TO HOST OR NOT?
AN ANALYSIS OF THE ECONOMIC IMPACT OF HOSTING THE OLYMPICS.
Senior Thesis in Economics: Collins Kalyebi, 2020
ABSTRACT:
Mega sporting events like the Olympics tend to attract a large pool of participants that compete with the
hope of winning the bid to host the games. Participants tend to prioritize different interests, though what
seems more apparent, for all of them, is their appetite for an economic windfall that is often associated with
hosting the games. Olympics have cost over USD 8.9 billion on average in the past decade and considering
this staggering amount, hosting the games could be the most financially risky mega project any city could
want to take on. In this paper, I focus on five hosts (USA, Canada, Japan, France, and South Korea) to build
on the findings of past research to further explore the economic worthiness of hosting the Olympic games.
Impact assessment will be done through GDP per capita, Unemployment rate and Export growth rate.
Results from the analysis show that host nations tend to see economic gains, irrespective of the high costs
involved.
INTRODUCTION
The Olympics is by far one of the largest mega-sporting events the world has yet to see with one
of the highest attendances and the largest budgets in terms of host expenditures. The modern Olympic
games started in 1896 and happen every four years, attracting a large pool of participants, who compete at
different levels, and supporters from almost every country in the world. Host cities are announced seven
years prior to the games to allow enough preparation time as a lot of infrastructure development may be
needed to see off the games. On average, over two hundred nations compete in the Olympic games and
over four billion viewers are estimated worldwide. Hosting the Olympic games is therefore prestigious, as
it allows host countries and cities to gain a platform on the international scene.
Before the competitions however and before the host cities are announced, countries have to face
off in a rigorous and quite competitive bidding process. The bidding process usually attracts a large pool of
countries that compete with hope to host the games and hopefully reap from the international exposure that
comes with it. The number of countries that participate in the bidding process took a sharp rise after a
successful and profitable execution of the 1984 games in Los Angeles, but recent statistics show that cities
are now turning away the opportunity to host the prestigious games. This shift is associated with the rise in
expectations from host countries and the high costs involved. This on the other hand means less or no profits
for the host cities and often huge debts left behind after the games.
The parties that say hosting the Olympic games is of economic benefit to the host countries often
argue that the games are a platform for international exposure which boosts tourism, drives growth of local
economies, increases the country’s global trade performance, and creates a sense of national pride for the
hosts. Host cities often see an increase in foreign visitors who mostly come to support the games. On the
international scene, winning the bid to host the games could also reflect a favorable economic environment,
which attracts investors and cities often end up with new infrastructure developments that could otherwise
not have come sooner. But many economists argue that these benefits are quite minimal given the high
costs incurred by host cities in preparation for the games and do not justify hosting the games as a good
investment under most circumstances. The high costs of setting up Olympic venues, putting up transport
infrastructure and housing facilities for the visitors usually outweigh the revenues generated from the games
and quite often cities are left with huge debt.
Staging of the Olympic games in the earlier stages was manageable to the host cities given the
smaller number of participants and the fact that most bidding cities were from already developed countries
and didn't need a lot of new infrastructure developments to host the games. These cities therefore had the
financial muscle that was needed to finance all the activities and for most of the early hosts, it wasn't much
about the financial gains, rather more about the Olympic pride and their political positioning at the world
stage. Past the 1970s though, the number of competitive games at the Olympics was increased and this
drove up participation to almost double the original numbers. This also drove up the costs of hosting the
games and the rise in expectations from the hosts continues to push up the price tag. The average cost for
the Olympic games that have been staged over the past decade is estimated at USD 8.9 billion. And this is
exclusive of costs incurred in preparation for the bidding process, which are in millions of dollars. This is
partly covered by the International Olympics Committee, though the largest portion of the bill remains
allocated to the hosts, who often foot that bill through public sector funding (taxes) or through loans.
Cities have to prove that they have the needed infrastructure and the financial muscle before they
are allowed into the final stages, which include a secret ballot vote by the IOC to decide the winner. The
final vote is politically influenced but the financial ability of the city is approved only after a careful
examination by the IOC. The examination takes a multiplier effect approach, using the estimated direct
expenditures to then estimate indirect expenditures associated with the games. The weakness of this
approach is the fact that it relies on estimates which are likely to be inaccurate. The excitement that comes
along with being a potential host often forces cities to underestimate their costs as a way of increasing their
chances since doing so elevates them to a better financial standing. As a result, cities experience cost
overruns 100% of the time with an average real overrun of 156% (Flyvbjerg and Stewart, 2016).
When considering the bid, different cities and countries tend to prioritize different interests, though
what seems more apparent, for all of them, is their appetite for an economic windfall that is often associated
with hosting the games. It’s no doubt that there is a substantial inflow of money from foreign visitors who
come to watch the games and from the selling of broadcasting rights, but that's usually not enough to cover
for the costs incurred. Los Angeles, who were the only city to bid for the 1984 summer games, turned out
to be the only city to profit from hosting the games. This was evidently due to the fact that their sole bid
put them at an advantage to negotiate favorable terms with the International Olympics committee. They
also relied on using existing facilities which together coupled with the growth in broadcasting revenue,
allowed them a clause of up to $215 million operating surplus.
Revenues from hosting the Olympic games are gained through tickets, sponsorships, television
rights, and licensing. There has been a fairly steady positive growth in revenue generated from the games
over time, but it still lags way behind the rising costs and because of that, host countries find themselves
recovering just a small fraction of their total expenditures, which also explains why host cities often end up
in debt. In 2012, London’s summer Olympics generated over $5.2 billion in revenue, which is far less when
compared to over $18 billion in the incurred costs. Even worse, this revenue is not just for the host cities.
A large portion of revenue generated from broadcasting rights remains with the International Olympics
committee. The most notable city to have fallen in the trap was Montreal, which ran into billions of dollars
in cost overruns and ended up with an estimated 1.5 billion in debt. It took nearly 30yrs to get it all cleared.
It is therefore imperative that a clear economic impact analysis is done and taken into consideration
before making the decision to take part in the bidding. Unfortunately, for the general public and the policy
makers who are usually very excited for the party to come, they often pay less attention to the financial
risks involved with hosting the games. This excitement often results in cities underestimating the costs of
hosting the games and on the other hand, overestimating the potential economic benefits of being the host.
And as a result, host cities run into negative fiscal shocks which alter economic performance and trap locals
in the financial burden resulting from hosting the games. This paper assesses the cause of excitement from
countries offering themselves as hosts while weighing into the costs of hosting the games. It examines
variance in the host country’s GDP% growth, as a result of changes in the host country’s exports,
unemployment rates, and government expenditure, between the years the country hosted the games and
when it didn't. It focuses on the USA, Canada, France, Japan and South Korea for the periods between 1969
and 2017.
LITERATURE REVIEW
The Oxford Olympics Study 2016: Cost and Cost Overrun at the Games: In this paper, Bent
Flyvbjerg and Allison Stewart of the University of Oxford carried out a study to weigh in on the costs
involved, the cost overruns incurred and the cost risks of hosting the Olympic games. This research was
done with the aim of informing and inspiring development of more reliable budgets for the games by
bringing forward the cost overruns from more than a third of the games between 1960 and 2016 that
seemed to be unknown to most people. The paper also aimed to test the effectiveness of the Olympic
Game Knowledge Management Program (OGKMP), established to transfer skills from former host cities
to new ones, in reducing cost risk from previous games while benchmarking cost and cost overrun for the
Rio 2016 against previous games. The study focused on 19games for which they could get all the data
points.
There has been prior research about the costs of the games, but there seem not to be attempts to
specifically and systematically evaluate the costs. Also most of the early research tends to focus on
understanding whether hosting the games is a viable investment from the perspective of cost-benefit
analysis. That though often leaves a debate on what exactly should be measured when determining the costs
and benefits of the games to the host country. Underlying results show that there could be different benefits,
some of which can’t be weighed in financial terms and it's very likely that countries have different priorities
when they chose to host such mega sports events. For example, legacy benefits or national pride that are
intangible make it hard to compare the valuations across countries. Comparisons were made between the
actual outturn costs with the estimated costs stated in the bid year to calculate the cost overruns. This was
followed by a number of statistical tests and comparisons across time and geographies in real terms to
ensure that fair comparisons are made between the hosts.
Results from this research show that the average actual outturn cost for summer games is USD 5.2
billion and 15 billion for winter games, after excluding wider capital costs for general infrastructure, which
are often larger than sports-related costs. Results also show that host cities experience cost overruns 100%
of the time with an average real overrun of 156%, the highest of any type of megaproject. The highest cost
overrun for the summer games was reported in Montreal at 720% and Lake Placid for winter games at
324%. On that note though, research shows that the OGKMP has helped reduce the cost risk for the games
as cost overruns have gone down since the establishment of the program. This alone is a reflection of the
risks of large cost overruns that are inherent to the Olympic games and should be a concern to countries
planning to host the games, especially those with relatively smaller and fragile economies.
It’s very unlikely for host countries to not go into cost overruns given the set-up of hosting the
Olympics. The bidding process of the games is set to look like a show where cities are likely to overestimate
their ability to host the games in order to lue those on the committee to vote for them. Cities that are often
selected to host the games have never hosted such a mega event and often have to start close to zero in
preparation to host the games. And those often in charge of the preparation process have no experience
organizing such a project before. In all these cases, most of the cost overruns come out as unexpected or
changes in project specifications. The time between the announcement of the next host and the actual year
for the games is also long and different events could alter the expected expenditures on the games. For
Example, the 9/11 incident that happened a year before the 2002 winter games in Salt Lake City meant that
the host city had to incur extra costs to beef-up security at the games.
The Olympic effect: This paper by Rose and Spiegel buries the general skepticism around the economic
benefits of hosting the Olympics that usually arises from the fact that hosting the event is associated with
huge cost demands and yet there are relatively fewer tangible benefits that come with it. It examines the
economic impact of hosting mega-sporting events like the Olympics with a focus on the impact on trade. It
uses a variety of trade models and employs the Gravity model of international trade, which expresses
bilateral trade flows between a pair of countries as a function of the distance between the two countries and
their economic masses, to empirically argue the assumption that hosting a mega-sporting event would
provide visibility to a host country and thus stimulate global demand for its exports. Analysis is also done
on countries that participated in the bidding but were not selected to host the games. The implicit assumption
in comparing the trade effects of Olympic hosting and candidacy was that failed candidacies form a valid
quasi-experimental counterfactual control group for Olympic hosts (Rose and Spiegel, 2010). Other factors
that affect trade intensity, like distance and existence of trade agreements are also considered.
Results showed that hosting a mega event has a positive impact on national exports and that
countries that participate in the bidding process but never win the vote to host the games experience similar
effects and on average see a general increase in their exports. Host countries are seen to register an increase
of about 36% in exports, a boost that is relatively more significant compared to factors like trade agreements
and the distance between the countries. The sensitivity analysis that followed the results also showed that
this enhancement in trade tends to be permanent and that hosting winter games had a relatively weaker
effect on exports. Winter games are however seen to be dwarfed by the summer games mainly because they
have specific geographic requirements that limit potential hosts.
Based on these observations, the paper concludes that the effect of hosting mega-sporting events is
attributable to the signal a country sends when bidding to host the games rather than the act of actually
hosting the game. The question that remains though is that, if countries don’t actually have to host to see
this impact on trade, why then do countries still compete fiercely for the right to host the games? Some of
that could be explained by factors like, national pride, boost in tourism and long-term infrastructure
developments, that may not have been considered in this paper.
A Ticket to the Olympics: An Assessment of the “Olympic effect” on Tourism (Laura Costanzo,
2010): This paper examines the magnitude of the Olympic effect on tourism in the host country. It uses
primarily OLS regressions to empirically examine whether hosting the Olympics boosts tourism, and if so,
to what magnitude is the effect and whether there is a significant difference between the tourism generated
by the summer and the winter Olympic games. It uses the White’s heteroskedasticity correction with
clustering on countries to take into account the possibility of within-country correlation in the errors. The
travel variables used are also modified to take into account the possible naturally occurring time trends that
may cause continual increases in tourism. And dummy variables are employed to specify the year and
country. These dummy variables are used to explain the variations that may be as a result of outside factors
that vary by country and year.
The research focused on 112 countries all with 2006 populations of three million people or more,
while excluding those with ten or more missing values of the travel variable (REAL_TRAVEL_ML), which
represents total expenditures (in millions) on all goods and services obtained from a country’s economy by
travelers within that country during a stay of 12months or less. This research is also extended to cities that
placed unsuccessful bids. The decision to look at countries with unsuccessful bids assesses the assumption
that placing a bid usually shows the openness and readiness of the country to take position on a world stage.
And even though these countries do not win the bid to host the games, they still get international attention
by simply participating in the bidding process, which may with time drive interest from international
visitors and boost the total tourism.
Results from this research show significant increases in tourism for the host countries. Surprisingly
though, similar increases are observed for countries with unsuccessful bids. On the other hand, summer
games are shown to have more global attention and therefore larger impacts on tourism compared to winter
games. Results also show that the visible increases in tourism for the host countries and the unsuccessful
bids are greatly dependent and influenced by the country profile and year factors. Beyond these variables,
most of the significant results and positive associations were found to be byproducts of natural occurring
time trends, which cause sectors like tourism to increase over time. In other words, this research showed
that there are other external and internal factors that weigh heavily on tourism trends which also means that
making a close to accurate quantification of the Olympic effect on tourism is to some extent impractical.
DATA/ METHODOLOGY
Theoretical expectations are that by hosting such mega events, countries should see a significant
growth in their economy. A positive relationship between economic performance and hosting the games is
therefore expected. Hosting the games comes with economic opportunity that drive exports of the host
country, attract foreign visitors and investors, and create employment for the locals. This growth in
economic activity usually translates into a boost in the country’s GDP per capita (GDPpc). Results that
show a positive impact of the Olympics on the economy of the host country would also be in support of the
results obtained by past researchers like Rose and Spiegel (2009) who found that hosting the games
positively drives the host country’s internal and external economic activities.
The original intention of this research paper was to use a model similar to the one used by Rose
and Spiegel (2009) where I track economic performance of the countries in question 8yrs before the games
and 8yrs after the games. Other variables like number of tourist arrivals, population, and Exports were to
also take a similar trend for fair comparison. But I later ran into data problems and realized that for countries
that have recently hosted the games, Brazil, China, and South Korea, it wouldn’t be possible to get
consolidated data for the 8yrs after the games. Other variables like the number of tourist arrivals also had
data that only dated back to 1995, and it would be hard to recover unbiased data for the earlier years given
there is no other database with reliable data for those years. To consolidate for all these challenges, I decided
to use panel data format and focused on the USA, Canada, Japan, France and South Korea as the host
countries. For these countries, I recorded their past data between 1969 and 2017 and that could allow me to
get a reasonable sample size for reliable results.
To quantify the economic impact of hosting the Olympic games, I used the World Bank and the
Federal Reserve Database (FRED) to trace annual data for GDPpc, Unemployment rates, and Exports for
countries that have hosted the Olympic games at least twice during the period between 1969 and 2017. The
World Bank and FRED are reliable sources, given their consistency in tracking development indicators like
GDP, Unemployment rates and the major trade indicators for different countries.
For the analysis, I use the Ordinary Least Squares (OLS) regression method. The choice of OLS as
the empirical method is based on its simplicity and its ability to minimize variance in addition to a number
of other useful characteristics that come with OLS estimates. The formulated multivariate regression
equation uses GDPpc as the dependent variable and Exports, Unemployment rates for the host countries as
independent variables. I also added dummy variables to the regression model to allow for a fair comparison
and minimize specification errors. The first dummy variable was to determine which year the country
hosted the Olympic games. ‘1’ is used to denote the year the country hosted the games and ‘0’ otherwise.
For the second dummy variable ‘1’ is used to denote summer Olympics and ‘0’ otherwise. I also added a
dummy to determine if the games were winter games and used ‘1’ to signal yes and ‘0’ otherwise.
GDPpc = ꞵo + ꞵ1(EXP) + ꞵ2(UNP) + ꞵ3(SUM) + ꞵ4(WIN) + ꞵ5(HY) + ε
RESULTS AND ANALYSIS
I initially did the analysis using the data as collected but noticed huge standard deviations within the data
sets. I decided to log the data as a remedy for the huge standard deviations and then followed up with other
diagnostic tests.
Diagnostic tests: To ensure the quality of my data, I performed a series of diagnostic tests. The first test
was to check for multicollinearity. Collinearity occurs when two or more variables in the regression
equation are closely related, which results in a lower and less reliable t-statistic of the independent variable.
I used the ‘corr’ function in STATA to get the numerical comparison of the collinearity between the
variables and then used scatter diagrams to visualize it. The scatter graphs output reflected different levels
of collinearity between the variables with the most significant being between the logged GDPpc and the
logged exports. Logged GDPpc and logged unemployment rates on the other hand show a quite minimal
collinearity which in this case didn’t call for attention. Similar to that was the displayed relationship
between the logged exports and the logged unemployment rates. I decided to do nothing in an attempt to
correct for the significant collinearity between the logged GDPpc and the logged exports. I believe exports
are a good measure of trade for a country and in this I couldn’t find a better way to track trade than through
the exchange of goods and services between countries. On the same note, I noticed that in whatever case,
improvements of trade positively correlate with GDPpc and that’s because generally, improvements in trade
increase GDP. Secondly, I ran the Heteroskedasticity test. Heteroskedasticity basically arises when the
variance of the error term differs across observations. For this I plotted potential residuals against the
potential fitted values and from the reflected graphs there seemed to be no signs of heteroskedasticity.
GDPpc Results: The results from the stated regression model show some expected findings. Countries that
host the games are seen to have a relative boost in GDPpc with improvements in trade driven by relative
growth in exports and reductions in unemployment. The host year variable is also seen to have significant
impact. There is no surprise that the coefficient for unemployment is negative. It proves that over the period
of hosting the games, unemployment rates will be lower. This can be due to the increase in project activities,
including construction, and increase in service activities like tourism which as a result boost demand in the
respective sectors and help reduce unemployment. Similarly, attempting to host the games is reflection of
a good and stable economy. This often attracts outside investments and signals readiness for trade with
other nations. This explains the positive relation of the games with exports as countries that host the games
find themselves involved in more trade with their near and sometimes far neighbors.
PRACTICAL APPLICATIONS
Hosting the olympic games would be any cities dream. It not only fun for the citizens but also gives
a chance to countries who want to show case their cities and, in the process, attract tourists, investment and
trade in general. Following the results of this research, cities that host the games benefit in most cases
irrespective of the high costs, but the risk of making losses and ending up in debt remains high and therefore
any cities that are thinking of bidding to host the games would need to clearly assess, not only their financial
ability to fund the games, but also their ability to recover in any case of such major losses that may hurt the
economy. Cases where countries have made large losses arent very rare. We have seen it happen in Brazil
(Rio) , Greece (Athens) , Canada (Montreal) and other places. Its unlikely that the IOC or any country
planning on bidding for the games will take a look at my paper, but in any case, I would give the following
recommendations.
To the IOC, I would argue that priority should be given to cities/countries that have hosted the
games before or have a reasonable infrastructure already in place to be used in the process. Infrastructure
developments are the major reasons for the high costs of hosting the games, and for cities/countries that
have never hosted the games, they would have to start from zero. On the other hand, countries that have
experience of hosting the games, are not only aware of any unexpected costs involved, they also don’t have
to build everything from scratch. This makes a very big difference both in terms of the final costs of hosting
the games and in the quality of cost estimations at the beginning of the bidding process.
To any city/country planning to take the first bid, I would recommend that rigorous, costs-benefit
assessments are made, and structures put in place to cater for recovery in case of an economic turn down
as a result of taking on the host responsibility. Cities usually fall short of this, and their cost-benefit
estimates are all but driven by excitement. Its definitely exciting to win the bid to host the games but paying
clear attention to the numbers can help save the host from the tuff economic times that usually follow the
games. It’s also hard to clearly forecast the economic position of the city/country given the seven-year lag
that lies between being announced and finally hosting the games. A lot of things can change in seven years
so it’s always important that a plan-B, usually reserve funds, is put in place to support the city/country when
economic conditions following the Olympic games turn out unfavorable.
LIMITATIONS
There are several limitations with this research. First is with the data and country specifications. I
limited my research to the period between 1969 and 2017 and to only five countries (USA, Japan, Canada,
France and South Korea). Data specifications were a result of the fact that most data only dates back to
1960 and for countries like South Korea, it only goes back to 1969. Few countries have hosted the games
between this time and for most of, it was only once and in the early years which means the economic
significance of hosting the games may have faded out with little impact to their current economic
performance. Countries like German and Russia also had incomplete data given the fact that during the
early days German was still dived into West and East German and Russia was still part of the Soviet Union.
Second is the fact that the research uses country data instead of data from the specific host cities.
The data from the specific cities isn’t available and the variance between country data and city data could
be significant to alter the results. The impact of the games on the specific city is also likely to be drastically
different. On the same note, its almost impossible to control for other factors that may affect the used
independent variables. For example, there are many other factors that may affect unemployment rates and
exports. That being said, there might be changes in these variables that are not connected to the Olympics.
The model also misses some of the important variables that are likely to be impacted by the
Olympic games. One of these factors is tourism, which I believe is one of the major income drivers for
Olympic hosting cities. Other benefits like Olympic legacy cannot be quantified which also limits the
impact coverage. Including data from tourism for example and finding a way to quantify some of these
other benefits would provide a wider coverage of the economic impact of the games.
CONCLUSIONS
The main objective of this research was to assess the economic impact of hosting the Olympic
games and whether it is a worthy investment any city. Broader studies have been done on the topic, but for
this research I chose to specifically focus on five countries, that had hosted two or more games between the
period of 1969 and 2017 and for which I could get the data. The data was collected mainly from the World
Bank database and FRED and qualitative analysis focused on GDP per capita growth as the dependent
variable and Unemployment rate, Exports as the independent variables. Dummy variables for the host year,
summer games and winter games were also added to allow for further specifications in the model.
The results from the model show that hosting the Olympic games ends up being beneficial to the
host nations anyway. This was evident with a clear cut in unemployment and positive change in exports for
the host nations. That being said however, its important that countries planning to participate in the bid to
host the games should always make a clear analysis to ensure that the expected benefit is obtained.
Otherwise, negative economic impacts as seen after the games in Rio, Athens and Montreal are still possible
and the ever-raising costs pose very high risks to the host nations.
REFERENCES
Andranovich, G., Burbank, M. J., & Heying, C. H. (2001). Olympic Cities: Lessons Learned from MegaEvent Politics. Journal of Urban Affairs, 23(2), 113. https://doi.org/10.1111/0735-2166.00079
Baade, R. A., & Matheson, V. A. (2016). Going for the Gold: The Economics of the Olympics†. Journal
of Economic Perspectives, 30(2), 201–218. https://doi.org/10.1257/jep.30.2.201
Barros, C. P., Ibrahım
́ o, M., & Szymanski, S. (2002). Transatlantic sport the comparative economics of
North American and European sports. Cheltenham, UK: E. Elgar.
Bista, R. (2017). Revisiting the Olympic Effect. Review of International Economics, 25(2), 279–291.
https://doi.org/10.1111/roie.12266
CHEN, Y., QU, L., & SPAANS, M. (2013). Framing the Long-Term Impact of Mega-Event Strategies on
the Development of Olympic Host Cities. Planning Practice & Research, 28(3), 340–359.
https://doi.org/10.1080/02697459.2013.750072
Costanzo, L.L. (2010). A Ticket to the Olympics: An Assessment of the "Olympic Effect" on Tourism.
Flyvbjerg, Bent & Stewart, Allison & Budzier, Alexander. (2016). The Oxford Olympics Study 2016:
Cost and Cost Overrun at the Games. Saïd Business School Working Papers.
10.2139/ssrn.2804554.
Fourie, J. (2017). Should South Africa host mega-events?: Evidence has shown that the costs of putting
on events such as the FIFA World Cup or Olympics outweigh the benefits. So why are countries
still so eager to host them? Finweek, 6.
McBride, J. (2018, January 19). The Economics of Hosting the Olympic Games. Retrieved January 2,
2019, from https://www.cfr.org/backgrounder/economics-hosting-olympic-games
McCoy, T. L. (2016). Why Brazil’s post-Olympic hangover will hit so hard. Finweek, 12–13.
Pagels, J. (2016). Olympic Ruins. Reason, 48(4), 44–49.
Rose, A., & Spiegel, M. (2011). THE OLYMPIC EFFECT. The Economic Journal, 121(553), 652-677.
Retrieved March 11, 2020, from www.jstor.org/stable/41236997
Streicher, T., Schmidt, S. L., Schreyer, D., & Torgler, B. (2017). Is it the economy, stupid? The role of
social versus economic factors in people’s support for hosting the Olympic Games: evidence
from 12 democratic countries. Applied Economics Letters, 24(3), 170–174.
https://doi.org/10.1080/13504851.2016.1173175
APENDIX
Data Summaries
. summarize logGDPpp logGNI logUNP logEXP, detail
logGDPpp
1%
5%
10%
25%
Percentiles
-2.662654
-.6979884
-.299105
.4927489
50%
.9911544
75%
90%
95%
99%
1.486814
1.944531
2.226467
2.4497
Smallest
-5.613308
-2.814983
-2.662654
-2.197561
Largest
2.430517
2.4497
2.483065
2.55061
Test Results
Obs
Sum of Wgt.
213
213
Mean
Std. Dev.
.871639
1.007735
Variance
Skewness
Kurtosis
1.015531
-1.926738
11.43107
1%
5%
10%
25%
50%
1.235558
75%
90%
95%
99%
1.683438
2.083657
2.309248
2.620401
Smallest
-6.453255
-4.932073
-2.346158
-1.930224
Largest
2.60842
2.620401
2.663687
2.685165
Number of obs
Number of groups
R-sq:
Obs per group:
within = 0.2173
between = 0.5743
overall = 0.3366
corr(u_i, X)
Coef.
224
224
logUNP
logEXP
HY
WIN
SUM
_cons
-.5873478
.3128542
.6192715
-.5694409
0
1.332658
Mean
Std. Dev.
1.115312
1.026011
sigma_u
sigma_e
rho
0
.60282651
0
Variance
Skewness
Kurtosis
1.052698
-3.135799
20.73256
Obs
Sum of Wgt.
=
=
193
5
min =
avg =
max =
35
38.6
43
=
=
95.39
0.0000
Wald chi2(4)
Prob > chi2
= 0 (assumed)
logGDPpp
logGNI
Percentiles
-2.346158
-.4845657
.1126374
.7918599
Random-effects GLS regression
Group variable: ID
Std. Err.
z
.0918735
.0558901
.3384484
.4732347
(omitted)
.2044768
P>|z|
[95% Conf. Interval]
-6.39
5.60
1.83
-1.20
0.000
0.000
0.067
0.229
-.7674166
.2033117
-.0440752
-1.496964
-.407279
.4223968
1.282618
.3580821
6.52
0.000
.9318912
1.733425
(fraction of variance due to u_i)
logUNP
1%
5%
10%
25%
Percentiles
.1823216
.7241612
.9066443
1.280934
50%
1.686399
75%
90%
95%
99%
2.04122
2.294654
2.373603
2.515032
Smallest
.0953102
.0953102
.1823216
.2623643
Largest
2.491137
2.515032
2.530995
2.533141
Heteroskedasticity Testing
Obs
Sum of Wgt.
245
245
Mean
Std. Dev.
1.636323
.5352312
Variance
Skewness
Kurtosis
.2864724
-.4915654
2.656246
Percentiles
-.8194209
.1289541
.5742862
1.380181
50%
2.010246
75%
90%
95%
99%
.
2.428042
2.807168
3.11906
3.662041
Smallest
-2.292272
-.9298691
-.8194209
-.7484584
Largest
3.597511
3.662041
3.758078
3.926374
Iteration
Iteration
Iteration
Iteration
0:
1:
2:
3:
log
log
log
log
likelihood
likelihood
likelihood
likelihood
=
=
=
=
-197.79922
-193.64875
-193.64744
-193.64744
Heteroskedastic linear regression
ML estimation
Log likelihood = -193.6474
logEXP
1%
5%
10%
25%
Fitting full model:
Obs
Sum of Wgt.
214
214
Mean
Std. Dev.
1.823786
.9218406
Variance
Skewness
Kurtosis
.8497902
-.8696173
4.656308
Number of obs
=
193
Wald chi2(4)
Prob > chi2
=
=
97.93
0.0000
logGDPpp
Coef.
logGDPpp
logUNP
logEXP
HY
WIN
SUM
_cons
-.5873478
.3128542
.6192715
-.5694409
0
1.332658
.0906757
.0551614
.3340356
.4670645
(omitted)
.2018107
-6.48
5.67
1.85
-1.22
0.000
0.000
0.064
0.223
-.7650688
.2047399
-.0354262
-1.484871
-.4096268
.4209685
1.273969
.3459887
6.60
0.000
.9371166
1.7282
lnsigma2
_cons
-.8311678
.1017973
-8.16
0.000
-1.030687
-.6316488
Std. Err.
z
P>|z|
[95% Conf. Interval]
Hausman test
Correcting for Heteroskedasticity
Random-effects GLS regression
Group variable: ID
Number of obs
Number of groups
R-sq:
Obs per group:
within = 0.2173
between = 0.5743
overall = 0.3366
corr(u_i, X)
=
=
193
5
min =
avg =
max =
35
38.6
43
=
=
415.26
0.0000
Wald chi2(4)
Prob > chi2
= 0 (assumed)
(Std. Err. adjusted for 5 clusters in ID)
logGDPpp
Coef.
logUNP
logEXP
HY
WIN
SUM
_cons
-.5873478
.3128542
.6192715
-.5694409
0
1.332658
sigma_u
sigma_e
rho
0
.60282651
0
Robust
Std. Err.
.162082
.1084466
.1409918
.1465003
(omitted)
.2204621
z
P>|z|
-3.62
2.88
4.39
-3.89
0.000
0.004
0.000
0.000
-.9050227
.1003028
.3429327
-.8565762
-.2696729
.5254056
.8956103
-.2823057
6.04
0.000
.9005607
1.764756
R-sq:
Obs per group:
within = 0.2173
between = 0.5743
overall = 0.3366
Coef.
sigma_u
sigma_e
rho
0
.60282651
0
=
=
193
5
min =
avg =
max =
35
38.6
43
=
=
95.39
0.0000
Wald chi2(4)
Prob > chi2
= 0 (assumed)
.3128542
-.5873478
.6192715
-.5694409
0
1.332658
.3128542
-.5873478
.6192715
-.5694409
Std. Err.
.0558901
.0918735
.3384484
.4732347
(omitted)
.2044768
z
P>|z|
[95% Conf. Interval]
5.60
-6.39
1.83
-1.20
0.000
0.000
0.067
0.229
.2033117
-.7674166
-.0440752
-1.496964
.4223968
-.407279
1.282618
.3580821
6.52
0.000
.9318912
1.733425
(fraction of variance due to u_i)
Robust test using GNI
Random-effects GLS regression
Group variable: ID
Number of obs
Number of groups
R-sq:
Obs per group:
within = 0.1132
between = 0.3631
overall = 0.1638
=
=
202
5
min =
avg =
max =
37
40.4
43
=
=
718.17
0.0000
Wald chi2(4)
Prob > chi2
= 0 (assumed)
(Std. Err. adjusted for 5 clusters in ID)
logGNI
Coef.
lohEXP
logUNP
HY
WIN
SUM
_cons
.2884059
-.4671568
.6412316
-.4632444
0
1.381396
sigma_u
sigma_e
rho
0
.87840652
0
-.0817278
-.0683149
-.1022321
.1870456
.
.0907301
.
.
b = consistent under Ho and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under Ho; obtained from xtreg
Test:
Ho:
difference in coefficients not systematic
chi2(4) = (b-B)'[(V_b-V_B)^(-1)](b-B)
=
-22.15
chi2<0 ==> model fitted on these
data fails to meet the asymptotic
assumptions of the Hausman test;
see suest for a generalized test
Fixed-effects (within) regression
Group variable: ID
Number of obs
Number of groups
R-sq:
Obs per group:
within = 0.2251
between = 0.5108
overall = 0.3280
corr(u_i, Xb)
logGDPpp
Coef.
lohEXP
logUNP
HY
WIN
SUM
_cons
.2311264
-.6556626
.5170393
-.3823953
0
1.596732
sigma_u
sigma_e
rho
.34251619
.60282651
.24404651
Robust
Std. Err.
.1064356
.1497153
.1692795
.1441249
(omitted)
.29563
z
P>|z|
[95% Conf. Interval]
2.71
-3.12
3.79
-3.21
0.007
0.002
0.000
0.001
.079796
-.7605935
.3094499
-.7457239
.4970157
-.1737202
.9730133
-.1807648
4.67
0.000
.8019722
1.960821
(fraction of variance due to u_i)
=
=
193
5
min =
avg =
max =
35
38.6
43
=
=
13.36
0.0000
F(4,184)
Prob > F
= 0.0501
Std. Err.
.0527676
.1291228
.3081788
.4320976
(omitted)
.2420268
t
P>|t|
[95% Conf. Interval]
4.38
-5.08
1.68
-0.88
0.000
0.000
0.095
0.377
.1270192
-.9104142
-.090979
-1.234898
.3352337
-.4009111
1.125058
.4701076
6.60
0.000
1.119227
2.074236
(fraction of variance due to u_i)
F test that all u_i=0: F(4, 184) = 11.83
corr(u_i, X)
sqrt(diag(V_b-V_B))
S.E.
Fixed effects
Number of obs
Number of groups
lohEXP
logUNP
HY
WIN
SUM
_cons
.2311264
-.6556626
.5170393
-.3823953
(b-B)
Difference
(fraction of variance due to u_i)
Random-effects GLS regression
Group variable: ID
logGDPpp
lohEXP
logUNP
HY
WIN
[95% Conf. Interval]
Random effects
corr(u_i, X)
Coefficients
(b)
(B)
fe
re
Prob > F = 0.0000