than any other sports event, the Olympic Games appeal to nationalistic feelings. People get interested in medal tallies by country, like to hear the national anthem after winning an event, etc. These arguments make the Olympic Games a special event that has economic causes and consequences. Hosting the Games can lead to an economic impulse of a significant magnitude. The investments needed often lead to fiscal expansion in the region. Besides, the huge media attention leads to a large cash flow. But also in participating countries economic links with the Games can be important. National Olympic Committees are able to attract large amounts of funds these days. These funds are partly conditional on future success, so the NOC’s strive for the best selection of athletes to the Games. Winning medals at a large scale is believed to affect consumer confidence and leads to a higher grading of sports in media. Since there are so many economic aspects of the Olympic Games, it is interesting to analyze a few highlights. In this paper we analyze the participation and success of national Olympic teams at all 18 editions of the Olympic Winter Games organized from 1924 up to and including 1998. The goal of the paper is to present the economic, demographic and geographic determinants of national Olympic Winter Games activity. Doing so we are able to describe and forecast Olympic Winter Games participation and success. To our knowledge this paper is one of the very few on the Winter Games so far. An exception is Balmer et al. (2001) who analyse the home advantage in the Winter Olympic Games at the event level. They find that events that are assessed by judges show a

bigger home advantage. We go beyond their scope however and analyse more elements than the home advantage alone (and our object of study is a country instead of an event). There are numerous papers that address the Summer Games. For the postWorld War II games sociologists and economist analyzed the impact of social and economic conditions on the results. Examples of these studies are Ball (1972), Grimes et al. (1974) and Levine (1974). The first study that restarts the performance analysis after two decades of radio silence is Slughart et al. (1993), who analyze the problem for transition economies. Recently two studies, by Johnson and Ali (2000) and Bernard and Busse (2000), relived the attention for this issue for the Sydney Summer Games. Why do we analyze the Winter Games instead of the Summer Games? The first argument, as shown above, is that there are typically more studies available for the Summer Games, so we fill a gap. Next, we think that the nature of the Winter Games is different from the Summer Games (we return to this issue later on), which warrants a separate analysis. Thirdly, we have a separate paper on the Summer Games (see Kuper and Sterken, 2001), using a comparable methodology as in this paper. The Summer Games are typically more popular than the Winter Games. Why would the Winter Games be less interesting than the Summer Games? One can think of a couple of arguments. First of all, the Winter Games are typically smaller in terms of events and participants (see Section 2 for a more detailed discussion). Secondly, participation in the Winter Games is typically restricted to countries with specific climatological conditions. This reduces the attention for the Games in the other countries. Also the nature of the events is typically clustered into a few branches of sports, where in some cases competition is less severe. Think of speed skating for instance, which is popular in a limited number of European countries, the U.S., Canada, and Japan only. But the recent rapid growth of the Winter Games justifies an economic analysis. To what extent are economic factors driving participation and success at the Winter Games? In other studies usually income per head is found to be crucial for a country in Olympic relations. Is it also true for the Winter Games? Is the home bias equally important in the Winter Games as it is for the Summer Games? These and related questions are answered in this paper. We begin with modeling participation. Next we model success in

terms of medal winnings. For a detailed description of our approach we refer to our paper on the Summer Olympics (see Kuper and Sterken, 2001). Finally we present forecasts for the Salt Lake City 2002 Winter Games. We sum up with conclusions. 2

Data

We include data for all modern Olympic Winter Games since 19242. This implies that we include all 18 events up to and including the Nagano 1998 Games in our sample. An overview of the events is presented in Table 1. Table 1 illustrates in a nutshell the statistical history of the Winter Games. It is interesting to compare the development of the Winter Games vis-à-vis the Summer Games. We do so in Figure 1, where we compare the editions in matching years (we match the Lillehammer 1994 Winter Games with the Atlanta 1996 Summer Games and the Nagano 1998 Winter Games with the Sydney 200 Summer Games). Figure 1 presents ratios of Winter Games versus Summer Games for four time series: the number of participants, the number of female participants, the number of participating countries, and the number of events. Figure 1 shows that the average size of the Winter Olympics is a little less than 20 percent of the Summer Games in terms of participants, female participants and events. Figure 1 shows that the participants ratio is rather constant. The lower left panel illustrates that women participated relatively more at earlier editions of the Winter Games. The events ratio has increased seriously though. The countries ratio shows that there has been a relative strong growth of the number of countries that participates in the Summer Olympics. In recent editions of the Games about one-third of the countries that participate in the Summer Olympics participates in the Winter Games. As we will show in the next sections it is better to model participation and medal earnings in shares instead of absolute values. Doing so we are able to forecast participation and medal shares and not the absolute number of athletes sent and the medal rankings. This requires having the absolute values of the aggregated number of participants and number of events. From this perspective it is good to describe the past growth rates of participants and events. Figure 2 gives the data for the post-WWII events (we exclude the pre-WWII events, because both versions of the Games were still 2

Winter sports made their debut at the London 1908 Summer Games in the Figure Skating events. At the 1912 Stockholm games no winter events were held. In Antwerp 1920 Figure Skating and Ice-Hockey were on the list of events.

developing there and demonstrate unstable growth rates). Figure 2 shows that the average growth rate of the number of participants is 7% for the Summer Games and 7.5% for the Winter Games. The number of events of the Summer Games grew by 6% per edition, while the Winter Games events growth rate was 10%. Figure 2 also shows that the editions 18 and 19 of both the Summer (Montreal and Moscow) and Winter Games (Innsbruck and Lake Placid) demonstrate low growth rates. This depression coincides with the development of the world economy in those years during the area of the two oil crises. This concludes our analysis of the Winter Games at the aggregate level. The main attention of our paper is focused on the national level. We collected the medal data from Wallechinsky (2001). We included the 80 countries that won at least one medal at one of the 18 editions. Data on participation by country are given by Kluge (1981) for the Olympic Games up to and including the Lake Placid 1980 games. For the Ice-Hockey participation we used the listing provided by the International Ice Hockey Federation (1984). For the later events we used data from Statistical Annexes of the Official Report of the Games (kindly provided by the International Society of Olympic Historians). Next we collected data on GDP. GDP data are typically hard to find for some countries, especially for those not included in the sets of the International Monetary Fund or World Bank. In our sample this typically holds for Liechtenstein, Monaco, and the Peoples Republic of Korea. The other problem is the provision of consistent estimates of GDP before the Second World War. We used Maddison (2001) for dollar weighed uniform priced GDP. Maddison gives estimates for about 15 countries in our sample back to 1870. This group of countries includes a majority of the countries that participated in the pre-WWII editions of the Winter Games. Data on population are provided by Maddison (2001) and by the World Bank. The World Bank provides moreover a data set on development indicators (see Easterly, 2000). We use this set for other geographical and demographic data, such as longitude and latitude, female labor participation (in 1980), legal system dummies, etc. Overall we are able to estimate the models with data for 69 countries over 18 events. At the end of the sample the number of countries increases, especially due to the participation of the former Soviet states and the splitting up of Yugoslavia.

3

Model and estimation results

Johnson and Ali (2000) propose to model the decision to participate and the probability of success separately. We model success conditional on participation. This is still a shortcut to reality, but makes sense at second thought. Indeed, like in a tournament, participation will be grounded on expected success. But in the decision to participate more motives than expected success alone play a role. In some cases national proud plays an important role. Since we model the results at the national and not at the individual level a strategy to model success conditional on participation seems to be the appropriate one. So we first we estimate participation. Next we model Olympic performance in terms of medal shares for gold, silver and bronze, conditional on participation. We estimate the model in a combined time-series cross-section form. First we use the events as cross-section, after that we use the countries as units to account for time-to-build effects. Throughout we estimate the models with the fixed-effects estimator. Alternative estimators would be the random coefficients model or instrumented panel estimators. The first estimator allows for stochastic differences between cross-section units. Experimenting with this estimator led to inferior results for our models though. The second estimator could be used to correct for endogeneity of the regressors. For instance if one includes a lagged dependent variable. We will use an IV-estimator in the model that explains the success rates conditional on participation. In those models we estimate the shares of e.g. gold medals on the endogenous participation rate. In our dynamic models we don’t correct for endogeneity, because our observation matrix is rather sparse. 3.1

Participation

The dependent variable is PSHit, which represents the fraction in percentages of athletes at game t (t=1,..,18) from country i (i=1,..,80) from the total number of participating athletes. Modeling in shares avoids problems of nonstationarity. Let Pit be the absolute potential number of athletes delegated by country i. Suppose now that each world citizen has an equal probability to become a top athlete. In that case Pit will be dependent on the size of the total population of a country at the time of the tth edition of the Summer Games POPit. There are several valid arguments why Olympic participation is not proportional to the absolute size of the population. Suppose we have a stochastic series X1,…, Xn which is identical independently standard normal

distributed N(0,1). The expected value of the supremum Xsup of all possible outcomes is of order √log(n) (see Reiss, 1989). So it is likely that the maximum performing individual of a population of size POPit will be of the order √ log(POPit). Since this result also holds at the world level, PSHit will be determined by the square root of the log of population share (POPSHit). Next we assume that income per capita (YSHit, in shares of total world income) will determine the training and health conditions of the potential athletes. We measure income by the 1990 Geary Khamis dollar denominated GDP per capita figures as presented by Maddison (1995). We average to Olympic GDP per capita series by taking the arithmetic averages over the last 4 years. We restrict the participation share to be positive (the upper bound of 100 per cent is not binding in any case) by taking the natural logarithm of the participation share: log PSHit = a √log(POPSHit) + b log(YSHit) + Ci

(1)

where Ci represents a country specific determinant. We expect both parameters to be positive. The potential share of athletes PSHit is disturbed by two effects. First we have the home advantage HOMEit (=1 if country i hosts Games t, = 0 in other cases). Home countries are allowed to send more athletes. Secondly, we have the distance DISTit to the Games. We measure the distance by taking the square root of the cubic terms that denote the differences in coordinates of latitude and longitude between the hosting and the visiting country. Note that we correct for taking the shortest route. This leads to the following specification: log PSHit = a √log(POPSHit) + b log(YSHit) + c* HOMEit+ d* DISTit + Ci + eit

(2)

where eit is a white noise residual. Finally, we can model the country specific effects Ci. Here we use geographical data like the latitude and the availability of mountains (for the Alpine Skiing events). In previous studies of the Summer Games, the socialist origin of countries has been found to be relevant to Olympic participation and success. We include all forms of legal systems, like defined by La Porta et al, 1997). For the Winter Games we might use the Scandinavian or German origin. Another determinant is emancipation, which we measure by female labor participation. Especially

for the more recent editions of the Games the number of women events increases. A last word on participation relates to the event Ice-Hockey. If a nation participates in the Ice-Hockey event it currently sends 20 athletes per team (in Nagano for the first time a Women Ice-Hockey event was organized, such that a country could send two teams). Since Ice-Hockey is one of the few team events at the Winter Games (Bob-sleighing is another one), this has a relatively large impact on participation (in a positive sense), but a negative impact on performance given participation. An Ice-Hockey team can only win one medal. This is the reason that in some of the models we include an Ice-Hockey dummy for each country-event observation. Table 2 presents the estimation results of the participation model with eventspecific fixed effects. We include two sets of results: the first with a dummyvariable for participation in the Ice-hockey event, and a second one with “endogenous” Ice-hockey participation. For each set of results we include regressions using all data available, one with pre-WWII events and results for the latest five editions of the Winter Games (the editions will be used to forecast the Salt Lake City 2002 participation). In panel A, the first column gives the results for all events. From this column we can observe that almost all the determinants as discussed have a significant impact on participation. For the legal dummy-variables we only find a significant result for the German legal countries. Also the home advantage is not convincingly determining participation (as opposed to results found for the Summer Games, see Kuper and Sterken, 2001). Income per capita, population, distance to the games, etc. do have their expected impact on Winter Game participation. At the first editions of the Games the model performs worse as can be seen from the second column. Only income per capita, population share and distance to the Games contribute in explaining the variance of particpation shares. For the last five editions we find that the model performs a little better again. Socialist countries perform relatively well, but the distance to the Games is no longer relevant. These results are rather robust to the choice of endogenous Ice-Hockey participation or not (see Panel B in Table 2). Excluding the Ice-Hockey dummy variable increases the elasticities of the other relevcant variables. The fir of the model although decreases. Next we estimate a similar model, but now with fixed effects for the countries (and so variation across the events). This allows for the analysis of

so-called time-to-build effects. If a country wants to participate with a large number of athletes this probably requires more than 4 years of training and acquiring experience, so that it will be likely that the previous participation share will explain the current one. On the other hand we now cannot use all the country-specific fixed effects variables any longer. Table 3 gives the results for both the exogenous (panel A) as well as the endogenous IceHockey participation (panel B). Since the observation matrix is sparse before the Second World War we now give results for all the events, the 1952-1972 and the 1984-1998 events. From the first column of panel A we can observe that the lagged participation share is significantly contributing to the model performance. Also the home advantage, the distance to the Games, and the Ice-Hockey dummy are significant, but the income and population effects are hardly significant. If we use the sub-samples only the home advantage in the period 1952-1972 and the Ice-Hockey dummy for the 1984-1998 Games have a significant impact. A similar picture emerges from the endogenous Ice-hockey participation models in Panel B. Generally, we observe some differences between the models with event-specific dummies and countryspecific dummy variables. One can see that the country-specific intercepts explain participation better than the event-specific intercepts. So the country variance seems to be bigger than the time variance in the Olympic Winter Games data. 3.2

Success

In the previous subsection we modeled national Olympic participation in the Winter Games as a function of income per capita, population, distance to the games and some country specific factors. Now we turn to Olympic success in terms of winning medals. We model the national shares in medal totals MSH. We distinguish the medals by type: gold, silver, and bronze. Our main innovation is that we model medal shares as a function of participation shares. Since participation is endogenous we use the estimated participation results from the previous section. As we illustrate below the data reveal that national medal success is dependent on participation, at least in eventspecific regressions. If we use country-specific intercepts estimated participation shares are no longer relevant in explaining medal success. Apparently, estimating participatiuon helps to explain cross-country variation but not the time variation of medal winnings. We model the share of medals (gold, silver, and bronze) as a function of the participation share, the home advantage, the legal systems and again income

per capita. The existence of home advantage in relation to success has been well documented (see Nevill and Holder, 1999). Courneya and Carron (1992) identify four factors to account for home advantage: crowd factors, learning or familiarity factors, travel factors and rule factors. The home advantage relates to the home crowd that supports the home team. The bigger the crowd the larger the home advantage gets. Familiarity with the venue contributes to the home advantage in e.g. alpine skiing (see for the home advantage in the world cup alpine skiing: Bray and Carron, 1993). Travel or time-zone effects might be another source of home advantage. Finally, officiating might be biased. Seltzer and Glass (1991) analysed the judging of skaters in Olympic Skating events from 1968 to 1988. They found that judges favored their own countrymen. The legal system relates to the fact that some countries might be more restrictive and selective to sending athletes, leading to a higher average quality of the team. In the country-specific regressions we include GDP per capita again as an additional determinant to indicate a higher average quality of a national team. In some of the models we include a dummy-variable for the U.S. and USSR, since those two countries show abnormal results (which might be caused by boycotts). We have again two types of regressions: event-specific fixed effects, where we can include variables indicating differences between countries, and country-specific effects, where we focus on the dynamics of medal winning. Analyzing the data both ways so reveals the importance of cross-sectional variation and time dynamics. Table 4 starts with the basic event-specific regressions. The model includes the home advantage, estimated participation, legal systems and two dummy-variables for the USA and USSR (in Panel C we also use a dummy variable for Russia). Since we include all the events in these regressions we need to correct for the fact that due to boycotts the USA and USSR probably won more medals than normal. Panel A shows that the home advantage seems to affect the winning of silver medals the most. This is quite opposite to the results for the Summer Games, where studies tend to find that there is a bias towards winning gold medals (see e.g. Kuper and Sterken, 2001). Panel A of Table 4 moreover shows that Scandinavian and German legal system countries typically earn more medals. Also the USA and USSR won more medals than the model would normally predict. Finally, estimated participation is an important determinant of success. Panel B of Table 4 shows that this characterisation of results also holds after the Second World War. Panel C shows that in the

last five editions of the Winter Games the home advantage is no longer relevant. Apparently competition is strong and other determinants are more important. The German legal family of countries still has the best performance. Panel C also shows that Russia seems to have taken over the role of the USSR in winning medals in the recent editions. Table 5 presents the results for the model with country-specific fixed effects for both the major post-WWII period and the latest five editions of the Winter Games. We model medal success conditional on the previous success in terms of winning gold, silver, and bronze medals. One could think that past success is a good predictor of future success. For the 1952-1998 sample we typically find only an impact of the bronze medal history. For the Summer Games we found a similar effect for the gold medals (see Kuper and Sterken, 2001). The role of forecasted participation and income per capita is insignificant. For the 1952-1998 we find a serious home effect. A similar picture emerges for the last five editions (1984-1998). The lagged medal success is not relevant (even for the bronze medal). Estimated participation is also not contributing to the explanation, while the home effect is only a little relevant to winning gold medals. The income effect though seems to regain significance (especially for silver and bronze medals). 4

Forecasts for Salt Lake City

How does our model perform out of sample? Here we use the Salt Lake City 2002 games as a test case. We use the models based on the last five editions of the games to predict the participation and success at the SLC event. We present results for both the models that try to explain cross-sectional variance as well as the time variance. We used the World Development Indicator 2001 of the World Bank to forecast income per capita and population averages over the period 1998-2001. First we present the participation results. We assume that there will compete 2500 athletes in SLC. Table 6 gives the forecasts for the national team sizes. Panel A presents the results based on the event-specific estimation results (see Table 2 Panel A last column: we know the Ice-hockey participation by now). Panel B presents the results based on Table 3 Panel A). In both cases the USA have the largest team size, closely followed by Canada, Germany and Russia. There is some variation between the two Panels though. In both panels we listed only teams with more than 10 athletes. In Panel A there

seems to be more concentration in the big teams. In Panel B the median team size is bigger. Next we predict the medal tallies using the same methodology. So we use Table 4 Panel C and Table 5 panel B to forecast success. Table 7 presents the results. In Panel A Russia wins the tournament, while in Panel B the USA is on top. In both cases the USA, Russia and Germany compete for the top positions. Italy typically benefits from the country-specific modelling (Panel B), while Japan is more successful in Panel A. Roughly spoken, Panel B correlates stronger with the Nagano results than Panel A. So both tables reveal a little of the uncertainty to predict medal success.

5

Summary and conclusions

This paper analyses the statistics of the Modern Winter Games since 1924. We present models that explain participation and performance in terms of medal winnings at the country level. Using data for 69 of the 80 countries that ever won a medal we estimate models per event and per country. We test for the impact of income per capita, population size, the home advantage, the distance to the games, legal systems, emancipation, and latitude. Moreover we correct for the impact of the Ice-Hockey event on participation. Finally we model success conditional on participation. We conclude that participation can be explained by basic economic and geographical data. Success is dependent on the home advantage. We don’t find serious evidence for hysteresis of winning medals at the Winter Games. Income per head though seems to get more power again at the recent editions of the games. Finally we present some forecasts for Salt Lake City. These simulations show that the model is quite sensitive to changes in structure. Luckily results in the past are no guarantee for future success, even in sports.

References Ball, D.W. (1972), Olympic Games Competition: Structural Correlates of National Success, International Journal of Comparative Sociology, 15, 186200. Balmer, N.J., A.M. Nevill, and A.M. Williams (2001), Home advantage in the Winter Olympics (1908-1998), Journal of Sports Sciences, 19, 129-139. Bernard, A.B. and Busse, M.R. (2000), Who Wins the Olympic Games: Economic Development and Medal Totals, Yale School of Organization and Management, mimeo. Bray, S.R. and Carron, A.V. (1993), Home Advantage in Alpine Skiing, Australian Journal of Science and Medicine Sport, 25, 76-81. Courneya, K.S. and Carron, A.V. (1992), Home Advantage in Sport Competitions: A Literature Review, Journal of Sport and Exercise Psychology, 14, 13-27. Grimes, A.R., Kelly, W.J., and Rubin, P.H. (1974), A Socioeconomic Model of National Olympic Performance, Social Science Quarterly, 55, 777-782. International Ice Hockey Federation (1984), Ice Hockey and Olympism, Olympic Review, 197, 181-203. Johnson, D.K.N., and Ali, A. (2000), Coming to Play or Coming to Win: Participation and Success at the Olympic Games, Wellesly College, Mimeo. Kluge, V. (1981), Die Olympischen Spiele von 1896 bis 1980, Sport Verlag, Berlin. Kuper, G.H., and E. Sterken (2001), Olympic Participation and Performance since 1896, Mimeo, University of Groningen. Levine, N. (1974), Why do Countries Win Olympic Medals? Some Structural Correlates of Olympic Games Success: 1972, Sociology and Social Research, 58, 353-360.

Maddison, A. (2001), The World Economy, A Millenial Perspective, Development Center Studies, OECD, Paris. Nevill, A.M. and Holder, R.L. (1999), Home Advantae in Sport: An Overview of Studies on the Advantage of Playing at Home, Sports Medicine, 28, 221-236. Reiss, R-D. (1989), Approximate Distributions of Order Statistics, SpringerVerlag, Berlin. Shughart, W.F. and Tollison, R. (1993), Going for the Gold: Property Rights and Athletic Effort in Transitional Economies, Kyklos, 46, 263-272. Wallechinsky, D. (2001), The Complete Book of the Winter Olympics, Salt Lake City 2002 Edition, New York: The Overlook Press.

Table 1 – Modern Olympic Winter Games Year City

Athletes

Female

Countries

Events

1924 1928 1932 1936 1948 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988 1992 1994 1998 2002

294 495 306 755 713 732 818 665 1186 1293 1232 1128 1067 1275 1427 1801 1738 2177

13 27 32 80 77 109 132 144 200 212 217 228 234 277 315 488 522 788

16 25 17 28 28 30 32 30 36 37 35 37 37 49 57 64 67 72

14 14 14 17 22 22 25 28 34 35 36 37 38 39 46 57 61 68 78

Chamonix St. Moritz Lake Placid Garmisch Partenkirchen St Moritz Oslo Cortina d’Ampezzo Squaw Valley Innsbruck Grenoble Sapporo Innsbruck Lake Placid Sarajevo Calgary Albertville Lillehammer Nagano Salt Lake City

Source: Up to and including 1984: V. Kluge (1981), De Olympische Spelen van 1896 tot heden, Elmar Sport, After 1984: Statistical Annexes of the Official Report of the Olympic Games.

Table 2 – Participation at the Games - Fixed effects for events Dependent variable is the log of the percentage participation share log(PSH); Home = 1 if a country hosts the Olympic Games, else 0; Female labor = percentage of female workers in the labor force in 1980; YSH = share of GDP per capita of a country as a percentage of the total worldGDP/capita; POPSH = country population share of the world population; Distance = distance in kilometers from the capital of the host country to the capital of the participating country; Ice-hockey = 1 if a country has sent an ice-hockey team; Mountain =1 if a country has an alpine mountain elevation; Socialist = 1 if the country is ruled under the socialist legal system; Scandinavian = 1 if the country is ruled under the Scandinavian Civil Law system; British =1 if the country is ruled under the British common law legal system; German = 1 if the country is ruled under the German legal system; Latitude = distance to the equator; R2 is the adjusted determination coefficient and SSR is the sum of squared residuals; The White-corrected standard errors are in parentheses.

Panel A – Exogenous participation in the Ice-Hockey event All Games Pre-WWII Last Five 1924-1998 1924-1936 1984-1998 log(Home+1) log(Female labor+1) log(YSH)

√log(100*POPSH) log(Distance/1000+1)

0.335 (0.158) 1.766 (0.380) 0.350 (0.081) 0.822 (0.133) -3.221

0.191 (0.247) 1.100 (1.514) 0.109 (0.215) 1.039 (0.371) -7.996

(0.925)

0.086 (0.316) 0.899 (0.570) 0.518 (0.101) 0.684 (0.165) -2.134

(2.070)

(1.365) log(Ice-Hockey+1) log(Mountain+1) log(Socialist+1) log(Scandinavian+1) log(British+1) log(German+1) log(Latitude2)

1.444 (0.117) 0.849 (0.189) -0.073 (0.245) -0.012 (0.307) 0.254 (0.219) 0.541 (0.182) 0.428

1.550 (0.258) 0.350 (0.353) -0.446 (0.457) 1.082 (0.515) -0.036 (0.351) 0.314 (0.239) -0.017

(0.097)

1.715 (0.202) 0.510 (0.256) 1.215 (0.378) 0.085 (0.484) 0.604 (0.346) 1.200 (0.309) 0.271

(0.257)

(0.076) R2 SSR # countries # country-events

0.703 318 69 565

0.699 16 26 75

0.711 140 68 230

Panel B – Endogenous participation in the Ice-Hockey event All Games Pre-WWII Last Five 1924-1998 1924-1936 1984-1998 log(Home+1) log(Female labor+1) log(YSH)

√log(100*POPSH) log(Distance/1000+1)

0.840 (0.194) 2.624 (0.413) 0.585 (0.084) 1.099 (0.141) -4.602

0.118 (0.381) 2.743 (1.615) 0.782 (0.212) 1.609 (0.534) -10.800

(1.041)

0.820 (0.230) 2.072 (0.676) 0.748 (0.107) 0.857 (0.180) -4.052

(3.284)

(1.500) log(Mountain+1) log(Socialist+1) log(Scandinavian+1) log(British+1) log(German+1) log(Latitude2)

1.206 (0.215) 0.075 (0.285) 0.188 (0.348) 0.315 (0.236) 0.704 (0.205) 0.409

0.499 (0.488) -0.306 (0.580) 0.803 (0.615) -0.004 (0.545) 0.499 (0.367) -0.096

(0.098)

1.049 (0.290) 0.993 (0.470) 0.360 (0.596) 0.473 (0.389) 1.010 (0.360) 0.245

(0.427)

(0.080) R2 SSR # countries # country-events

0.631 396 69 565

0.530 26 26 75

0.632 179 68 230

Table 3 – Participation at the Games - Fixed effects for countries Dependent variable is the log of the percentage participation share log(PSH); Home = 1 if a country hosts the Olympic Games, else 0; YSH = share of GDP per capita of a country as a percentage of the total world GDP/capita; POPSH = country population share of the world population; Distance = distance in kilometers from the capital of the host country to the capital of the participating country; Ice-hockey = 1 if a country has sent an ice-hockey team; R2 is the adjusted determination coefficient and SSR is the sum of squared residuals; The White-corrected standard errors are in parentheses. Panel A – Exogenous participation in the Ice-Hockey event All Games 1924-1998 1952-1972 1984-1998 log(PSH(-1)) log(Home+1) log(YSH)

√log(100*POPSH) log(Distance/1000+1) log(Ice-hockey+1) R2 SSR # countries # country-events

0.206

0.015

0.018

(0.050) 0.390 (0.132) 0.115 (0.095) 1.141 (0.410) -1.899 (0.463) 0.792 (0.081)

(0.095) 0.456 (0.193) 0.201 (0.227) 3.199 (2.246) -0.763 (0.662) 0.935 (0.200)

(0.096) 0.288 (0.166) -0.203 (0.231) 0.678 (1.108) -1.996 (0.661) 0.568 (0.086)

0.876 87 65 476

0.872 22 33 151

0.909 23 63 198

Panel B – Endogenous participation in the Ice-Hockey event All Games 1924-1998 1952-1972 1984-1998 log(PSH(-1)) log(Home+1) log(YSH)

√log(100*POPSH) log(Distance/1000+1) R2 SSR # countries # country-events

0.218

-0.041

(0.055) 0.638 (0.138) 0.227 (0.098) 0.590 (0.442) -2.540 (0.518)

(0.105) 0.798 (0.175) 0.367 (0.221) 0.253 (2.387) -1.826 (0.791)

(0.097) 0.590 (0.229) -0.104 (0.235) 0.589 (1.118) -2.117 (0.683)

0.856 102 65 476

0.843 27 33 151

0.904 24 63 198

0.025

Table 4 – Medal counts with event-specific intercepts Dependent variable: log of the percentage medal share log(MSH+1); Home = 1 if a country hosts the Olympic Games, else 0; Socialist = 1 if the country is ruled under the socialist legal system; Scandinavian = 1 if the country is ruled under the Scandinavian Civil Law system; British =1 if the country is ruled under the British common law system; German = 1 if the country is ruled under the German legal system; USA = dummy-variable representing the USA; USSR = dummy-variable representing the Soviet Union; PSHe = estimated participation share of athletes (results of Table 2 for Panel A and corresponding results for Panel B, both with exogenous Ice-Hockey participation); R2 is the determination coefficient and SSR is the sum of squared residuals; The White-corrected standard errors are in parentheses. Panel A - All Games: 1924-1998

log(Home+1) log(Socialist+1) log(Scandinavian+1) log(British+1) log(German+1) log(USA+1) log(USSR+1) log(PSHe)

Gold

Silver

Bronze

0.678 (0.291) -0.395 (0.138) 1.531 (0.189) -0.071 (0.163) 0.573 (0.182) 1.949 (0.317) 3.594 (0.391) 0.248

0.816 (0.297) -0.212 (0.141) 1.659 (0.193) -0.144 (0.169) 0.770 (0.185) 2.052 (0.322) 2.999 (0.399) 0.252

0.041 (0.282) -0.481 (0.134) 1.093 (0.183) -0.184 (0.158) 0.381 (0.176) 1.078 (0.307) 2.657 (0.379) 0.443

(0.042)

(0.040)

0.429 353 69 565

0.467 320 69 565

(0.041) R2 SSR # countries # country-events

0.449 341 69 565

Panel B - Post-World War II Games: 1948-1998

log(Home+1) log(Socialist+1) log(Scandinavian+1) log(British+1) log(German+1) log(USA+1) log(USSR+1) log(PSHe)

Gold

Silver

Bronze

0.797 (0.315) -0.444 (0.144) 1.047 (0.201) -0.220 (0.167) 0.480 (0.194) 1.764 (0.343) 3.430 (0.380) 0.303

0.953 (0.317) -0.298 (0.144) 1.137 (0.202) -0.359 (0.168) 0.556 (0.195) 1.972 (0.345) 2.794 (0.382) 0.302

0.120 (0.296) -0.539 (0.135) 0.850 (0.189) -0.318 (0.158) 0.340 (0.183) 0.837 (0.322) 2.594 (0.357) 0.451

(0.042)

(0.042)

(0.039) R2 SSR # countries # country-events

0.450 275 69 490

0.443 277 69 490

0.496 243 69 490

Panel C - Post-World War II Games: 1984-1998

log(Home+1) log(Socialist+1) log(Scandinavian+1) log(British+1) log(German+1) log(USA+1) log(USSR+1) log(RUS+1) log(PSHe)

Gold

Silver

Bronze

0.234 (0.577) -0.454 (0.180) 0.800 (0.408) -0.204 (0.168) 0.965 (0.319) 1.907 (0.317) 3.201 (0.290) 3.132 (0.235) 0.275

0.817 (0.618) -0.263 (0.185) 0.601 (0.356) -0.166 (0.167) 0.808 (0.328) 1.349 (0.420) 2.876 (0.390) 2.376 (0.266) 0.299

-0.202 (0.407) -0.526 (0.168) 0.453 (0.336) -0.245 (0.160) 0.544 (0.272) 0.052 (0.440) 2.632 (0.307) 1.157 (0.254) 0.422

(0.054)

(0.055)

(0.053) R2 SSR # countries # country-events

0.520 93 68 230

0.471 99 68 230

0.551 82 68 230

Table 5 – Medal counts with country-specific intercepts Dependent variable is the log of the percentage medal share log(MSH+1); Home = 1 if a country hosts the Olympic Games, else 0; YSH is the share of GDP per capita of a country as a percentage of the total world GDP/capita; Goldsh = share of gold medals; Silversh = share of silver medals; Bronzesh = share of bronze medals; PSHe = estimated participation share of athletes sent by a country (results of Table 4); R2 is the determination coefficient and SSR is the sum of squared residuals; The White-corrected standard errors are in parentheses.

Panel A – Postwar Games: 1952-1998 Gold

Silver

Bronze

log(Home+1)

0.997 (0.295)

1.097 (0.285)

0.122 (0.282)

log(Goldsh(-1)+1)

-0.016 (0.059)

-0.007 (0.057)

0.034 (0.056)

log(Silversh(-1)+1)

0.062 (0.059)

-0.019 (0.057)

0.181 (0.056)

log(Bronzesh(-1)+1)

0.198 (0.055)

0.175 (0.053)

0.125 (0.053)

log(YSH)

0.070 (0.151)

0.129 (0.145)

0.203 (0.143)

log(PSHe)

-0.024

-0.194

0.275

(0.169)

(0.163)

(0.161) R2 SSR # countries # country-events

0.678 119 65 408

0.701 110 65 408

0.691 108 65 408

Panel B – Last Five Games: 1984-1998 Gold

Silver

Bronze

log(Home+1)

1.017 (0.535)

0.699 (0.489)

-0.590 (0.475)

log(Goldsh(-1)+1)

-0.075 (0.101)

0.074 (0.092)

0.138 (0.090)

log(Silversh(-1)+1)

0.106 (0.108)

-0.195 (0.099)

0.036 (0.096)

log(Bronzesh(-1)+1)

0.136 (0.102)

0.070 (0.093)

-0.050 (0.090)

log(YSH)

0.208 (0.365)

0.918 (0.333)

1.194 (0.324)

log(PSHe)

-0.359

-0.392

0.377

(0.338)

(0.308)

(0.300) R2 SSR # countries # country-events

0.693 38 63 198

0.731 32 63 198

0.741 30 63 198

Table 6 – Forecasted participation in Salt Lake City Panel A – Event-specific intercepts Country USA GER CAN RUS SWE AUT CHN FIN SUI CZE FRA SLO JPN BLR UKR KAZ KOR POL LAT GBR HUN ITA AUS BUL NOR CRO ROM NZL SLV DEN ESP IRL NED

Participants 304 304 193 175 148 121 117 116 102 99 71 67 54 44 39 38 33 32 24 23 23 20 19 17 17 16 15 13 13 13 11 10 10

Panel B – Country-specific intercepts Country USA CAN GER RUS SWE FRA AUT SUI FIN UKR ITA JPN CZE NOR BLR KAZ GBR CHN SLO LAT POL KOR AUS SLV BUL EST ROM NED ESP ARG HUN NZL TUR GRE

Participants 373 195 171 160 115 103 102 102 90 90 83 82 77 64 62 57 54 46 46 45 35 31 30 28 28 24 24 21 17 16 16 12 10 10

Table 7 – Forecasted Medal Tally for Salt Lake City Panel A – Event-specific intercepts Country RUS USA GER AUT SUI SWE FIN JPN KOR NOR CAN DEN LIE FRA ICE CHN CZE SLO ITA BLR GBR UKR KAZ POL LAT AUS HUN ESP NED NZL BEL TUR

Gold

Silver 18 12 6 4 4 4 4 3 3 2 2 2 2 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0

Bronze 13 14 6 4 4 4 3 3 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0

7 4 8 5 5 5 5 3 3 2 4 1 1 2 1 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Panel B – Country specific intercepts Country USA RUS GER ITA NOR AUT FIN SUI SWE KOR CAN NED FRA JPN GBR CHN CZE BLR KAZ UKR SLO AUS

Gold

Silver 14 11 10 8 7 4 4 4 4 2 2 2 1 1 1 0 0 0 0 0 0 0

Bronze 9 9 9 4 12 6 3 3 1 2 5 2 2 2 0 3 1 1 1 1 0 0

2 4 8 3 8 7 8 4 2 2 7 3 5 3 0 2 2 2 2 0 2 1

Figure 1 – Comparison of Winter- en Summer Olympic Games 28

24 22

24

20 20 18 16

16 14

12 12 8 8

10

12

14

16

18

20

22

24

10 8

10

12

14

PART

16

18

20

22

24

EVE NT

35

60

30

55 50

25

45 20 40 15

35

10

30 25

5 8

10

12

14

16 FE M

18

20

22

24

8

10

12

14

16

18

20

22

24

C OU N

PART = ratio of the total number of participants of the Winter Games and Summer Games; EVENT = ratio of the total number of events of the Winter Games and Summer Games; FEM = ratio of the total number of female participants of the Winter Games and Summer Games; COUN = ratio of the number of participating countries at the Winter Games and Summer Games. Horizontal axis numbering denotes the numbering of the Olympiads (7 is Paris 1924 for the Summer Games and Chamonix 1924 for the Winter Games, 24 is Sydney 2000 for the Summer Games and 1998 for the Nagano Winter Games).

Figure 2 – Growth Rates of Number of Participants and Events 60

60

40

40

20 20 0 0 -20 -20

-40 -60

-40 11 12 13 14 15 16 17 18 19 20 21 22 23 24

11 12 13 14 15 16 17 18 19 20 21 22 23 24

GR P A RTS UM

14

GRP A RTW IN

30

12

25

10 20

8 6

15

4

10

2 5

0 -2

0 11 12 13 14 15 16 17 18 19 20 21 22 23 24 GRE V E NTS UM

11 12 13 14 15 16 17 18 19 20 21 22 23 24 GRE V E NTW IN

GRPARTSUM = growth rate of the number of participants at the Summer Games; GRPARTWIN = growth rate of the number of participants at the Winter Games; GREVENTSUM = growth rate of the number of events at the Summer Games; GREVENTWIN = growth rate of the number of events at the Winter Games. Horizontal axis numbering denotes the numbering of the Olympiads (11 is London 1948 for the Summer Games and St Moritz 1948 for the Winter Games, 24 is Sydney 2000 for the Summer Games and 1998 for the Nagano Winter Games).