The city of Detroit has had a banner year and a half in hosting big-time sporting events. In July 2005, the city hosted the MLB All Star Game. In February 2006, the city hosted the Super Bowl XL. In May 2006, the city hosted a number of post-season NBA games, with the Pistons making it to the Eastern Conference Finals. During October and November, the city hosted a number of post-season MLB games, with the Tigers making it to the 2006 World Series.

If we believe the politicians and convention bureau folks, with all the big-time sports activity in the Motor City, Detroit should be raking in the dough. Why? Because Detroit isn't on most people's must-go tourism list and thus the additional people coming to the city would add substantially to the local economic activity. This is in contrast to other cities, such as Miami or pre-Katrina New Orleans, where sports fans likely just replace would-be sun tanners or Mardis Gras celebrants, respectively.

How can we get a handle on the impact of these events? One way would be to grab hotel occupancy rates, retail sales tax returns, or other local economic activity measures and estimate whether they experienced an appreciable change in those weeks and months during which the events took place. Perhaps this entry will inspire someone to do so, or I will offer the idea as a potential Master's thesis to one of our students.

In the meanwhile (and because I am lazy) we can look at Comerica Bank's Detroit Business Activity Index. This index is a composite of eight different sectors of the greater Detroit-area's economy. Tourism and tourism related spending is therefore only a portion of the index and it is likely that the manufacturing sector's decline will outweigh anything additional tourism might bring to the city, but in a given month who knows?. Nevertheless, this index provides an initial pass as to what these mega-events meant for the Detroit economy:

The 2006 Super Bowl might have had a slight impact on the local economy but the remaining events didn't seem to have a large contemporaneous impact on the business index. Moreover, the ocular estimator suggests that the mega-events didn't seem to dramatically alter the slope of the index's decline since just after the MLB Allstar Game.

I put together the time series of the index from January 2005 through October 2006 (STATA file here). I first looked at the difference in means between those months in which Detroit hosted a mega-event and those when it did not:

. reg index megaevent 

      Source |       SS       df       MS              Number of obs =      22
-------------+------------------------------           F(  1,    20) =    1.88
       Model |  31.6767677     1  31.6767677           Prob > F      =  0.1860
    Residual |  337.777778    20  16.8888889           R-squared     =  0.0857
-------------+------------------------------           Adj R-squared =  0.0400
       Total |  369.454545    21  17.5930736           Root MSE      =  4.1096

------------------------------------------------------------------------------
       index |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
   megaevent |  -3.111111   2.271672    -1.37   0.186    -7.849736    1.627514
       _cons |   110.1111   .9686442   113.68   0.000     108.0906    112.1317

Those months with a megaevent average approximately 3 points lower than those without, but it is not possible to distinguish that difference from zero (p-value = 0.186). That said, the trend is downward to begin with and most of the events happen towards the end of the time series. I therefore regressed the index against a time trend and threw in the megaevent dummy variable:

      Source |       SS       df       MS              Number of obs =      22
-------------+------------------------------           F(  2,    19) =   11.51
       Model |  202.409125     2  101.204562           Prob > F      =  0.0005
    Residual |  167.045421    19  8.79186426           R-squared     =  0.5479
-------------+------------------------------           Adj R-squared =  0.5003
       Total |  369.454545    21  17.5930736           Root MSE      =  2.9651

------------------------------------------------------------------------------
       index |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       timeb |  -.4547473   .1031936    -4.41   0.000     -.670734   -.2387607
   megaevent |  -1.165803   1.697431    -0.69   0.501    -4.718567    2.386961
       _cons |    114.987   1.308702    87.86   0.000     112.2479    117.7262

The results suggest that during the sample period the index declined .45 each month and that megaevents did not have a significant effect on the business index (p-value = 0.50).

I then added an interaction between the megaevent dummy variable and the time trend to see if the megaevents impacted the intertemporal rate of change of the index:

      Source |       SS       df       MS              Number of obs =      22
-------------+------------------------------           F(  3,    18) =   10.34
       Model |  233.794673     3  77.9315576           Prob > F      =  0.0003
    Residual |  135.659873    18  7.53665959           R-squared     =  0.6328
-------------+------------------------------           Adj R-squared =  0.5716
       Total |  369.454545    21  17.5930736           Root MSE      =  2.7453

------------------------------------------------------------------------------
       index |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       timeb |  -.3751276    .103203    -3.63   0.002    -.5919489   -.1583062
   megaevent |   6.849738   4.230615     1.62   0.123    -2.038454    15.73793
    megatime |  -.5570758   .2729851    -2.04   0.056    -1.130596    .0164445
       _cons |   114.1333    1.28187    89.04   0.000     111.4402    116.8264

The megaevent dummy variable remains insignficant (although it is now positive), the parameter estimate on time remains significant although it declines slightly in absolute value, and the interaction term is negative and slightly significant (p=0.056). The results suggest that when a megaevent occured in Detroit, the level of the index might have increased (the parameter estimate might be positive in a one-tailed test with a p-value around .06, indicating that 6 times out of 100 we would be incorrect stating the parameter estimate was positive). However, the intertemporal decline in the business index was perhaps more than double of what it was in the non-mega-event months. The upshot is that there doesn't seem to have been a detectably large and positive impact of the unprecedented 17 months' worth of sporting events on the Detroit metro-area's business activity index.

While this might not be the best news for those in Detroit, these preliminary results jive with recent work with Dennis Coates (presented last weekend at the Southern Economic Association meetings) suggesting that the impact of mega-events is not tremendously large compared to the local economy (the 2004 Houston Super Bowl is estimated to have generated taxable activity equivalent to approximately 0.02% of the Houston economy).