Andrew Rose describes his new paper with Mark Spiegel:
Trade is around 30% higher for countries that have hosted the Olympics… It turns out that unsuccessful bids to host the Olympics also have an impact on trade, one every bit as big as the effect of actually hosting the games. That is, the Olympic effect on trade is attributable to the signal a country sends when bidding to host the games, rather than the act of actually holding a mega-event. A country that wishes to liberalise its trade may want to signal this by bidding to host a mega-event…
Rome was awarded the 1960 games in 1955, the same year that Italy started to move towards currency convertibility, joined the UN, and, most importantly, began the negotiations that lead two years later to the Treaty of Rome and the creation of the European Economic Community. The Tokyo games of 1964 coincided with Japanese entry into the IMF and the OECD. Barcelona was awarded the 1992 games in 1986, the same year Spain joined the EEC…
Our empirical findings suggest that bidding for the Olympics is a costly policy signal that is followed by future liberalisation. For a country pursuing a trade-oriented development strategy, such an outcome would clearly be attractive.
So they are saying that trade liberalization signaled by an Olympic bid is more effective than regular liberalization?
At first glance, I don’t buy it. Their strategy is to measure the permanent effect of bidding for the Olympics by a dummy variable. So, for example, the Olympics dummy is turned on for all British observations from 1948 (when London hosted) through the end of the sample. Why should I believe that there is a causal signaling effect rather than mere correlation?
I turn to the robustness checks section of the paper and find this (p.18):
We next use a treatment methodology, comparing exports for either hosts or candidate countries with exports for matched counterparts. This allows us to better handle the problem that candidate and host countries for the Olympic games are not randomly selected from our sample. We match observations using a stratiﬁcation technique. Our variables used for matching country-pair*year observations include the logs of: distance, exporter and importer populations, exporter and importer real GDPs per capita; and dummy variables for sharing a common language or border.
As best I can tell, this approach controls for some characteristics that result in bidding for the Olympics, but potentially important variables — like trade policy — are not included. In fact, the only variable in the entire paper that describes trade policy is a regional trade agreement dummy that appears in the gravity regressions. As Spiegel and Rose explain in footnote 3:
We focus our attention on the effects of mega-events on trade rather than trade policy since the latter is difficult to measure. We have experimented with the Wacziarg-Welch measure of trade liberalization, and ﬁnd that it is signiﬁcantly and positively correlated with past Olympic hosting or candidacy, taking into account time- and country-effects and controlling for country size and income. This result is quite consistent with the model we develop below. However, we do not consider this avenue of research to be worth pursuing until we have better empirical measures and models of trade liberalization.
If I read this right, while it’s true that their signaling model makes the prediction that trade liberalization and past Olympic hosting will be correlated, it also suggests to me that you ought to control for trade policy before evaluating the Olympics’ effect on trade. If their story is that bidding for such an event makes forthcoming trade liberalization more credible and causes it to have a greater effect, then shouldn’t we use something like a measure of trade policy plus the interaction of the Olympics dummy and the policy variable?
The authors have clearly put a fair amount of work into this paper, but I hold my prior fairly strongly – I suspect that, while bidding to host the Olympics may often coincide with major trade liberalization episodes, Olympic bids have no meaningful causal impact on trade flows. At present, I don’t understand how their identification strategy would produce evidence that shows otherwise. What am I missing?