Macroeconomic news and bank stock returns

Transcription

1 Macroeconomic news and bank stock returns Tilburg University Faculty of Economics and Business Administration Business administration (BE) Bachelor thesis Finance Supervisor : Cisil Sarisoy Robert Melis

2 Abstract Using monthly macroeconomic data together with their expectations of the Consumer Price Index, the nonfarm payrolls and the new home sales for a period from 1980 to 2010, this paper analyzes the impact of macroeconomic news on stock returns of the two biggest banks of America, Bank of America and JPMorgan Chase, and two smaller banks, City National and Commerce Bancshares. Both with this empirical study and a discussion of existing literature relative to this subject, I will try to provide a clear view of the relation between the macroeconomic environment and the stock market. The empirical part gives evidence for the influence of macroeconomic news of the Consumer Price Index on the stock returns of JPMorgan Chase and Commerce Bancshares. Macroeconomic news of the nonfarm payrolls and of the new home sales only seems to have a significant impact on the stock returns of JPMorgan Chase. For the other two banks there is not enough evidence to say that news of one of these macroeconomic indicators has influence on their stock returns.

3 Introduction Many financial experts have tried to explain the influence of macroeconomic news on returns on the stock market. But because of the fact that there is no clear answer to the question how asset prices will react on for example changes in the interest rate, there will always be risk on the stock market. Therefore, the macroeconomic environment is important to consider in financial opportunities. The aim of this paper is to test whether there is any significant relation between macroeconomic news and bank stock returns. According to the efficient market hypothesis (EMH) security markets are efficient in reflecting all available information into asset prices, even without delay (Fama, 1970). If this is true, one could expect that there should be a clear model for estimating asset prices. But why is it then that not all financial literature is consentient about how asset prices react to random variables like macroeconomic indicators? Of course the stock market is an important source for companies to raise money, but the stock market can also be considered as an indicator for how a country s economy is performing. Therefore it is good to understand how macroeconomic news influences stock market returns. There are already many studies related to this issue like for instance the study of Chen, Roll & Ross (1986) who tried to find an answer to the question if risks that come along with macroeconomic frictions pay off on the stock market. Other researchers, like Funke & Matsuda (2006), went further on this with the assumption that the state of the economy, as well relevant international macroeconomic news is important to consider in estimating the influence of macroeconomic news on stock returns. These studies are all to be considered in understanding the influence of the macroeconomic environment on the stock market. But due to the fact that not all studies correspond to each other, it is clear that there is still some disagreement among financial experts about how asset prices react to macroeconomic news. In general this paper will try to provide the answer to the question: Is there any significant relation between macroeconomic news and bank stock returns in the United States?

4 Literature review Funke & Matsuda (2006) are two of many financial researchers who investigated the influence of macroeconomic news on stock markets. They investigated the stock markets of the United States and those of Germany. One of their statements was that macroeconomic news that affects the global economy as a whole, is likely to be more important if you want to explain fluctuations of stock returns in one country. In their study they also made a distinction between three states of the economy: a period of weak economic growth (recession), a period of strong economic growth (expansion), and a period of normal economic growth. They do so because they assume that stock markets may react in a different way to financial news in times of an economic contraction than in times of an economic expansion. A clear example is that a higher than expected unemployment rate, which normally is bad news, can be good news for economic growth expectations. A higher than expected unemployment rate in times of an economic expansion could cause the interest rate to decline, which in turn could be good news for investors and economic growth (Boyd, Hu & Jagannathan, 2005). Boyd et al. (2005) did an empirical research to this phenomenon and they indeed found evidence to say that a higher than expected unemployment rate is good news in times of expansion, but they did not find significant results for stock reactions to unemployment rate news in times of contraction. But of course, the effect of macroeconomic news on stock markets does not only vary depending on the state of the economy. If you want to find out whether there is a relation between, for instance, the interest rate and the returns on the stock market, you should also take the inflation into account. In general, when the interest rate is high it can be assumed that people will save their money rather than spending it to benefit from the high interest rate. The other way around, investors will be more restrained to borrow money when they have to pay more interest on their loan. Therefore, the cash flows of banks will be partly determined by the interest rate. However, when the inflation is higher than the interest rate, i.e. a negative real interest rate, it is to be expected that people are more likely to spend their money rather than bring it to the bank, because in that case people can buy less with their money in the future. Therefore, if macroeconomic indicators are somehow related to each other, the effect of macroeconomic news of one indicator can also depend on news of another indicator. Nevertheless the real financial world is not that simple to describe because the financial world is much more complex. An interesting finding of Elyasiani and Mansur (1998) with respect to the interest rate was that the volatility of bank stock returns of large to medium large banks dropped after a period

5 of high volatile interest rates. One could assume that bank stock returns react heavily on changes in the interest rate, and in fact they do. But because of the fact that banks are relatively dependant of the interest rate due to the fact that the interest rate determines for a large part what people and companies will do with their money, it may be clear that banks want to protect themselves from uncertain interest rate movements. So Elyasiani and Mansur explain their finding by the fact that banks buy derivatives when the volatility of the interest rate is high, causing the volatility of the bank stock returns to drop. The fact that large banks invest money in derivatives to protect themselves for a volatile interest rate, is enough to assume that banks financial performances are dependant of macroeconomic variables like the interest rate. Like mentioned before, Funke & Matsuda (2006) emphasized that not only the domestic financial situation is important to determine how stock prices will react, but also macroeconomic news from abroad. Nikkinen, Omran, Sahlström & Äijo (2006) went further on this. In general, they investigated what influence macroeconomic news of the United States has on global financial markets. They assumed that because of the fact that the economy of the United States is one of the leading economies in the world, macroeconomic news concerning the United States does not only affect returns on the stock markets in America, but also returns of stock markets in other continents or economic regions around the world. But America is also curious about financial news of, for example, the Euro zone. A recent example is the financial crisis in the Euro zone, where in fact the whole world follows the financial news of the Euro zone about what decisions will be made to rescue Greece, or even the Euro itself. These decisions can, for example, affect the exchange rates with other currencies. Therefore, the global macroeconomic environment is also of great importance in understanding stock market reaction around the world. Furthermore, Nikkinen et al. (2006) made a distinction between stock markets in order of their size. For instance, if you compare the S&P500 with a random local stock market, it is obvious that the S&P500, which contains stocks many multinational companies, will react differently to macroeconomic news than some local stock market, which could be more influenced by the situations within its direct environment. All this information from earlier studies makes it clear that you have to take many perspectives into account in estimating the influence of macroeconomic news on the financial world.

6 The efficient market hypothesis The efficient market hypothesis is a clear model which gives evidence that stock prices certainly do reflect relevant information like macroeconomic figures. Eugene Fama (1970) wrote an extensive article about this theory and he divides the empirical researches on the efficient market hypothesis into three different categories, called the strong form tests, the semi-strongform tests and the weak form tests. Without explaining the differences between these tests extensively, the main difference between these tests is the assumed incorporated information in the stock prices. The weak form test claims that publicly available information up to the recent past is incorporated into asset prices, the semi-strong test adds the assumption of immediate incorporation of everyday s market events into asset prices, and the strong form test adds the assumption that asset prices also reflect hidden information only known by a select group of investors. Of course, this theory can also be applied for macroeconomic news and stock prices, in a way that stock prices should immediately reflect macroeconomic announcements made on a specific day. In the empirical part of this paper, macroeconomic data and stock returns from the past will be analyzed. The macroeconomic announcements, which will be described in further detail later, were given at a specific time on a specific day. However, the stock returns that will be used are based on the closing prices of the stocks per day. So there will be some time spread between the macroeconomic announcements and the closing stock price data. So it could be that other events that could influence the stock prices happened between the time of the announcement and the time of the closing price estimation. Therefore this research can be best described as a weak form test of the efficient market hypothesis. Nevertheless, Fama (1970) claimed that there is enough significant evidence for this weak form test. So according to these findings, either positive or negative macroeconomic announcements should be reflected in related stock price data. However, speculators always have an expectation about what the macroeconomic figures will be at next announcement day. So if we assume that the theory of the efficient market hypothesis holds, earlier expectations about these actual announcement should also and already be incorporated into asset prices. In other words, according to the efficient market hypothesis, stock markets should not react anymore on the already expected part of the actual macroeconomic announcement. Therefore, stock prices should only react on the remaining news part of the announcement on a specific announcement day.

7 Still, although the theory of the efficient market hypothesis seems legit, it may be wise to not take this theory for granted in any case. For instance, Subramanian (2010) came with evidence that the model failed in the recent financial crisis. According to Subramanian (2010), the efficient market hypothesis was based on the 19 th century economical environment. But with the upcoming new generation of financial products like derivatives, swaps and other complex products which most of the time are not even comprehended by economists and bankers themselves, the economic environment changed in the 20 th century. One of the reasons of the recent financial crisis was that banks started to offer the so called sub-prime mortgages. Many economists claim that the housing bubble in the United States of the last years was caused by the low interest rate policy invented and defended in the beginning of the 21 st century by Federal Reserve chairman Alan Greenspan. So because of the fact that people could easily get a mortgage due to the low interest rate, nearly everyone that qualified for a mortgage had one in the years after the invention of the low interest rate policy. Therefore, banks began to offer mortgages to people who are in fact not qualified to get a mortgage. These mortgages were called the sub-prime mortgages. Of course it is clear that these mortgages were much more risky than the normal prime mortgages. This resulted in more and more people who were not able to pay back their loans anymore, whereby banks were losing money on a big scale. And together with falling house prices due to a low demand, the financial crisis had started. To make this piece complete, the assumption of the efficient market hypothesis was that asset prices fully reflect all available information that is relevant to its value. So if assets are mispriced, there will be rational players arbitraging the assets to its correct price. However, offering sub-prime mortgages described above is not very rational. According to Subramanian (2010), the efficient market hypothesis was built on the desire of rationality and certainties which simply cannot hold in a highly uncertain economic environment with intervening political influences. Still, with the findings of Fama (1970) in mind, a reflection of the macroeconomic data in the stock returns could be evidence for the efficient market hypothesis. But moreover, significant results for a relation between macroeconomic indicators and stock returns will result in a better understanding of the financial world.

8 Data description For the empirical part, I will try to find out whether the Consumer Price Index (CPI), the nonfarm payrolls and the new home sales have a significant influence on the stock returns of four American banks which I will describe later. In an earlier study, Flannery & Protopapadakis (2002) investigated the influence of these and some other different macroeconomic variables on aggregated stock returns, but they did not find significant results for every variable they used. Of course, these results could be different by investigating only bank stock returns rather than aggregated stock returns. It is to be expected that the Consumer Price Index, the nonfarm payrolls and the new home sales will influence the performance of financial firms. Consumer Price Index It is highly likely that the inflation rate will affect stock returns. As discussed before, the interest rate together with the inflation determines whether people can buy more or less with their money in the future. Nevertheless, if the inflation rate is high in one period it is likely that the purchasing power will decrease. If this is the case, companies will sell less products or services and profits will decrease, which could result in lower or negative returns on the stock markets. Of course this is also relevant for the bank sector which provides, for instance, loans to these companies. To set boundaries, I will only investigate the influence of the Consumer Price Index to bank stock returns rather than also take the Producer Price Index (PPI) into account. Because the CPI affects the economy as a whole rather than only the firms, it is to be expected that the CPI has a higher influence on the performance of the economy than the PPI has. Nonfarm Payrolls Furthermore the nonfarm payroll employment is also likely to be related to the performance of the financial sector, but in an indirect way. The unemployment rate, more or less the opposite of the nonfarm payrolls, is negatively related with the interest rate. The interest rate is likely to drop when the unemployment rate is high to keep the economy strong. With a low interest rate, firms can borrow money at lower costs to expand, and therefore it is to be expected that firms will hire more employees resulting in a lower unemployment rate (Storey, 1991). Shown the relation between the unemployment rate and the interest rate, it can be assumed that nonfarm payroll employment could influence the performance of banks, in a way that banks will provide more loans after a period of high unemployment and vice versa. Furthermore, by taking the nonfarm payroll employment into account together with stock price data in times of a healthy economy and in times of a recession, the statement of Boyd, Hu & Jaganathan (2005) mentioned earlier could be confirmed.

9 Home Sales And of course, the new home sales are also important for the financial sector. Financial institutes like banks offer mortgages to finance the purchase of a house. When home sales are expected to raise, it could also be expected that banks will offer more mortgages in the future. If this is the case, banks will earn more interest income in the future which could result in higher returns on the stock market. Although interest income also depends on the interest rate, it can be assumed that new home sales have a positive effect on bank stock returns. Macroeconomic data clarification Bloomberg and Money Market Services provide data on the macroeconomic indicators described above for the United States of America. Apart from the actual monthly macroeconomic announcements, these institutes also provide survey data of forecasts of economists about these announcements. With the theory of the efficient market hypothesis in mind, these expectations should already be incorporated into asset prices on the day of the actual macroeconomic announcement. Therefore, stock prices should only react on the news part of the actual announcements. So, the macroeconomic data I will actually use for the empirical research is the surprise part of the actual monthly announcements, which is the actual announcement minus the already expected part of this announcement. However, the difference between the actual announcements and the survey forecasts cannot optimally reflect to what extent the macroeconomic announcements contain new information. Beber and Brandt (2010), who investigated reaction of bonds and their volatility to macroeconomic news, chose to standardize the news parts of the macroeconomic announcements by the standard deviation of these news parts. In short, they calculated the standardized measure of macroeconomic news using the following equation: kt kt kt k So, subtracting the expectations of each announcement k at time t (X kt ) from each corresponding actual announcement (A kt ), and dividing each resulting surprise part by the standard deviation of all these surprise parts ( k ), will result in the standardized surprise parts of each announcement per announcement day (S kt ). According to Beber and Brandt (2010), an advantage of this methodology is that it makes it easier to compare the surprise parts of different macroeconomic variables. For instance, the Consumer Price Index is given in a

10 percentage change from previous period, while the new home sales are given in absolute numbers. With standardized measures, these differences among different variables will be eliminated. Therefore, in the empirical part of this study I will also use the standardized news parts of the macroeconomic announcements. Furthermore, for the nonfarm payrolls and the new home sales it is obvious that if the actual figures are higher than the expected ones, this can be defined as a positive shock for the economy. However, a higher than expected inflation (CPI) can be defined as a negative shock for the economy. But since deflation is assumed to be bad (Bordo et al., 2004), a higher than expected deflation, where in fact the actual value turns out to be lower than the expected one, will also be bad news. Therefore, a higher than expected inflation, and a higher than expected deflation will be indicated as a negative shock for the economy in the data analysis. Eventually, Bloomberg provided data for the period of 1997 till 2010 for all of these macroeconomic indicators, and Money Market Services provided this data for the time before. However, the time range of the data received from Money Market Services varies per indicator. After merging the macroeconomic data, this paper uses data of the CPI from 1980 till 2010, data of the nonfarm payrolls from 1985 till 2010, and data of the new home sales from 1988 till 2010.

11 Stock price data The financial data which will be analyzed in this paper is stock price data of four American banks listed in the top fifty of largest banks of the United States. SNL Financial is a provider of financial data and composed a list with mer ica s largest banks based on total assets and deposits data as of December 31, This list* shows major differences between the banks on top of the list and those at the bottom in terms of total assets and deposits. The four largest banks alone hold more than 65 percent of the accumulated total assets of these fifty banks together. It can be expected that a bank that holds a lot of assets is more dependent of macroeconomic variables than, for instance, a local bank. To see if this is true, I will use stock price data of the two largest banks of America, which are the Bank of America Corporation and JPMorgan Chase & Co., and two banks at the bottom of the list, which will be City National Corporation and Commerce Bancshares Incorporation. All of these banks are commercial banks. One may notice that there are other banks ranked between City National and Commerce Bancshares, respectively ranked 46 and 50 on the list. But, due to the fact that the banks in between could not provide financial data for the time range I will use, I chose to include the bank City National in my empirical research. Still, to show the major differences between banks on top of the list and those at the bottom, the Bank of America and JPMorgan Chase together hold almost 38 percent of the accumulated total assets of all fifty banks, while City National and Commerce Bancshares together only hold 0,34 percent of the accumulated assets on the list. The Center for Research in Security Prices (CRSP) provides data for the stock prices together with their daily returns without dividends of these four American banks for a period of 31 years, from the year 1980 up to and including This quite large sample period contains multiple contraction and expansion periods, of course up to the recent financial crisis which affected stock markets worldwide drastically, and where American banks like Lehman Brothers even collapsed. Thereby it may be possible to investigate whether macroeconomic news has a different effect in different times of the business cycle, which could lead to a better understanding in the influence of macroeconomic indicators on stock prices. The National Bureau of Economic Research (NBER) provides information about the business cycle of the United States from the mid-fifties till now**. * The list with the top 50 largest banks of the United States is enclosed in the appendix (table 1) ** The data of the business cycle of the United States is enclosed in the appendix (table 2)

12 Stock price data clarification Some further information about the stock price data I will use must be mentioned briefly. First of all, the current Bank of America Corporation was established in October 1998 after NationsBank acquired BankAmerica Corporation*. After the acquisition, NationsBank went further under the name Bank of America Corporation. Therefore, for the period of January 1980 till November 1998 I will use stock price data of NationsBank, and data of the Bank of America itself for the period thereafter. Second JPMorgan Chase, which of course has a long history as the second leading bank of America. The bank is a composition of many mergers and acquisitions from the past, but relevant information for the period of 1980 till 2010 is the acquisition of Chase Manhattan by Chemical Bank in March These banks were two of the key predecessors in the formation of JPMorgan Chase. After the acquisition they took over the name of Chase Manhattan, and at the end of the year 2000, the same Chase Manhattan merged with the formerly JPMorgan & Co. to become JPMorgan Chase. In order of these events, CRSP provided the stock price data for JPMorgan Chase, adjusted at times of an acquisition or merger. I will use stock price data of Chemical Bank, which was the acquirer of Chase Manhattan, from 1980 till March For the period of April 1996 till the end of 2000 I will use the stock price data of Chase Manhattan, and data of J.P. Morgan Chase itself from the year 2001 till the end of For the banks City National and Commerce Bancshares no further information about mergers or acquisitions is required for the period of 1980 till Still, it must be mentioned that for both banks there were no closing stock prices available for the days between 1980 and Therefore the stock returns in that period were estimated on the stocks bid/ask averages for these days. * Note that there is a difference between BankAmerica and the Bank of America

13 Theoretical Framework In short, with the information mentioned above the simple theoretical framework for this paper will be as follows: CPI Nonfarm payrolls Bank stock returns Home sales State of the economy The Consumer Price Index, the nonfarm payrolls and the new home sales will be the independent variables. I assume that that they all will have a significant influence on the dependent variable, which will be the bank stock returns. As described above, I will not generalize the bank stock returns but I will make a distinction between larger and smaller banks in terms of total assets. This because macroeconomic news could have a different effect on large banks like the Bank of America, than it has on more local banks like Commerce Bancshares. Still I assume that all of these macroeconomic variables will have an influence on bank stock returns overall. Furthermore, in theory the business cycle can be divided in more than just a contraction and an expansion period. Still, I will only use data of contraction and expansion periods to describe the state of the economy of the United States of America as a moderating variable in the empirical research.

14 Estimation technique One of the most used models among financial researchers who investigated the impact of macroeconomic news on stock returns are the autoregressive conditional heteroskedasticity (ARCH) and the generalized autoregressive conditional heteroskedasticity (GARCH) model. Both these models were designed to forecast volatility, which of course is of great importance on stock markets. The main difference between the ARCH and the GARCH model is that the GARCH model tries to forecast volatility not only on the past value of the conditional variance like the ARCH model does, but also by taking the past values of the squared errors into account (Engle, 2001). By doing so, you will have a long memory model and thus be able to forecast volatility more precisely in the long run. Elyasiani and Mansur (1998), who investigated the influence of the interest rate on bank stock returns, used an even more expanded GARCH model, GARCH-M. This generalized autoregressive conditional heteroskedasticity in mean model is able to show the relation between volatility and expected returns by modelling the mean of the past returns as a function of the volatility measure. Due to the fact that the GARCH-M model incorporates a link between mean returns and volatility measures, Elyasiani and Mansur (1998) claim that the empirical GARCH-M model is close related to asset pricing theories like the capital asset pricing model (CAPM). Therefore, a GARCH model seems to be the model to use to investigate the influence of macroeconomic news on bank stock returns, when dealing with a dataset with a broad time range. The difference between the study of Elyasiani and Mansur (1998) with respect to this paper, apart from the fact that they investigate the effect of the interest rate on bank stock returns rather than the effect of macroeconomic news, is that they do not make a distinction between different states of the economy in their empirical research. When using data containing different periods of the business cycle, it could be that independent variables have a different influence on data in one period of the business cycle than in another period. The findings of Boyd et al. (2005) mentioned earlier are a good example for this. In other words, a dummy variable has to be implemented to find out whether macroeconomic news has a different effect on stock returns in different times of the business cycle.

15 The GARCH model can be described as: (1) Y t = α + x t b + ε t (2) h t = α 0 + α 1 ε 2 t-1 + ß 1 h t-1 (3) ε t ~ N(o, h t ) Equation one is the regression equation to be derived from the macroeconomic data and the bank stock returns, where Y t is the expected return at time t, and x t will be the macroeconomic news of the Consumer Price Index, the nonfarm payrolls and the new home sales at time t. The error term (ε t ) is normally distributed with a mean of zero, and the volatility measure at time t (h t ) is determined by the squared error of the previous period (ε 2 t-1) and by the volatility of the previous period (h t-1 ) So, with the regression line equation, the expected return is determined by a constant (α), plus the function of a variable at time t times a constant, plus a random error. However, like mentioned before, Funke and Matsuda (2006) took the state of the economy into account in their research. They divided the economy into three different states, and therefore they implemented a dummy variable to describe the state of the economy in their model : (1) Y t = α + ΣD i x t b + ε t i Where the term D i is the dummy variable. Despite of the fact that this study will be limited to just two different periods of the business cycle rather than three, the model described above will be the appropriate model to investigate the influence of macroeconomic news on stock returns.

16 With respect to this paper, I will describe the relation between the bank stock returns and macroeconomic news of the Consumer Price Index (CPI), the Nonfarm Payrolls (NFP) and the New Home Sales (NHS), using the following regression: BSRTRN t = α + ß Cont * (SE t ) * MEN t + ß Exp * (1-SE t ) * MEN t + ε t Where: BSRTRN t = The specific bank stock return without dividends at announcement day t given as a percentage. MEN t = The standardized macroeconomic news part at announcement day t of the CPI, NFP and NHS. SE t = The state of the economy at announcement day t, where 0 will stand for an expansion phase, and 1 for a contraction phase. ε t = N(0, 2 )

17 Summary statistics Before showing the empirical results, I will provide the summary statistics of the bank stock returns and of the macroeconomic data. Bank stock returns Like mentioned before, the daily stock returns of the Bank of America, JPMorgan Chase, City National and Commerce Bancshares from 1980 till 2010 will be analyzed. Within this time range, there are 7843 observations for the Bank of America and 7844 observations for the other three banks. Bank of America JPMorgan Chase City National Commerce Bancshares Daily returns without dividends (%) N Minimum Maximum Mean Std. Dev , ,2691 0,0588 2, , ,0967 0,0561 2, , ,2456 0,0607 2, , ,3147 0,0561 1,4180 The minimum and the maximum daily return for each bank given above are exceptionally low or high for the specific time range. In fact, with a mean of around 0,06 percent, the average daily return for each bank is quite low. Therefore, with the macroeconomic data which will be described next, I hope to find a relation between macroeconomic news and exceptional returns at times of an macroeconomic announcement.

18 Macroeconomic data Consumer Price Index For the Consumer Price Index, 372 announcements were made in the time between 1980 and Due to the fact that the CPI is given as a percentage, the difference between the actual CPI and the expected CPI is given as a percentage point difference for each announcement day. The descriptive statistics of the standardized differences are given apart because this will be the data that actually will be used in the regression analysis. Descriptive Statistics N Minimum Maximum Mean Std. Dev. Expected (%) 372-1,2 1,4 0, , Actual (%) 372-1,7 1,4 0, , Difference (% point) 372-0,6 0,5-0, , Standardized Difference 372-4,0365 3,3638-0,0497 0,9999 Announcement days Contraction 54 Expansion 318 Total of the CPI announcements between 1980 and 2010 were made in times of an economic contraction, and 318 announcements were made in times of an economic expansion. The descriptive statistics of the CPI in times of an economic contraction, as well in times of an economic expansion are given separately below. Descriptive statistics (Contraction versus Expansion) Expected (%) Actual (%) Difference (% point) St. Difference Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Contraction 0,3593 0,4624 0,3796 0,5635 0,0204 0,2167 0,1370 1,4575 Expansion 0,2825 0,1956 0,2704 0,2485-0,0118 0,1336-0,0814 0,8990

19 Nonfarm Payrolls Like mentioned before, Bloomberg and Money Market Service could not provide macroeconomic data of the nonfarm payrolls back to Therefore, for the empirical analysis I will use data of the nonfarm payrolls from 1985 to Within this period, 312 announcements of the nonfarm payrolls were made. In contrast to the CPI, changes in the nonfarm payroll employment are given in thousands. N Descriptive Statistics Minimum Maximum Mean Std. Dev. Expected (1000s) Actual (1000s) Difference (1000s) Standardized Difference 312-0,0279 0,0347-0,0007 0,0092 Announcement days Contraction 33 Expansion 279 Total 312 Descriptive statistics (Contraction versus Expansion) Expected (1000s) Actual (1000s) Difference (1000s) St. Difference Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Contraction ,0053 0,0073 Expansion ,0002 0,0093

20 New Home Sales Like the data of the nonfarm payrolls, Bloomberg and Money Market Service could not provide macroeconomic data of the nonfarm payrolls back to Therefore I will use macroeconomic data of the new home sales from 1988 to Within this period, 273 announcements of the new home sales were made. N Descriptive Statistics Minimum Maximum Mean Std. Dev. Expected (1000s) Actual (1000s) Difference (1000s) Standardized Difference 273-2,7149 4,1541 0, Announcement days Contraction 33 Expansion 240 Total 273 Descriptive statistics (Contraction versus Expansion) Expected (1000s) Actual (1000s) Difference (1000s) St. Difference Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Contraction ,0449 0,6112 Expansion ,1270 1,0414

21 Empirical Results Before discussing the empirical results, I will provide the statistical results from the regression analysis in the summary table below. Statistical Results Bank of America JPMorgan Chase City National Commerce Bancshares CPI α 0,723 * 1,160*** 0,536 0,500*** ß Cont 0,022-0,114-0,044-0,130* ß exp - 0,717* - 1,297*** - 0,433-0,427** R 2 0,008 0,032 0,004 0,018 Nonfarm Payrolls α 0,608 0,535 0,508 0,310 ß Cont - 28,504* -52,324*** - 14,767-10,995 ß exp - 0,469-0,458-0,401-0,120 R 2 0,015 0,043 0,009 0,006 New Home Sales α 0,549 1,076*** 0,489 0,435* ß Cont - 0,034-0,063 0,167-0,027 ß exp - 0,473-1,027*** - 0,433-0,425* R 2 0,005 0,028 0,008 0,012 * Significant at a 90% confidence level ** Significant at a 95% confidence level *** Significant at a 99% confidence level

22 Consumer Price Index The unexpected news part of the CPI seems to have a significant effect on the returns of JPMorgan Chase and on the returns of Commerce Bancshares. The model of the regression between the news part of the CPI and the returns of JPMorgan Chase is significant at a confidence level of 99 percent, while the model of the regression between the news part of the CPI and the returns of Commerce Bancshares is significant at a confidence level of 95 percent. For the other two banks the models are not significant and thus useless. Unfortunately, these models have a very low explanatory power. The model regarding JPMorgan Chase has a R square of 0.032, and the model regarding Commerce Bancshares has a R square of only 0,018. Thereby, the models explain respectively only 3,2 percent and 1,8 percent of the variation. by the model. Furthermore, for the model regarding JPMorgan Chase, ß Cont is not given significant. Therefore, the slope of the regression line of in times of an economic contraction must not be taken for granted. Still, the hypothesis that a higher than expected inflation will have a negative impact on bank stock returns can be accepted for JPMorgan Chase and Commerce Bancshares. An interesting fact is that since ß exp is significant and more negative than ß Cont for both banks, the CPI has a more negative effect on the returns of these banks when the economy is in an expansion. Nonfarm Payrolls Also the nonfarm payrolls seems to have a significant effect only on the returns of JPMorgan Chase. The model is significant at a confidence level of 99 percent. Nevertheless, although the R square of this model is some bigger than in the model regarding to the CPI and JPMorgan Chase, the explanatory power is still not very high. Interesting fact for this model is that a higher than expected value for the nonfarm payrolls seems to lead to lower returns for this bank. Unfortunately, since the value of ß exp is insignificant, the assumption of Boyd et al. that a higher than expected unemployment rate can be good news in times of economic expansion cannot be approved. On the other hand, since ß Cont is negative and also significant, there seems to be a negative relation between nonfarm payrolls news and the stock returns for JPMorgan Chase. The models of the other three banks regarding to the nonfarm payrolls are not significant.

23 New Home Sales Again, with a 95 percent confidence level, the macroeconomic news of the new home sales only has a significant effect on the stock returns of JPMorgan Chase. Also in this model, more than expected new home sales seem to have a negative impact on the stock returns of JPMorgan Chase, and even a higher negative impact in times of an economic expansion. But since ß Cont is insignificant, there is not enough evidence that more new home sales in times of an economic contraction definitely will have a negative impact on the stock returns of JPMorgan Chase. And unfortunately, also this model has a very low explanatory power of only 2,8 percent. Conclusion The most interesting result of the empirical research is that the macroeconomic news parts of the Consumer Price Index, the Nonfarm Payrolls and the New Home Sales all have a significant influence on the stock returns of JPMorgan Chase. Unfortunately, due to the fact that these variables did not have a significant effect on the bank of America there is too little evidence to say that bigger banks are more dependant of macroeconomic news than smaller banks. A recommendation for further research therefore should be to investigate the bank stock returns of more different big banks in contrast to more different smaller banks. Furthermore, the low explanatory power found in each of the significant models can be explained by the fact that bank stock returns are relatively volatile due to the fact that daily bank stock returns are dependent of much more factors than just one macroeconomic announcement alone. In this study, I used macroeconomic announcements which often were published in the morning on a specific announcement day, while for the stock price data I used the closing prices for the same announcement days. Therefore, there is a possibility that other events relative to the bank stock returns had taken place in the time between the macroeconomic announcement and the closing prices on these days. So, to explain bank stock fluctuations more precisely, it may be better to take other factors relative to the financial sector into account rather than only macroeconomic indicators. Nevertheless, due to the fact that I did found significant results for all the macroeconomic indicators on bank stock returns, and due to the assumed hypotheses underpinned with theory of earlier researches, I still assume that the Consumer Price Index, the nonfarm payrolls and the new home sales do influence bank stock returns. Further investigation should incorporate aggregate returns of the whole financial sector to prove this assumption even more.

24 Appendix Table 1 (Source: SNL Finance)

25 Table 2 (Source: The National Bureau of Economic Research) BUSINESS CYCLE REFERENCE DATES DURATION IN MONTHS Peak Trough Contraction Expansion Cycle Quarterly dates Peak Previous Trough Peak are in parentheses to trough from from Trough to Previous Previous this peak Trough Peak June 1857(II) October 1860(III) April 1865(I) June 1869(II) October 1873(III) March 1882(I) March 1887(II) July 1890(III) January 1893(I) December 1895(IV) June 1899(III) September 1902(IV) May 1907(II) January 1910(I) January 1913(I) August 1918(III) January 1920(I) May 1923(II) October 1926(III) August 1929(III) May 1937(II) February 1945(I) November 1948(IV) July 1953(II) August 1957(III) April 1960(II) December 1969(IV) November 1973(IV) January 1980(I) July 1981(III) July 1990(III) March 2001(I) December 2007 (IV) December 1854 (IV) December 1858 (IV) June 1861 (III) December 1867 (I) December 1870 (IV) March 1879 (I) May 1885 (II) April 1888 (I) May 1891 (II) June 1894 (II) June 1897 (II) December 1900 (IV) August 1904 (III) June 1908 (II) January 1912 (IV) December 1914 (IV) March 1919 (I) July 1921 (III) July 1924 (III) November 1927 (IV) March 1933 (I) June 1938 (II) October 1945 (IV) October 1949 (IV) May 1954 (II) April 1958 (II) February 1961 (I) November 1970 (IV) March 1975 (I) July 1980 (III) November 1982 (IV) March 1991(I) November 2001 (IV) June 2009 (II)

26 References Aarti Kanjani (2011) Top 50 US banks and thrifts by assets, Beber, A.; Brandt, M.W. (2010) When It Cannot Get Better or Worse: The Asymmetric Impact of Good and Bad News on Bond Returns in Expansions and Recessions, Review of finance, 14-1, Belgacem, A. (2009) Fundamentals, Macroeconomic Announcements and Asset Prices, EconomiX Working Papers, Bordo, M. D.; Lane, J. L.; Redish, A. (2004) Good versus Bad Deflation: Lessons from the Gold Standard Era, NBER Working Papers: Boyd, J. H., Hu, J., Jagannathan, R. (2005), The Stock Market's Reaction to Unemployment News: Why Bad News Is Usually Good for Stocks, The journal of finance 60-2, Chen, N.; Roll, R.; Ross,S. A. (1986) Economic forces and the stock market, Journal of business, 59-3, Dokko, J.; Doyle, B. M.; Kiley, M. T.; Kim, J.; Sherlund, S.; Sim, J.; Van Den Heuvel, S. (2011) Monetary policy and the global housing bubble, Economic policy : a European forum, 26-66, Elyasiani, E.; Mansur, I. (1998) Sensitivity of the bank stock returns distribution to changes in the level and volatility of interest rate: A GARCH-M model, Journal of Banking & Finance, 22-5, Engle, R. (2001) GARCH 101: The Use of ARCH/GARCH Models in Applied Econometrics, Journal of Economic Perspectives, 15-4, Fama, E. (1970) Efficient Capital Markets: A Review of Theory and Empirical Work, The journal of Finance 25-2, Flannery, M. J.; Protopapadakis, A. A. (2002) Macroeconomic Factors Do Influence Aggregate Stock Returns, The review of financial studies, 15-3, Funke, N.; Matsuda, A. (2006) Macroeconomic News and Stock Returns in the United States and Germany, German Economic Review, 7-5, Malkiel, B. G. (2003) The Efficient Market Hypothesis and Its Critics, Journal of Economic Perspectives, 17-1, Nikkinen, J.; Omran, M.; Sahlström, P.; Äijö, J. (2006) Global stock market reactions to scheduled U.S. macroeconomic news announcements, Global Finance Journal, 17-1, Shanken, I.; Weinstein, M.I. (2006) Economic forces and the stock market revisited, Journal of empirical finance, 13-2, Storey, D.J. (1991) The Birth of New Firms - Does Unemployment Matter? A Review of the Evidence, Small business economics : an international journal, 3-3, Subramanian, K (2010) Efficient Market Hypothesis: The Model That Failed, Economic and political weekly : a Sameeksha Trust publication, 45-31,