1
Government shutdowns in the United States: Effects on stock returns and bond markets
Student number: S2399547 Author: Sander Muurman
Study: Msc Finance Rijksuniversiteit Groningen
Date: 09-01-2015
Abstract
2 1. Introduction
The past years of the US political landscape have been troubled by debt ceilings,
sequestrations and most striking, government shutdowns. Large parts of the government were closed and employees were send home, services and other government funded or affiliated organisations grounded to a halt. Millions within the US were affected directly or indirectly. What caused this government shutdown to happen and why does it keeps occurring
throughout recent history? What are its consequences on the US stocks and bond markets? In my research I want to clarify some of these questions. The research conducted by Fisher and Peters (2010) provide us with some answers to these questions. The political structure of the U.S. enables the shutdowns to happen and thus more shutdowns are likely in the future. In addition, during 2013 there were a lot of messages of stock markets hitting all-time highs. Does this make sense with a government which fails to agree upon a budget? Does such a major event even have an impact on stocks? Thus, this study has two objectives. First, does a government shutdown have a significant effect on stock returns? And second, what effect does it have on bonds and is the movement observed in line with the expected movement with stocks?
3 2. Literature review
First of all, what exactly are government shutdowns and why do they appear? In the past few years the term often occurred in the news. Most often the effects of a government shutdown are discussed, but the reason of occurrence often is not. The paper written by Meyers (1997) may elaborate on these questions, as he indicates that government shutdowns are a
consequence of national voting patterns, which in the last decades divided control of the democratic entities between two large parties. And since there still are two major political parties in the US politics, it is therefore reasonable to assume more shutdowns are likely to follow in the future. As of 5 November 2014, the Republicans control both the Senate and the House of Representatives. This leaves the Democrats, who still control the White House, impaired to issue new legislation unless the Republicans agree. The President of the US can fall back on executive action, meaning that he passes legislation without the approval of Congress. On the other hand the President can veto legislation on which he does not agree. Both executive action and vetoes are likely to happen more often as long as political power remains fractured. What consequences do more shutdowns have for shareholders of US stocks or the holders of US bonds? In his paper Meyers addresses some consequences for parties that are more likely to be affected, like government employees and the military. But to what extend is the hold and delay of large sums of government spending reflected in the overall economy? And what does this mean for the implied risk of the traded US debt treasuries?
2.1 Affected companies
When thinking about companies that will likely be affected by a government shutdown, likely companies are those who deal a lot with the government. These can be military contractors, certain utility companies or construction. Fisher and Peters (2010) have conducted research on government spending shocks and focused on the returns of military contractors. They imply that after a positive shock returns are, with a delay, higher. However, they also state that these outcomes should not be used to identify consequences on spending shocks in other sectors, like education, healthcare or infrastructure. This fact is important for the research I want to conduct, since I want to check returns in the broader economy.
But how broad should a selection of the broader economy be? It must not be too broad since there are companies who hardly deal with the government and are thus probably (almost) not affected. The paper written by Bowen, Castanias and Lane (1983) might provide the
4 investigates the effects of the Three Miles Island nuclear disaster on stock returns of energy companies. They do this by selecting subclasses of energy companies based on the degree of nuclear power in their energy mix. This principle is applicable on my research too, since I can create subclasses of companies that have a large exposure to government spending or where the government is one of the greatest contractors. An important thing to learn from their paper is that they control for confounding events. In order to make sure the found price shifts are in majority attributable to the nuclear accident, they scan for potentially other impacting events and filter these out of the events that may cause conflicting observations. They do this by creating control groups. This might be applicable to my research too, since there can be various events that have a large impact on stock prices.
2.2. Bond Returns
Next to the returns for stocks, I want to investigate if there are abnormal returns in bonds. It is possible that the bond returns exhibit abnormal returns due to the government shutdown. Since it is often said stocks and bonds move in opposite direction, we would expect this to be true even in the event of a government shutdown. Baele, Bekaert and Inghelbrecht (2010) have conducted research to the comovement of stock and bond returns. They find that macroeconomic fundamentals provide little explanatory power in explaining stock and bond correlations. Liquidity proxies however, play a more important role in explaining the stock-bond correlation. So when we follow these conclusions it would be possible that government shutdown do provide correlation.
5 On the question what happens with the government bonds, Liu, Shao and Yeager (2009) provide some answers. They have done research at whether government shutdowns raise the default risk premium on US Treasury bonds. They have tested this for several occasions, the shutdown of 1995-1996 being the first. Their findings are very surprising, namely that in this first year there is an additional risk premium on US Treasury bonds. When the shutdown was resolved, the risk premium disappeared. However, after the shutdown was solved it did not reoccur, not even in new shutdowns. This suggests that investors, at least at the bond market, are indifferent in their view towards government shutdowns. Liu, Shau and Yeager give two possible explanations for this rather odd fact. The first one is that apparently investors expect that the political differences that caused the shutdown are solved soon and no serious harm will come from the shutdown. Another explanation for this lack of default risk premium could be that there actually is a default risk premium, but this premium is so small that the models used in the paper are not able to detect it.
Another paper which investigates the movement between stocks and bonds is that of Yang, Zhau and Wang (2009). They have done research to the correlation between stock and bond markets. They found that for large cap stocks the correlation with bonds also depends on monetary policy and economic conditions. During restrictive monetary policy (or economic expansion) the correlations between stocks and bonds is lower than in periods of expansive monetary policy (or economic recessions). This may be interesting since during the crisis of 2008 central banks have reacted with drastic cuts in interest rates and monetary expansion. Based on the findings of Yang, Zhau and Wang we can expect a higher correlation between the movement in stocks and bonds during especially the past few years.
6 retail sales. A government shutdown has the potential to affect all these economic indicators, thus having the power to alter the probability of cojumping. This can indicate altered
expectations of investors, which seems highly likely in the event of a government shutdown.
2.3. Event study
Event study returns can be calculated by different approaches. It starts with the simple event study as performed by Brown and Warner (1980). They performed three different methods to calculate the returns of an event. These include the constant mean return model and the market model. The market model later has been improved by MacKinlay (1997). He states that the market model provides a potential improvement over the constant mean return model. The advantage is in the fact that the market model takes away the portion of return that that is related to the variance of the market. This property can lead to an increased ability to detect the effects of an event.
This leads me to set up hypotheses which I want to test in my research:
H0: A US government shutdown has no significant effect on the selected S&P500 company returns and bond prices.
H1: A US government shutdown has a significant effect on the selected S&P500 company returns and bond prices.
7 3. Methodology
For this event study I use data of selected S&P 500 company returns surrounding government shutdowns and near government shutdowns. This allows me to determine the effect of an actual government shutdown as specific as possible. Off course, this method is not flawless since there are always underlying motives and expectations on the stock market, but this is the most pure form to determine the stock reactions on an actual government shutdown. To select which companies of the S&P 500 I can use best, I use the economy Input-Output tables from the U.S. Bureau of Economic Analysis (BEA). The tables indicate which industries are doing a lot of business with the US government. I select the ten industries with the largest
percentages, since if there is a significant effect caused by the shutdowns, these industries are most likely to show signs. Then I select the companies which fit these industries from the S&P 500. I use the Datastream database a my main source for the S&P500 since this accurately reflects the current components along with a company’s SIC and NAISC codes. This makes it easy to select companies in combination with the Input-Output tables.
The reason why I choose for the returns of the S&P500 is that it is the broadest index, also with regards to industries. Also, the data is widely available. As an estimation period, I use the daily returns over a period of 500 days. This might seem much, but I have to keep in mind that a government shutdown is not coming out of the blue and in many occasions the market knows what is going on quite a while before the actual event. With an estimation period of 500 days I expect to account for the fact shutdowns are ‘foreseen’ and I am confident early expectations drop out in the 500-day average.
The normal procedure for such a study would be to follow the simple Brown and Warner (1980) procedure of creating an estimation period:
𝐸𝑟 = ∑ 𝑅𝑠 𝑛
𝑛
𝑖=250
( 1 )
Equation (1) is used for all the estimation periods, both actual and near shutdowns. Here, 𝐸𝑟 stands for the estimated return, 𝑅𝑠 for the return of the stock and n for the number of daily returns. The next step is to compare the estimation average with the event window average to compute the abnormal returns. This method is given in formula 2.
8 Since the events have different duration and thus different durations, I have to consider the test to use well. Kolari & Pynnonen (2011) has shown in their paper that an event study with an event window of 21 days gives inaccurate results, so I should shorten some of the events I use in order to make the test results more reliable. The statistical test for the event period will have an event window of 3 days. The motivation for the choice of 3 days is that so far the US government has never defaulted on their debt and the government shutdown always is a political game, so it is likely that they eventually solve their differences. On top of that, in the past shutdowns hardly ever lasted any longer than a couple of days, so I expect the biggest abnormal returns in the first 3 days after the shutdown. To be more specific, the event window consists of the day before the actual shutdown and the first and second day of the actual shutdown. With this most expectations of investors are captured in my opinion. The day before the actual shutdown is included since the deadline for a government shutdown is known to investors and thus they anticipate during the day on the outcome of negotiations between political parties.
I use one of the methods formulated by Brown and Warner (1980) if the data is distributed normal. To find out how the data is distributed, I use the Jarque-Bera test. If the data is not normally distributed, it is best to use the nonparametric rank test as formulated by Cowan (1992). 𝑍𝑟 = 𝑑1 2 𝐾𝑑 − 𝜇𝑅 ∑𝑛 (𝐾𝑡 − 𝜇𝑅)2/𝑛12 𝑡=1 ( 3 )
In equation 3, d stands for the number of days in the event period, n for the total observations, Kd for the average rank across stocks, 𝜇𝑅 stands for the mean rank and Kt stands for the average rank across stocks on the event day.
9 𝛽̂𝑖 = ∑𝑇𝜏=𝑇1 0+1(𝑅𝑖𝜏− 𝜇̂𝑖)(𝑅𝑚𝜏− 𝜇̂𝑚) ∑𝑇𝜏=𝑇1 0+1(𝑅𝑚𝜏 − 𝜇̂𝑚)2 ( 4 ) 𝛼̂𝑖 = 𝜇̂𝑖 − 𝛽̂𝑖𝜇̂𝑚 ( 5 ) Where 𝜇̂𝑖 = 1 𝐿1 ∑ 𝑅𝑖𝜏 𝑇1 𝜏=𝑇0+1 ( 6 ) and 𝜇̂𝑚= 1 𝐿1 ∑ 𝑅𝑚𝜏 𝑇1 𝜏=𝑇0+1 ( 7 )
These equations (4,5,6 and 7) provide the tools needed to estimate the necessary parameters. Here L stands for the length of the estimation period. With these parameters the abnormal return within the market model can be calculated.
𝐴𝑅𝑖𝜏 = 𝑅𝑖𝜏− 𝛼̂𝑖− 𝛽̂𝑖𝑅𝑚𝜏 ( 8 )
With equation 8 it is possible to calculate all the abnormal returns for the stocks. The cumulative abnormal return is then calculated by equation 9.
𝐶𝐴𝑅𝑖(𝜏1, 𝜏2) = ∑ 𝐴𝑅𝑖𝜏 𝑇2
𝜏=𝜏1
( 9 )
3.1. Bond markets
Also, I want to see whether there are abnormal returns in the bond market. It is interesting to see if, like one would expect, a flight out of stocks would give rise to bond prices and vice versa, or if both assets show the same pattern in the event of a government shutdown. For the bond returns, I use the same methodology as with the S&P500 index, I use a 500 day
10 power of the price. For the calculation of bond returns, I rely on the method as described by Bessembinder, Kahle, Maxwell and Xu (2009). They use two methods, one including accrued interest and one without accrued interest for those who are not able to get details about the accrued interest. I will be using the clean version (without accrued interest), as I am not able to get details about the accrued interest. The bond returns will be calculated by using equation 10.
𝑅𝑏 =𝑃𝑡− (𝑃𝑡−1) 𝑃𝑡−1
( 10 )
Equation 10 leaves out the accrued interest part since this is often hard to collect. After collecting the returns of the bonds it is important to calculate the abnormal returns. Once again, I use the method as described by Bessembinder, Kahle, Maxwell and Xu (2009), in particular the mean-adjusted method. First they calculate the premium holding period return (PBR), here given in equation 11.
𝑃𝐵𝑅𝑖 = 𝐵𝑅𝑖 − 𝑇𝑅𝑖 ( 11 )
The PBR is calculated by taking the bond return (BR) and subtracting the return for a matched treasury (TR). The next thing to calculate is the mean expected excess returns (EBR), for which the equation is presented in equation 12.
𝐸𝐵𝑅𝑖 = ( ∑ 𝑃𝐵𝑅𝑖,𝑡 −𝑦 𝑡=−1 )1 𝑦 ( 12 )
Where y stand for the estimation period. The abnormal return for bond i are then calculated by performing the calculations in equation 13.
𝐴𝐵𝑅𝑖 = 𝑃𝐵𝑅𝑖 − 𝐸𝐵𝑅𝑖 ( 13 )
Now there are abnormal returns available to use in an analysis. These returns will be used in the same way as with the stocks. If the data is normally distributed, then a t-test will be used. If the case of non-normality a rank test will be used.
12 4. Data
For the events where there actually is a government shutdown, I use two events:
The government shutdown of 1995-1996 under President Clinton, which lasted from December 19 till January 6th, a total of 21 days.
The government shutdown of 2013 under president Obama, which lasted from September 30 till October 17, a total of 16 days.
The reason I choose these two shutdowns is because these two are the most recent and also the longest shutdowns, which gives an opportunity to research stock market effects in the longer run. However, since both of these events-windows are too long for a traditional event study, I will select a shorter window via a cumulative average abnormal return graph.
For the events where there is the threat of a government shutdown, but which does not actually happens, I also use two events:
At the 8th of April of 2011, there was the threat of a government shutdown. With only 11 hours left until the debt ceiling was reached and automatic budget cuts kicked in, a compromise was reached and the shutdown was averted by raising the debt ceiling. The debt ceiling raise of late 2004. By October 14, the debt limit was about reached at
$25 million left. The Treasury took measures to delay hitting the debt ceiling, but expected it could only keep this up until half of November. Finally, at 16 November the debt ceiling was finally raised and the US government was again able to fulfil its obligations.
The reason why I choose these two specific events is because they are important. Especially the event in 2011 was very alarming and provided an insight in politics. Republicans
13 These two events mark the data I require. For both stocks and bonds, I need at least 500 returns prior to the actual shutdown or the data it was averted, plus the returns in the
shutdown events itself. In addition, it seems wise to collect some more returns after the event to have the possibility open to do a post-event test.
For all the four periods I select, it is important to make sure as little external events as possible influence the stock and bond returns. For the screening of possible economic or political events I use the website www.infoplease.com/yearbyyear.html. The reason for choosing this website is that although maybe not entirely complete, it provides a world overview as well as a US overview. This means that I will be able to judge both national as worldwide events which might affect stock and bond returns. Another reason, maybe even more important, is that all data on the website is gathered from trusted sources, which makes it reliable.
For the shutdown of 1995/1996 it shows no other significant effects that would be able to influence stock and bond markets. For the near-shutdown in 2004 there are some factors to consider, such as the military presence in Iraq and maybe in a lesser extent the Tsunami in Asia. Specific US events are the re-election of G.W. Bush as president and the south-east of the US is hit by two hurricanes. The year 2011 is a relatively calm year with no significant high impact effects other than the ongoing threat of a possible government default. For convenience, I leave out the fact that world growth outlook is bad since this has been then case in the years before and should thus be considered common knowledge in the financial markets. Finally, 2013 is a year in which happens a lot; President Obama is sworn in for a second term, the bombing at the Boston Marathon, the threat of going to war with President Assad of Syria. However, the question is whether the events are significant enough to affect financial markets substantially and individually.
For the selection of the firms I first qualified several sectors by using the Input-Output tables from the BEA. The outcomes are shown in table 1. I used data for 2011, since 1995 and 2013 were not available in the database. In my opinion 2011 provides a good insight in the sectors which deal a lot with the federal government, since it accounts for new technological
14 After the selection of firms is made, I can collect the data necessary. There is however an important thing to keep in mind, which is that there can be interferences of other events other than (the threat of) a government shutdown. These events can be political, social or economic by nature and identifying them can be hard. The way I intent to tackle this problem is by scanning the returns data for relatively abnormal outcomes. Then I use the LexisNexis database to see whether there have happened certain events which may explain the abnormal outcome. Should this be the case, then I will attribute a dummy variable to these days in the regression analysis to compensate for these events. This approach certainly isn’t completely right, as minor events slip through. But I’m confident it will deal with all the larger events and correct for this.
15 5. Results
This paragraph is about the results of my research and the outcomes of the statistical tests performed. First the mean return method will be discussed, followed by the market return model, the largest Input Output sector and finally the bond returns.
Mean Return method
Firstly, the outcomes of the performed mean return method are discussed. As mentioned in the methodology, this method follows a setup which uses the methodology as described by Brown and Warner (1980). The estimation periods are 500 days and are performed for each of the four events, both for the use and make selection. Each event has a three day event
window, due to the fact that the day before the actual shutdown markets are very aware of the threat and the limited timeframe still available to come to a solution.
Table 1: Descriptive statistics Mean Return event study
Use Table SD 1995 NSD 2004 NSD 2011 SD 2013 Companies 159 189 205 211 Mean return 0,082% 0,085% 0,097% 0,082% Minimum return -39,448% -56,554% -25,333% -37,569% Maximum return 45,454% 41,502% 58,810% 45,517% Stdev 0,020548482 0,018169398 0,01938139 0,015088357 Skewness 0,411501119 0,341087163 0,264550605 0,673495666 Kurtosis 20,92060622 26,55621986 13,44175056 29,85228535 Jarque Bera statistic 38132,48281 102939,6818 9726,785568 170237,6714
16 Make Table SD 1995 NSD 2004 NSD 2011 SD 2013 Companies 34 44 50 51 Mean return 0,085% 0,099% 0,102% 0,077% Minimum return -34,553% -27,234% -27,625% -29,008% Maximum return 54,480% 23,539% 62,725% 14,109% Stdev 0,017419936 0,015758289 0,023115353 0,014618309 Skewness 1,071531309 0,175522668 1,248459072 -0,27767724 Kurtosis 56,58467985 16,63192199 27,80609108 11,00688149 Jarque Bera statistic 217972,9289 4644,445144 31813,47528 1091,465454
Chi-Squared 0 0 0 9,8014E-238
As can be concluded from the summary statistics, none of the event studies follows a normal distribution. This means that a nonparametric rank test will have to be used to calculate if the events are significant. The rank test as performed by Cowan (1992) provides us with the following results.
Table 2: Outcomes Mean Return Rank Test
Use Table SD 1995 NSD 2004 NSD 2011 SD 2013 Z value 0,001440859 0,000670017 0,000542535 0,001381252 Significance 0,500574819 0,500267298 0,50021644 0,50055104 Make Table SD 1995 NSD 2004 NSD 2011 SD 2013 Z value 0,001835655 0,001674624 0,001517908 0,00306737 Significance 0,50073232 0,500668078 0,500605557 0,501223702
17 Are these results in line with the actual abnormal return data? The best way to check this is via a cumulative return graph.
Graph 1: Cumulative Abnormal Average Returns Use Group
As can be concluded from this graph, for the use group it appears to be correct. There is no significant returns drop visible around day 501, the centre of the event period. The decline seems to be limited to roughly 1-2%, which can hardly be called significant on the stock markets.
18 The graph for the Make group provides the same conclusions. Around the event date there is no significant drop in stock returns visible. All the CAAR’s seem to be rising even within the event date, with the only noteworthy exception the 2011 line which obviously includes the worst of the 2008 financial crisis. It appears that for the selected stocks (the threat of) a government shutdown is not something which is seen as determining over other economic factors.
Market Return method
Secondly, I performed a test based on the market model by MacKinlay (1997).In his paper MacKinlay suggests that it is possible this method has some advantages over the mean return method. Since the mean return method showed there is almost no effect on stock returns in the event of a government shutdown, it is worth it to do a market return method. Perhaps this method shows some effect or provides new insights.
Firstly, calculations for the Beta and Alpha are required in order the get the market abnormal return. I did these calculations on the returns of the last 400 days available, and then used the outcomes in each event period. The summary statistics of the market returns are presented in table 3.
Table 3: Summary statistics market returns
Use Table SD 1995 NSD 2004 NSD 2011 SD 2013 Companies 159 189 205 211 Mean return -0,040% 0,027% 0,008% -0,003% Minimum return -38,930% -53,270% -23,340% -36,945% Maximum return 43,078% 46,137% 56,245% 44,709% Stdev 0,022254323 0,016150025 0,014599659 0,01257417 Skewness 0,292670508 0,441227951 0,584121999 1,000264657 Kurtosis 15,37212197 39,46923499 32,23156897 56,0073523 Jarque Bera statistic 12548,65058 381977,2475 213364,7774 1309456,932
19 Make Table SD 1995 NSD 2004 NSD 2011 SD 2013 Companies 34 44 50 51 Mean return 0,014% 0,052% 0,023% 0,005% Minimum return -35,491% -25,879% -28,116% -29,029% Maximum return 53,965% 21,410% 63,080% 14,462% Stdev 0,016611926 0,013906414 0,017398681 0,011465368 Skewness 1,150735546 0,183481698 2,895918711 -0,592520603 Kurtosis 64,76847756 23,21547872 79,77947303 25,29387618 Jarque Bera statistic 333870,5617 15146,09249 943031,7824 23548,90551
Chi-Squared 0 0 0 0
As can be concluded from the summary statistics, none of the event periods is normally distributed. So for all the events ranks will have to be used. The outcomes of this rank test are presented in table 4.
Table 4: Test statistics Market Return Model
Use Table SD 1995 NSD 2004 NSD 2011 SD 2013 Z value 0,001559231 0,000432584 0,000384239 0,000926384 Significance 0,500622043 0,500172576 0,500153289 0,500369574 Make Table SD 1995 NSD 2004 NSD 2011 SD 2013 Z value 0,003323218 0,003422703 0,001349203 0,000811941 Significance 0,50132577 0,501365458 0,500538254 0,500323918
20 Graph 3: Cumulative Abnormal Returns Market Model Use
As can be seen, the cumulative abnormal return is much less extreme then it is in the mean return model. Also, now there is a downtrend visible in the graph. This is not visible in graph 1. This is due to the fact that the market model depicts the abnormal returns as a under- or over performance in relation to the market return. Apparently the selection of companies in 1995 performed well, given the increasing trend in the overall market, but not as good as the market itself. The less deviating returns can also be observed in the graph for the make group. Here there are overall positive returns, indicating that the selected stocks performed better than the market as a whole.
21 Despite the less extreme CAR’s in the graphs, it appears that the market model is not able to detect more than the mean return model. Both the statistical values are nowhere near being significant and are very close. It is therefore safe to say that investors do not view a
government shutdown as a threat for company revenues and thus future dividends. Between the use and make groups there also is very little difference. Perhaps the expectation that a solution will be found before companies come into more serious problems rather than some inconvenience or later payment of invoices is strongly present in the market.
Largest share in Input-Output
A final attempt to investigate if there is any effect resulting from a government shutdown is to check to largest sectors from the Input-Output tables. For the Use table this is NAICS code 6, the Educational services, health care and social assistance. For the Make table this is the government, but since this is impossible to detect since the government has no shares enlisted on stock markets, I choose the second group, which is NAICS code 2 Utilities. Appendix E shows the selected firms resulting from these parameters. The returns of the selected firms are calculated on a market returns basis, of which the summary statistics are presented in table 5.
Table 5: Summary statistics largest Input-Output firms
22 Use Table SD 1995 NSD 2004 NSD 2011 SD 2013 Companies 27 28 28 28 Mean return -0,030% 0,110% 0,067% 0,052% Minimum return -12,995% -28,921% -13,184% -7,532% Maximum return 13,470% 26,542% 18,180% 17,420% Stdev 0,010905408 0,0142542 0,012833128 0,00973022 Skewness 0,086234747 0,6690937 0,038521878 0,37110294 Kurtosis 8,84022965 32,493514 6,655907874 8,44668184 Jarque Bera statistic 224,1336911 29933,443 57,01449376 189,156335 Chi-Squared 2,1379E-49 0 4,16351E-13 8,4183E-42
As can be derived from the summary statistics, also this return data is not normally
distributed. So once more I use ranks in order to obtain a test statistic. The results of the rank tests are presented in table 6.
Table 6: Test statistics Market Return largest Input-Output
Use Table SD 1995 NSD 2004 NSD 2011 SD 2013 Z value 0,02244 0,001663 0,001809 0,012047 Significance 0,508951 0,500664 0,500722 0,504806 Make Table SD 1995 NSD 2004 NSD 2011 SD 2013 Z value 0,007391 0,007173 0,00775 0,002354 Significance 0,502948 0,502862 0,503092 0,500939
It appears that even for the companies that have the most trade contact with the government a substantial stock return is not noticeable in the returns. And once again the differences
between the use and make group are very small, indicating investors do not actively select on the trade relation with the government in the case a government shutdown happens.
23 Bond Returns
If we hold the view that bonds often depict movements in returns opposite to stocks, we do not have to expect much as the outcomes for the stocks are of minor movement. In that view, it would be surprising if the bond returns do provide to be significant, even more so since the literature states risk premia do not increase significant. The summary statistics of the bond returns are given in the next table.
Table 7: Summary statistics bond returns
SD 1995 NSD 2004 NSD 2011 SD 2013 Mean return 0,041% -0,085% -1,709% -2,748% Minimum return -5,016% -11,635% -399,105% -400,000% Maximum return 4,528% 8,662% 100,582% 100,017% Stdev 0,010907894 0,019111787 0,296379713 0,302714276 Skewness -0,604847799 -0,685739977 -4,602060282 -5,220964357 Kurtosis 2,630996726 5,626272973 49,01765641 55,50825962 Jarque Bera statistic 0,058879946 0,833132308 4063,868407 6036,686099 Chi-Squared 0,970989162 0,659306895 0 0
The outcomes of the tests for normality indicate that the events of 1995 and 2004 are
normally distributed and the events of 2011 and 2013 are not. So for the normally distributed events I simply use a two-tailed t-test. For the nonparametric events I will use the rank procedure as described in the methodology section for stocks. The outcomes are presented in the following table.
Table 8: Outcomes bond returns
SD 1995 NSD 2004 NSD 2011 SD 2013 2-T-Statistic 0,999712 0,999396
Z value 2,77735 0,437651
Significance 0,99726 0,66918
24 rejecting the null-hypothesis. This can be because the shutdown of 2013 was quite
long-lasting and therefore more impacting than the earlier shutdowns. An important thing to be noted with this respect is the fact that the events of 2011 and 2013 have some extreme abnormal returns, especially when compared with 1995 and 2004. This is due to the
calculations for abnormal returns. As described in the methodology section abnormal returns are calculated by subtracting the return of a matched security from the return of the bonds. Since the interest rate of the 3-month Treasury Bill has dropped severely during the economic crisis of the past years, the interest rate approaches zero. This means that even a 0,01% change in interest rate provides significant percentage changes, and therefore huge abnormal returns.
This fact has also affected the post event study I intended to do, since the cumulative
abnormal returns within 100 days after the start of the event could go as far as -800%, making it very hard to say anything sense full about it. Hence I first made a graph of the bond returns from the day before the first event date till 101 days after the event period. The outcomes of this are much less extreme and therefore more usable. Table 9 provides the t-statistics for both actual shutdowns, which are both not significant, indicating any effects already wore off.
Graph 5: Cumulative Bond Returns post-event
Table 9: T-test post-event bond returns
Shutdown 1995 2013 T-test 0,992647749 0,997555063 -6,00% -4,00% -2,00% 0,00% 2,00% 4,00% 6,00% 1 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 10 3
25 6. Conclusions
This paragraph sets out the findings of my research and summarizes the results of the previous paragraph.
The consulted literature in the literature already presented some ideas to which direction the outcomes of my research could go. For example Liu, Shao and Yeager (2009) already found in their study that markets do not exhibit abnormal returns in bond markets in the case of a government shutdown. On the other hand Fisher and Peters (2010) found that there can be significant abnormal returns after a government spending shock, which a government
shutdown obviously can be. So only looking at the literature is not unambiguously an answer to the main hypothesis of this paper. I hope my research proves to be an addition to existing literature by answering the main hypothesis, namely does a US government shutdown cause abnormal returns in stock and bond markets? In short, this question can be answered with no. Both the mean return method and the market return method find returns in the event window which are not significant at any level. Also the most likely candidate to exhibit significant abnormal returns, the selection based on the largest share in the Input-Output tables of the Bureau of Economic Analysis, exhibits almost the same returns as the first two methods based on the top 10 sectors. Apparently a government shutdown does not cause any significant returns in the stock market. Also interesting is the fact that there is almost no difference in outcomes between the use and make group. It would be logical to expect some difference here, as the make group ‘simply’ gets paid after the shutdown ends. Obviously this can be problematic for some firms in financial distress, but for most firms it is limited to some inconvenience. For firms that are dependent on some government services the inconvenience can be much greater, affecting their daily operations.
If the stock market is not affected, then perhaps the bond market is a more likely candidate to show signs of investor concerns. Or is it as mentioned in the literature by Liu, Shao and Yeager (2009) and do bond returns not exhibit abnormal returns. The results of my research are that in the bond market there is no significant effect on returns to explain a possible increased concern from bond investors.
So in none of the events het null-hypothesis can be rejected and we can thus assume a
26 significantly sell or buy stocks or bonds. Perhaps a possible explanation to this seemingly indifferent attitude might be the fact, as stated by Meyers (1997), that government shutdowns are the result of voting patterns. Investors know there is a real possibility that government shutdowns occur and probably continue to occur in the future. They also know that until now, every government shutdown is solved because neither of the political parties involved can afford to be blamed for causing real harm to the economy due to not willing to resolve the shutdown. Therefore, this reoccurring pattern might be the reason investors are no longer impressed by a government shutdown. The same belief must be held in the bond market, since government bonds are directly related to the government. As recent history tells us that
investors can be very well aware of increased default possibilities, for example in the case of Greece, Italy and Spain, the same response does not happen in the case of the US. So bond holders do not believe their investment is at serious risk and thus do not sell of bonds.
27 7. References
Austin, D. A., Levit, M. R. 2013. The debt limit: History and recent increases. Congressional
Research Service.
Beale, L., Bekaert, G., Inghelbrecht, K. 2010. The determinants of stock and bond return comovements. The Review of Financial Studies, 23(6): 2374-2428.
Bessembinder, H., Kahle, K.M., Maxwell, W.F., Xu, D. 2009. Measuring abnormal bond performance. The Review of Financial Studies, 22(10): 4219-4258.
Bowen, R.M., Castanias, R.P., Daley, L.A. 1983. Intra-industry effects of the accident at Three Miles Island. The Journal of Financial and Quantitative Analysis, 18(1): 87-111.
Brown, S.J., Warner, J.B. 1980. Measuring security price performance. Journal of Financial
Econonomics, 8: 205-258.
Connoly, R., Stivers, C, Sun, L. 2005. Stock market uncertainty and the stock-bond return relation. Journal of Financial and Quantitative Analysis, 40(1): 161-194.
Cowan, A.R., 1992. Nonparametric event study tests. Review of Qualitative Finance &
Accounting, 2(4): 343-358.
Dungey, M., Hvozdyk, L. 2012. Cojumping: Evidence from the US Treasury bond and futures markets. Journal of Banking and Finance, 36(5): 1563-1575.
Fisher, J.D.M., Peters, R 2010. Using stock returns to identify government spending shocks.
The Economic Journal, 120(544): 414-436.
Kolari J.W., Pynnonen S. 2011. Nonparametric rank tests for event studies. Journal of
Empirical Studies, 18(5): 953-971.
28 MacKinlay, A.C. 1997. Event studies in economics and finance. Journal of Economic
Literature, 35(1): 13-39.
Meyers, R.T. 1997. Late appropriations of government shutdowns: Frequency, causes, consequences and remedies. Public Budgeting and Finance, 17(3): 25-38.
29 8. Appendixes
Appendix A. Selection of Purchasors’ sectors
1997 2004 2011 2013 In millions Use Input Sector Input Percentage Use Input Sector Input Percentage Use Input Sector Input Percentage Use Input Sector Input Percentage Agriculture, forestry, fishing,
and hunting 1369 211833 0,646% 1318 247356 0,533% 3072 354600 0,866% 2663 389890 0,683% Mining 7139 195816 3,646% 17569 384610 4,568% 19545 718448 2,720% 18612 698563 2,664% Utilities 16817 166586 10,095% 28940 224413 12,896% 29742 252897 11,761% 29092 240199 12,112% Construction 30813 94387 32,645% 44157 165799 26,633% 66059 229156 28,827% 67644 241677 27,989% Manufacturing 155289 2363855 6,569% 243802 2752227 8,858% 366526 3528086 10,389% 383791 3711463 10,341% Wholesale trade 20613 359641 5,732% 30216 457719 6,601% 42201 585068 7,213% 46384 648278 7,155% Retail trade 311 76302 0,408% 495 107181 0,462% 568 115564 0,492% 596 128053 0,465%
Transportation and warehousing 30008 311168 9,644% 45643 438282 10,414% 65436 596770 10,965% 65677 644227 10,195%
Information 43386 298068 14,556% 64675 416968 15,511% 79282 495985 15,985% 76332 543298 14,050%
Finance, insurance, real estate,
rental, and leasing 44248 1024880 4,317% 84888 1809808 4,690% 128031 1968364 6,504% 137787 2218887 6,210%
Professional and business services 97193 1159852 8,380% 197535 1806357 10,936% 289137 2449316 11,805% 268699 2684783 10,008% Educational services, health care,
and social assistance 12807 37760 33,917% 13435 41554 32,331% 25382 66206 38,338% 32838 72189 45,489%
Arts, entertainment, recreation,
accommodation 9949 134689 7,387% 18980 202078 9,392% 31275 247105 12,657% 34370 281067 12,228%
Other services, except government 26941 167083 16,124% 27080 177751 15,235% 26396 187517 14,077% 27677 211354 13,095%
Government 8349 62111 13,442% 8680 76648 11,324% 9044 85585 10,567% 9392 86189 10,897%
30 Appendix B. Selection of Manufactors’ sectors
1997 2004 2011 2013 In millions Make Input Sector Input Percentage Make Input Sector Input Percentage Make Input Sector Input Percentage Make Input Sector Input Percentage Agriculture, forestry, fishing,
and hunting 1893 259112 0,73% 2821 308784 0,91% 4016 437162 0,92% 4105 481026 0,85% Mining 202 163093 0,12% 303 273703 0,11% 417 510213 0,08% 452 580284 0,08% Utilities 62800 314628 19,96% 78555 420158 18,70% 110528 506813 21,81% 114930 496708 23,14% Construction 9636 763039 1,26% 14774 1220055 1,21% 17765 1077792 1,65% 16287 1206919 1,35% Manufacturing 3132 3721772 0,08% 4679 4200824 0,11% 6547 5432507 0,12% 9316 5787100 0,16% Wholesale trade 0 731029 0,00% 0 1003640 0,00% 0 1327333 0,00% 0 1458137 0,00% Retail trade 2710 767650 0,35% 4059 1073281 0,38% 6110 1254267 0,49% 6028 1392835 0,43%
Transportation and warehousing 12905 511087 2,53% 18073 703816 2,57% 24874 949919 2,62% 27023 1024517 2,64%
Information 1592 610871 0,26% 2321 897460 0,26% 3393 1113232 0,30% 3221 1223897 0,26%
Finance, insurance, real estate,
rental, and leasing 23261 2453544 0,95% 34668 3945103 0,88% 53953 4684471 1,15% 55357 5133698 1,08%
Professional and business services 57852 1646429 3,51% 88014 2539651 3,47% 114937 3397640 3,38% 119583 3691586 3,24% Educational services, health care,
and social assistance 131682 1036331 12,71% 196391 1631396 12,04% 273197 2351742 11,62% 290809 2564883 11,34%
Arts, entertainment, recreation,
accommodation 25461 578042 4,40% 46787 861999 5,43% 62577 1104021 5,67% 66073 1218544 5,42%
Other services, except government 1789 461335 0,39% 2502 612134 0,41% 3781 691215 0,55% 4324 740624 0,58%
Government 1327803 1328544 99,94% 1995717 1996809 99,95% 2682670 2684423 99,93% 2704281 2706124 99,93%
31 Appendix C. Selected Use companies
3M COACH HOSPIRA PERRIGO
AT&T COCA COLA HUMANA PFIZER
ABBVIE COGNIZANT TECH.SLTN.'A' INGERSOLL-RAND PHILLIPS 66
ABBOTT LABORATORIES COLGATE-PALM. INTEGRYS ENERGY GROUP PITNEY-BOWES
ACCENTURE CLASS A COMCAST 'A' INTEL PRAXAIR
ACTAVIS COMPUTER SCIS. INTERPUBLIC GROUP PROCTER & GAMBLE
ADOBE SYSTEMS CONSOLIDATED EDISON INTL.FLAVORS & FRAG. PUB.SER.ENTER.GP. AGILENT TECHS. CONSTELLATION BRANDS 'A' INTERNATIONAL PAPER QUALCOMM AIR PRDS.& CHEMS. CORNING JOHNSON & JOHNSON QUEST DIAGNOSTICS AKAMAI TECHS. CROWN CASTLE INTL. JOHNSON CONTROLS RALPH LAUREN CL.A
ALCOA CUMMINS KANSAS CITY SOUTHERN RED HAT
ALEXION PHARMS. DTE ENERGY KELLOGG REGENERON PHARMS.
ALLEGHENY TECHS. DARDEN RESTAURANTS KIMBERLY-CLARK REYNOLDS AMERICAN
ALLEGION DEERE KINDER MORGAN ROBERT HALF INTL.
ALLERGAN DELPHI AUTOMOTIVE LABORATORY CORP.OF AM. HDG. ROCKWELL COLLINS
ALLIANCE DATA SYSTEMS DELTA AIR LINES LAM RESEARCH SCANA
ALTRIA GROUP DISCOVERY COMMS.'A' LINEAR TECHNOLOGY SEAGATE TECH.
AMER.ELEC.PWR. DOMINION RESOURCES LOCKHEED MARTIN SEMPRA EN.
AMERICAN EXPRESS DOVER LYONDELLBASELL INDS.CL.A SIGMA ALDRICH
ANALOG DEVICES DOW CHEMICAL MALLINCKRODT SOUTHERN
APPLE DR PEPPER SNAPPLE GROUP MARRIOTT INTL.'A' SOUTHWEST AIRLINES
APPLIED MATS. DUKE ENERGY MASCO SPECTRA ENERGY
ARCHER-DANLS.-MIDL. E I DU PONT DE NEMOURS MCCORMICK & COMPANY NV. STARBUCKS
AUTODESK EASTMAN CHEMICAL MCDONALDS STARWOOD H&R.WORLDWIDE
AUTOMATIC DATA PROC. EATON MCGRAW HILL FINANCIAL STERICYCLE
AVAGO TECHNOLOGIES ECOLAB MEAD JOHNSON NUTRITION STRYKER
AVERY DENNISON EDISON INTL. MEADWESTVACO SYMANTEC
AVON PRODUCTS EDWARDS LIFESCIENCES MEDTRONIC TE CONNECTIVITY
C R BARD ELECTRONIC ARTS MERCK & COMPANY TERADATA
BAXTER INTL. EXELON MICROCHIP TECH. TEXAS INSTRUMENTS
BECTON DICKINSON EXPEDIA MICROSOFT TEXTRON
BEMIS EXXON MOBIL MOLSON COORS BREWING 'B' THERMO FISHER SCIENTIFIC
BERKSHIRE HATHAWAY 'B' FMC MONSTER BEVERAGE TYCO INTERNATIONAL
BIOGEN IDEC FEDEX MOODY'S TYSON FOODS 'A'
H&R BLOCK FIDELITY NAT.INFO.SVS. MOSAIC UNION PACIFIC
BORGWARNER FIFTH THIRD BANCORP MOTOROLA SOLUTIONS UNITED PARCEL SER.'B'
BOSTON SCIENTIFIC FIRST SOLAR MURPHY OIL UNITED TECHNOLOGIES
BRISTOL MYERS SQUIBB FIRSTENERGY MYLAN UNIVERSAL HEALTH SVS.'B'
BROWN-FORMAN 'B' FORD MOTOR NIKE 'B' VARIAN MEDICAL SYSTEMS
CA FOSSIL GROUP NETAPP VERISIGN
32
CONSOL EN. GANNETT NORFOLK SOUTHERN WHIRLPOOL
CSX GARMIN NORTHROP GRUMMAN WINDSTREAM HOLDINGS
CAMERON INTERNATIONAL GENERAL DYNAMICS ONEOK WYNN RESORTS
CAMPBELL SOUP GENERAL ELECTRIC PACCAR XCEL ENERGY
CAPITAL ONE FINL. GENERAL MILLS PG&E XEROX
CATERPILLAR GENERAL MOTORS PPG INDUSTRIES XILINX
CENTERPOINT EN. GILEAD SCIENCES PPL XYLEM
CHEVRON GOODYEAR TIRE & RUB. PARKER-HANNIFIN YAHOO
CHIPOTLE MEXN.GRILL HERSHEY PAYCHEX YUM! BRANDS
CINTAS HESS PEPCO HOLDINGS ZIMMER HOLDINGS
CISCO SYSTEMS HEWLETT-PACKARD PEPSICO ZOETIS
CLOROX HONEYWELL INTL. PERKINELMER EBAY
Appendix D. Selected Make companies
ACE INVESCO SALESFORCE.COM
AFLAC KINDER MORGAN SANDISK
AMER.ELEC.PWR. LABORATORY CORP.OF AM. HDG. SCHLUMBERGER AMERICAN INTL.GP. LINCOLN NATIONAL CHARLES SCHWAB
AMERIPRISE FINL. METLIFE SCRIPPS NETWORKS INTACT. 'A'
BB&T MONSANTO SEMPRA EN.
BANK OF AMERICA MOODY'S SIMON PROPERTY GROUP
CITIGROUP NISOURCE STARBUCKS
CROWN CASTLE INTL. PNC FINL.SVS.GP. STARWOOD H&R.WORLDWIDE
DELTA AIR LINES PPL TECO ENERGY
DUN & BRADSTREET DEL. PINNACLE WEST CAP. ADT
EQUITY RESD.TST.PROPS. SHBI PIONEER NTRL.RES. UNITEDHEALTH GROUP
EXELON PLUM CREEK TIMBER WESTERN UNION
EXPEDITOR INTL.OF WASH. PUBLIC STORAGE WISCONSIN ENERGY
FLUOR ROBERT HALF INTL. XL GROUP
GENERAL GW.PROPS. SCANA YUM! BRANDS
33
Appendix E. Selected companies largest values Input-Output
Companies Use Companies Make
DAVITA HEALTHCARE PTNS. AES NEXTERA ENERGY
HUMANA AMEREN NRG ENERGY
LABORATORY CORP.OF AM. HDG. AMER.ELEC.PWR. NORTHEAST UTILITIES QUEST DIAGNOSTICS CENTERPOINT EN. PEPCO HOLDINGS TENET HEALTHCARE CMS ENERGY PINNACLE WEST CAP. CVS HEALTH CONSOLIDATED EDISON PG&E
DOMINION RESOURCES PPL
DTE ENERGY PUB.SER.ENTER.GP. DUKE ENERGY SCANA
EDISON INTL. SEMPRA EN.
ENTERGY SOUTHERN
EXELON TECO ENERGY
