University of Amsterdam
Graduate Thesis 2015
Monetary Policy Shocks and Stock Returns: Evidence
from Highly Distressed and Less Distressed
Companies
Ashutosh Shahi
Student Number: 10660933
Supervisor: Chris Florakis
Abstract
In this thesis, I use an event-study based empirical analysis to study the effect of Bank of England, Monetary Policy Committee’s interest rate decisions on the returns of UK companies sorted based on their probability of financial distress. I document strong evidence of relationship between monetary policy shocks of Bank of England and its differential effect on the returns of companies sorted based on their level of financial distress. Specifically, the effect of surprise interest rate changes which are extracted from the actual rate changes on MPC meeting dates relative to interest rate changes expected by the market (which are derived from a short term future contract that tracks LIBOR) are transmitted differentially to companies based on their level of financial distress – with the effect more pronounced on companies which have higher chance of default. The result has important implications for active fund managers. For example, if the manager expects a rate hike unexpected by the markets from the central bank, he should consider moving assets to companies, which have less chance of default. Similarly, if the manager expects an unexpected rate cut, he should consider moving assets to companies that have higher default risk.
Keywords: Interest Rate Shocks, Stock Returns, Default Risk, Altman Z-Score, Ohlson O-Score
TABLE OF CONTENTS
1. INTRODUCTION ... 5
2. LITERATURE REVIEW ... 7
2.1 Initial Studies ... 8
2.2 Event-‐study based research ... 8
2.3 Vector Autoregressive (VAR) analysis based studies ... 9
2.4 Recent empirical studies that this thesis builds on ... 10
3. DATA AND METHODOLOGY ... 12
3.1 Interest Rate Decisions by Bank of England MPC ... 12
3.2 Calculating Unexpected and Expected Components of Interest Rate Change ... 13
3.3 Stock Market Data ... 16
3.4 Altman Z-‐Score ... 18
3.5 Ohlson O-‐Score ... 19
3.6 Which measure is used? ... 20
3.7 Construction of Portfolios for the two measures ... 21
4. RESPONSE OF STOCK MARKET RETURNS TO INTEREST RATE CHANGES ... 23
4.1 Results ... 24
4.2 Comparison of O-‐Score vs Z-‐Score – which is superior ... 29
5. ROBUSTNESS CHECKS ... 29
5.1 Alternate Definition of Crisis – from September 2008 to March 2009 ... 29
5.2 Alternate Definition of Crisis – covering the official period of recession in UK from March 2008 to June 2009 ... 31
5.3 Alternate Description of Portfolio – Equal Sized Quartile portfolios ... 33
6. LIMITATIONS OF CURRENT RESEARCH ... 33
6.1 Size Effect ... 33
6.2 Does not consider some recent oddities regarding companies’ access to capital ... 35
7. POTENTIAL EXTENSIONS FOR FUTURE WORK ... 36
7.1 Consider all UK listed stocks ... 37
7.2 Consider alternate definitions of financial distress ... 37
7.3 Quantitative Easing and its differential effects on Stock Returns ... 37
8. CONCLUSIONS ... 38
REFERENCES ... 40
APPENDIX 1: RESULTS WITHOUT CONSIDERATION OF A STRUCTURAL BREAK OF
RETURNS VERSUS RATE CHANGE DURING CREDIT CRISIS ... 43
1. Introduction
Central Banks around the world use monetary policy as an important tool to influence outcomes such as economic growth, inflation and unemployment to maintain
monetary stability. Monetary stability implies stable prices and confidence in the currency. Stable prices are defined by the Government’s inflation target, which the central banks seek to meet through typically their interest rate target. When a central bank changes its official interest rate, it is attempting to influence the overall level of activity in the economy in order to keep the demand for, and supply of, goods and services roughly in balance. Doing so results in a rate of inflation in the economy consistent with the Bank’s inflation target.
When demand for goods and services in the economy exceed supply, inflation tends to rise above the Bank’s target. On the other hand, when supply exceeds demand, inflation tends to fall below the Bank’s target. By changing the interest rate that the central bank pays on reserve balances held by commercial banks, the bank is able to influence a range of other borrowing and lending rates, and hence spending in the economy. A reduction in interest rates makes saving less attractive and borrowing more attractive, which stimulates spending. Lower interest rates can also affect consumers’ and firms’ cash flow – a fall in interest rates reduces the income from savings and the interest payments due on loans. Borrowers tend to spend more of any extra money they have than lenders, so the net effect of lower interest rates through this cash-flow channel is to encourage higher spending in aggregate. The opposite occurs when interest rates are increased.
There are time lags before changes in interest rates affect spending and saving decisions, and longer still before they affect consumer prices. While it is difficult to be too precise about the size or timing of all these channels, the overall effect of a change in interest rates on output is estimated to take up to about one year, and the overall impact on consumer price inflation takes up to about two years [see e.g Friedman (1948); Bernanke and Gertler (1995)]. So interest rates have to be set based on judgments about what inflation might be – the outlook over the coming few years – not what it is today.1
1
Based on Bank of England’s definition of how it decides on target policy rate.
The effect of the monetary policy on inflation though is indirect, by controlling the rate of growth of money supply, and economic activity. A direct and immediate effect of this change in monetary policy is on financial markets – by affecting the asset prices. The asset prices change in anticipation of modified expectations of economic activity and output as a result of monetary policy change.
Economists have tended to focus on the question whether monetary policy has any effect on real stock prices – or the change in stock prices just incorporate the change in inflation expectations as a result of monetary policy change as described above. Many studies [see e.g. Lynge Jr (1981) ;Berkman (1978); Thorbecke (1997); Patelis (1997);Kuttner (2001); Bernanke and Kuttner (2005)] have been done to show the effect of monetary policy shocks on real stock returns. Most of these suggest that a positive monetary policy shock (e.g. a cut in interest rates) leads to increase in stock prices and similarly a negative monetary policy shock leads to decrease in stock prices. Distinction is made between a change in monetary policy that was expected by the markets versus a “surprise” or shock in the monetary policy. This is because if the market already anticipated a change, its effect would have already been priced in asset returns.
According to theory (e.g. Residual Income Model, Dividend Discount Model or Discounted Cash Flow Model), stock price of a security is equal to the present value of expected future net cash flows. Thus the evidence that expansionary monetary policy leads to higher stock prices should suggest that the policy change should affect the company positively by increasing future cash flows or by decreasing the discount rate at which those cash flows are capitalized.
So, if as per theory the monetary policy does affect the firm’s balance sheets, a monetary tightening, by increasing interest rates, can worsen the cash flow of a firm net of interest and thus the value of the firm. This decline in the value of the firm can reduce its ability to borrow and thus spend and invest. The effect of this tightening should be more severe on firms, which are already in financial distress – by affecting their access to credit, compared to firms not in financial distress.
In this thesis, I explore the differential effect of monetary policy shocks on a portfolio of stocks sorted by their likelihood of default. By examining the cross-section of default-risk based sorted portfolios, we can test whether the monetary policy shock
has larger impact on firms in financial distress compared to firms not in distress by, for example, affecting their access to credit.
Of particular importance to policymakers and investment managers would also be a test if this relationship is subject to structural breaks or if it stands across various economic environments. Recent empirical studies [see Florackis, Kontonikas, and Kostakis (2014)] have also suggested a structural break in shock-return relationship, reversing its sign, during the financial crisis of 2007-2009. During this period, a interest rate cut was perceived by market participants as a negative news and a signal of deteriorating economy, leading to “flight-to-safety” trading towards safer assets, and decrease in stock prices. I also investigate the effect of financial crisis on stock portfolios, sorted by their probability of default and see if this structural break is more pronounced for the companies in financial distress. The empirical study will be an event study conducted on the stocks part of the FTSE All-Share Index, listed on the London Stock Exchange, during the period June 1999 - April 2015, on Bank of England (BoE) Monetary Policy Committee’s (MPC) meeting days. The current study extends the recent studies done by Bernanke and Kuttner (2005) – which explains the market’s reaction to unexpected rate changes by Fed in the US stock market, and Florackis, Kontonikas, and Kostakis (2014) – which shows the effect of BoE MPC rate decisions on liquidity sorted portfolios in the UK stock market. The current study can have important implications for active fund managers in their decision to rebalance their portfolio in anticipation of rate change by central banks.
2. Literature Review
The effect of monetary policy shocks on stock returns has been well studied.
Macroeconomists have for long focused on the question whether money or monetary policy of central banks has any effect on real stock prices. In this literature review section, I first discuss the major empirical studies which looked at the effect of monetary policy on stock returns and then focus on two recent studies – Bernanke and Kuttner (2005) and Florackis, Kontonikas, and Kostakis (2014) that this thesis follows closely and extends to study the differential effect of monetary policy on companies in financial distress.
2.1 Initial Studies
Earliest studies done to study the effect of money supply predicted that money supply shocks would result in investors substituting between money and other assets to reestablish their desired money holdings. Investors will typically respond with a lag, which would imply that money could help predict stock returns. This was supported by some empirical studies [Sprinkel (1964); Keran (1971); Homa and Jaffee (1971); Brunie, Hamburger, and Kochin (1972)]. This however contradicted with the theory of efficient markets developed by Fama (1970). The findings were however disputed by subsequent research, which showed that there is no predictive content in past changes in money supply but there is a reverse causality from stock returns to money supply [Cooper (1974); Rozeff (1974); Rogalski and Vinso (1977)].
2.2 Event-‐study based research
This problem with endogeneity between monetary shocks and stock returns led researchers to consider using an event-study methodology – where they look at the effect on stock returns immediately after a monetary policy announcement. The argument was that stock prices would immediately adjust to the monetary shocks and hence the announcement could be viewed as exogenous to market reaction on the event day. Berkman (1978) and Lynge Jr (1981) found that stock prices reacted negatively to money supply announcements. Lynge did not distinguish between changes expected and unexpected by the market, while this was a major point of Berkman’s study. Berkman showed that stock prices only react to unanticipated changes. Pearce and Roley (1983) studied how stock prices react to unanticipated money supply announcements. Using weekly data from 1977 – 1982, they estimated the following model:
∆𝑃!= 𝑎 + 𝑏 ∆𝑀!!− ∆𝑀!! + ∈!
where ΔPt is the percentage change in stock price, ΔMat is the announcement
change in money stock, and ΔMe
t is the expected change in money stock. The
expected change was obtained from survey data. They obtained a negative estimate of b parameter showing unexpected monetary shocks had a negative effect on equity prices.
Jensen and Johnson (1995) considered the returns on stocks in different time periods surrounding a change in the discount rate. They investigated the effect on equity returns of a discount rate change on the pre announcement days, on the announcement day, and the post announcement days. There was a negative effect on stock returns in all three periods. The pre announcement returns indicated that the stock market anticipates the change in the discount rate. They did not interpret post announcement reaction as a delayed response, but as a sign that reinforcing events tend to follow a rate change.
Tarhan (1995) examined the impact of Federal Reserve Open Market Operations (FOMC) on financial asset prices for periods between 1979 and 1984, when the growth rate of money was the target of Fed. He found no evidence that Fed influences stock prices. Thorbecke (1997), on the other hand, found significant negative responses on the percentage change in Dow Jones Industrial Average to Fed’s actions. The focus of Thorbecke’s study was the period when the Fed targeted Fed Funds Rate as part of their monetary actions however, which might explain the difference from Tarhan’s results. Thorbecke shows using Fed Funds Rate as the measure of monetary policy, in the short run, has real and significant effect on stock returns.
2.3 Vector Autoregressive (VAR) analysis based studies
In the 1990s, many studies [e.g. Thorbecke (1997), Patelis (1997), Lastrapes (1998)] applied vector autoregressive (VAR) technique to determine the effect of monetary policy on stock returns. VAR models have become increasingly popular in recent decades in empirical econometric studies to summarize the dynamics of
macroeconomic data. They are estimated to provide empirical evidence on the response of macroeconomic variables to various exogenous impulses in order to discriminate between alternative theoretical models. A VAR can be an equation, n-variable model in which each n-variable is in turn explained by its own lagged values, plus the current and past values of the remaining n-1 variables. VAR modeling does not require as much knowledge about the forces influencing a variable, as do structural models with simultaneous equations.
Thorbecke (1997) used the following variables in a monthly VAR: the growth rate of industrial production, inflation rate, commodity price index, federal funds rate, Fed
non borrowed reserves (NBR)2, total reserves and stock returns. The federal funds
rate is the monetary variable used. He split the stocks into ten portfolios based on size, and found that monetary tightening has strongest negative impact on stocks of small firms. He explains this as due to monetary tightening affecting the small firm’s ability to borrow more than larger firms.
Patelis (1997) used the following variables in his VAR setup: excess stock returns, real interest rate, dividend yield, term spread, the growth in fed funds rate, and the Strongin indicator3. He finds that monetary policy (fed funds rate and Strongin
indicator) influence expected excess returns and expected dividend yield growth. There is a negative short-run response of expected excess return to monetary tightening. The results are thus quite similar to Thorbecke (1997), but the model is different in the choice of variables to include in VAR.
Lastrapes (1998) confirms the findings of Thorbecke and Patelis in an international setting by estimating the response of equity prices and interest rates to money supply shocks in G-7 countries. He used a VAR with a long-term interest rate, real output, real equity price index, real money balances and nominal stock of money.
2.4 Recent empirical studies that this thesis builds on
Bernanke and Kuttner (2005) document a strong and consistent response of stock market to unexpected monetary policy actions, using Federal funds futures data to gauge policy expectations. They derive the unexpected or surprise element of the Federal Funds target from the change in the futures contract’s price relative to the day prior to the policy action. The separation of expected and unexpected component of the rate change decision was done as follows: If the rate decision was taken on day d of month m, the unexpected target funds rate change can be calculated from the change in the rate implied by the current-month futures contract. But because the contract’s settlement price is based on the monthly average Federal funds rate, the change in implied rate is scaled up by a factor related to the number of days in the month affected by the change,
∆𝑖!= ! !!! (𝑓!,!
! − 𝑓
!,!!!! )
2 Non-borrowed reserves represent the numerical difference between total reserves
minus funds that have been borrowed from the Fed discount window.
where Δiu is the unexpected target rate change, f0
m, d is the current-month futures
rate, and D is the number of days in the month. The expected component of the change is defined simply as the actual change minus the unexpected change:
∆𝑖! = ∆𝑖 − ∆𝑖!
They conducted an event study on FOMC meeting days when the rate was cut, and regressed the CRSP value-weighted index return Ht against Δie and Δiu:
𝐻! = 𝑎 + 𝑏!∆𝑖!!+ 𝑏!∆𝑖!!+ 𝜀!
According to them, an unexpected 25-basis-point rate cut would typically lead to an increase in stock prices on the order of 1%. They also noticed evidence of larger market response to rate changes that are perceived more permanent, and a smaller response to unexpected inaction on part of FOMC.
Florackis, Kontonikas, and Kostakis (2014) study UK stock market and show that the effect of monetary policy shocks are transmitted in a differential manner to the cross-section of liquidity sorted portfolios. The monetary policy shock is extracted on the meeting days of Bank of England Monetary Policy Committee (MPC), relative to market expectations embedded in 3-month LIBOR future prices. More importantly, they also show that the effect of monetary policy shocks reversed its sign during the financial crisis of 2007-2009. Their empirical results show that expansionary rate surprises led to lower prices for liquid stocks during the crisis. This was because investors perceived the rate cut as bad news for future economic prospects, leading to rebalancing of portfolios away from stocks and towards safer assets such as government bonds.
To account for the structural break in relationship between the monetary policy shock, and portfolio returns, they extend the model of Bernanke and Kuttner by introducing a crisis period dummy variable. The crisis period used by them was from August 2007 to March 2009. The regression model used for their liquidity-sorted portfolios is:
𝑟!,! =
𝛼 + 𝛽!! 1 − 𝐷
!!"#$#$ ∆𝑖!!+ 𝛽!! 𝐷!!"#$#$ ∆𝑖!!+ 𝛽!! 1 − 𝐷!!"#$#$ ∆𝑖!!+ 𝛽!! 𝐷!!"#$#$ ∆𝑖!!+ 𝜀!
where Dcrisis stands for crisis-period dummy variable that takes a value of 1 from August 2007 to March 2009, and 0 otherwise.
Similar to studies done in the US by Kuttner (2001) and Bernanke and Kuttner (2005), they use interest-rate futures to extract monetary policy shocks on BoE’s MPC meeting days. However, there is no futures contract which tracks BoE’s policy rate. So short sterling futures contract that settles on 3-month British Banker’s Association (BBA) London Interbank Offer Rate (LIBOR) is used as a substitute as it is widely used to hedge against or speculate on interest rate movements. Its change in price on the MPC meeting days is considered as the monetary policy shock initiated by the central bank action or inaction.
3. Data and Methodology
Below I summarize the interest rate and stock financial data used, their construction from the data sourced from Thomson DataStream, and filtering applied, and the descriptive statistics of the datasets, before moving onto using this data in my event study analysis in the next section.
3.1 Interest Rate Decisions by Bank of England MPC
For my event study analysis, I consider the BoE meeting days from June 99 to April 2015, for a total of 192 meetings. In this period there were 14 times that the rate was increased, 20 times the rate was decreased and 158 times the rate was maintained
4. The start of the study is from June 99 because LIBOR future contracts that I use to
extract the unexpected component of interest rate change (see section 3.2 below) did not settle on a monthly basis before June 99, only quarterly delivery existed. This lack of correspondence between the frequency of MPC meeting date (monthly), and the future’s settlement may lead to biased estimates of unexpected rate change before June 1999.
Figure 1 shows the plot of interest rate decisions of BoE MPC during this period. The interest rate during this period has varied from a high of 6%, maintained from
February 2000 to January 2001, to a historical low of 0.5%, maintained from March 2009 till now. Figure 2 shows the magnitude of the interest rate changes during this period. Most of the changes have been of 25 basis points, however during the credit crisis (period shown in gray background), there were two interest rate cuts of 100 basis points and 150 basis points, followed by 3 more 50 basis point decreases till March 2009, when the rates reached a low of 0.5%.
Figure 1 Plotted from data obtained from 4 BoE MPC Rate decisions from June 1999 to April 2015.
There were in total 192 MPC meetings during this period. The interest rate varies from a high of 6% in February 2000 to a historical low of 0.5% that has been maintained from March 2009 till now. The last 6 years of this time period have been a quite extra ordinary as the BoE has kept the low rate of 0.5% for a record 73 months, the longest period of no rate change since the Second World War. This period has also been accompanied by a recent policy of central banks around the developed world to inject money directly into the economy by purchasing financial assets – a phenomenon known as
Quantitative Easing (QE). QE is an unconventional form of monetary policy where a Central Bank creates new money electronically to buy financial assets, like
government bonds. This process aims to directly increase private sector spending in the economy and return inflation to target. The BoE started with a QE of £75 billions in March 2009, and it stands at £375 billion at end of my test period (April 2015).
The total period however covers the various phases of economic cycle quite well – with the first expansionary phase ending around 2001, followed by a contraction till 2004, and subsequent recovery and period of high economic growth till 2007. This was followed by one of the worse recessions of recent times, which coincided with the credit crisis, and now the economy is in a recovery phase again.
3.2 Calculating Unexpected and Expected Components of Interest Rate
Change
Similar to the method used by Florackis, Kontonikas, and Kostakis (2014), I use the short sterling futures contract traded on LIFFE that settles on 3-month LIBOR, to extract monetary-policy shock on BoE’s MPC meeting days. The 3-month LIBOR is one of the instruments used by the BoE to access the market expectations regarding the future interest rate in its policy decision-making (Brooke, Cooper, and Scholtes (2000); Bredin et al. (2007)). 0.0%$ 1.0%$ 2.0%$ 3.0%$ 4.0%$ 5.0%$ 6.0%$ 7.0%$ May/99 $ May/00 $ May/01 $ May/02 $ May/03 $ May/04 $ May/05 $ May/06 $ May/07 $ May/08 $ May/09 $ May/10 $ May/11 $ May/12 $ May/13 $ May/14 $ May/15 $
BoE$MPC$Rate$Decision$
Crisis$200752009$ Interest$Rate$Figure2: Interest Rate Change Decisions between June-‐99 and April-‐2015. In this period there has been 192 MPC meetings, in which the 14 times the interest rate was increased, 20 times the rate was decreased and 158 times the rate was maintained
Figure 3 shows the plot of the interest rate implied from the short sterling future contract price changes during the test period. This plot is based on daily settlement prices of the 3-month LIBOR future (not limited to the MPC meeting dates). The implied interest rate is 100 minus LIFFE future contract price. As can be seen visually, the implied interest rate graph follows closely the interest rate change decisions of BoE plotted in Figure 1.
Figure 3: Plot of daily Implied Interest Rate (100 – Short Sterling Futures Contract) from June 1999 to April 2015
Figure 4 plots the corresponding implied rate changes during the same period. Again, as can be seen, the implied rate changes closely track the actual changes – the graph is however much more volatile because the prices are based on supply and demand, with other factors determining the changes on days other than the MPC
!1.60%' !1.40%' !1.20%' !1.00%' !0.80%' !0.60%' !0.40%' !0.20%' 0.00%' 0.20%' 0.40%'
Jun!99' Jun!00' Jun!01' Jun!02' Jun!03' Jun!04' Jun!05' Jun!06' Jun!07' Jun!08' Jun!09' Jun!10' Jun!11' Jun!12' Jun!13' Jun!14' Jun!15'
Interest'Rate'Change'Decisions'by'BoE'MPC'
Crisis'2007:2009' Interest'Rate'Change' !1# 0# 1# 2# 3# 4# 5# 6# 7# 8#Jun!99# Jun!00# Jun!01# Jun!02# Jun!03# Jun!04# Jun!05# Jun!06# Jun!07# Jun!08# Jun!09# Jun!10# Jun!11# Jun!12# Jun!13# Jun!14# Jun!15#
100#$#3#month#LIBOR#Future##
meeting days. More importantly, we can see that the implied rate changes become much more volatile during the credit crisis – defined in my study as the period between August 2007 and March 2009.
Figure 4: Implied rate change derived from the 3-‐month Short Sterling Future Contract Price Change form June 1999 to April 2015
As suggested by Florackis, Kontonikas, and Kostakis (2014), the unexpected interest rate change, Δiud, is calculated as the change in the implied 3-month LIBOR rate on
the MPC meeting day, d, relative to the previous day, d-1: ∆𝑖!!= 𝑓
!,!− 𝑓!,!!!
where fm,d is the implied rate from the future contract, i.e. 100 minus the LIFFE
futures contract price extracted from the contract with the delivery month m nearest to the MPC meeting day d. The source of LIBOR futures price on the meeting day, and the day before the meeting is Thomson DataStream5. The expected component
of the interest rate is defined as the actual change in interest rate (the change made by the MPC on the meeting date) minus the unexpected rate change calculated above:
∆𝑖! = ∆𝑖 − ∆𝑖!
I split the interest rate decision made by BoE on the meeting dates into unexpected and expected components using the above two formulas.
5
The name of future contract in Thomson DataStream is LIFFE-3MTH STERLING
CONTINUOUS, and the prices used are settlement price at the end of trading day.
!1.5% !1% !0.5% 0% 0.5% 1% 1.5% 1999!06!09%2000!06!09%2001!06!09%2002!06!09%2003!06!09%2004!06!09%2005!06!09%2006!06!09%2007!06!09%2008!06!09%2009!06!09%2010!06!09%2011!06!09%2012!06!09%2013!06!09%2014!06!09%
3"month(LIBOR(Future(Price(Change(
Crisis(2007"2009( LIBOR(Future(Change(Figure 5: Actual Interest Rate Change vs Unanticipated Interest Rate Change calculated from price change of sterling futures contract traded at LIFFE, which settles on 3-‐month LIBOR
Figure 5 shows the plot of actual interest rate change versus the unexpected rate change during the study period. Table 1 shows the corresponding descriptive statistics for the interest rate change, their expected and unexpected components, and the corresponding summary statistics for the 5-year and 10-year bond yield changes on the MPC meeting dates, also sourced from Thomson DataStream. The average unexpected interest rate change is close to zero, ranging from a minimum of -44 basis points, to a maximum of 27 basis points. Also, Figure 5 indicates, the interest rate changes became highly volatile during the financial crisis, but have remained almost unchanged since then – indicating that the market participant haven’t expected a rate raise yet.
Table 1
Interest Rates - Summary Statistics
Interest Rate Changes (Basis Points) Mean Median Min Max St.Dev. Interest Rate Change -‐2.47 0 -‐150 25 17.8 Unexpected rate change -‐0.58 0 -‐44 27 5.92 Expected rate change -‐1.9 0 -‐118 37 16.7 5-‐year bond yield change -‐0.64 -‐1.1 -‐17.9 16.7 5.33 10-‐year bond yield change -‐0.35 -‐0.65 -‐29.4 16.6 5.56
3.3 Stock Market Data
To look at the differential effect of monetary policy on companies in financial distress vis-à-vis companies not in financial distress, I consider the common stocks part of
!160% !140% !120% !100% !80% !60% !40% !20% 0% 20% 40%
Jun!99% Jun!00% Jun!01%Jun!02%Jun!03%Jun!04% Jun!05% Jun!06% Jun!07%Jun!08%Jun!09%Jun!10% Jun!11% Jun!12% Jun!13%Jun!14%Jun!15%
In te re st 'Ra te 'C ha ng e' (B as is 'P oi nt s) ' Crisis'200762009' Interest'Rate'Change'(Basis'Points)' Unan9cipated'Change'(Basis'Points)'
FTSE All-Share Index from June 1999 to April 2015. FTSE All-Share Index is an aggregation of FTSE 100 Index, FTSE 250 Index and FTSE Small Cap Index, and it aims to represent at least 98% of full capital values of UK companies that qualify as eligible for selection. To qualify, the companies must have a full listing on LSE, with a Sterling or Euro denominated price.
The data includes both the presently listed companies, along with the companies, which got de-listed at some point during the sample period. Thus the dataset is free from any survivorship bias – this is particularly important as we wish to differentiate the companies based on their probability of default, hence there is a high likelihood of introducing a potential bias if we exclude companies which get de-listed from the exchange after a bankruptcy or default.6
The stocks considered in the empirical analysis are subjected to some filtering to remove potential bias. As is customary in UK stock market studies, I exclude unit trusts, investment trusts and ADRs. Also, the academically popular measures of financial distress that I propose to use (Z-score and O-score) do not work well for financial companies (financials tend to be highly levered and their risks are not well disclosed in the statements ), so I limit the empirical analysis to non-financial companies only. In addition, I had to exclude the companies for which not all the financial data was available to calculate the Z-Score or O-Score. The final dataset consisted of an average of 425 companies each year for the analysis using Z-Score, and 428 companies for the analysis using O-Score.
In my study, I have split the companies selected above into 4 portfolios, based on their probability of default. To estimate the probability of default, I use the two well-known measures of estimating the bankruptcy risk – Altman’s Z-score and Ohlson’s O-score. Both the measures are widely used in the academic circle to estimate the financial health of a company, and they use multiple corporate income and balance sheet values to measure the financial health of a company. Their popularity stems from the fact that they are easy to calculate, based on publicly available data from the financial statements of the companies, and have historically given good estimate about the financial health of a company. Below I give a brief description of the two
6
In Thomson Datastream one can append the month and year code at the end of an
index list to get the historical constituent list. For example, LFTALLSH gives the list of companies, which are currently part of FTSE All Share Index, whereas
LFTALLSH0199 will give us the list of companies part of the index in January 1999 – including the ones which got de-listed at some point.
measures, before discussing how I create the 4 portfolios for running my analysis, and which measure is likely to give us a better prediction of financial distress.
3.4 Altman Z-‐Score
Altman (1968) proposed Z-score to predict the probability that a firm will go into bankruptcy within 2 years. Z-scores are popular method in academic studies to measure the financial distress of companies. Z-score is calculated as:
𝑍 = 1.2 𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠 + 1.4 𝑅𝑒𝑡𝑎𝑖𝑛𝑒𝑑 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠 + 3.3 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝐵𝑒𝑓𝑜𝑟𝑒 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑎𝑛𝑑 𝑇𝑎𝑥𝑒𝑠 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠 + 0.6 𝑀𝑎𝑟𝑘𝑒𝑡 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝐸𝑞𝑢𝑖𝑡𝑦 𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝑇𝑜𝑡𝑎𝑙 𝐿𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠+ 0.99 𝑆𝑎𝑙𝑒𝑠 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠
(𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝐶𝑎𝑝𝑖𝑡𝑎𝑙)/(𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠) measures the net liquid assets of a firm relative to the total capitalization. This ratio provides information about the short term
financial position of the business. The more working capital there is compared to the total assets, the better the liquidity situation.
(𝑅𝑒𝑡𝑎𝑖𝑛𝑒𝑑 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠)/(𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠) measures profitability that reflects the company's age and earning power. The company will use retained earnings to operate the business; it is either reinvested or used to pay off debt. The lower the ratio, the more the company is funding assets by borrowing instead of through retained earnings.
(𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝐵𝑒𝑓𝑜𝑟𝑒 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑎𝑛𝑑 𝑇𝑎𝑥𝑒𝑠)/(𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠) measures operating efficiency apart from tax and leveraging factors. It recognizes operating earnings as being important to long-term viability.
(𝑀𝑎𝑟𝑘𝑒𝑡 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝐸𝑞𝑢𝑖𝑡𝑦)/(𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝑇𝑜𝑡𝑎𝑙 𝐿𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠) adds market dimension that can show up security price fluctuation as a possible red flag. This ration is supposed to measure how much of the company’s market value could decline before liabilities exceed assets.
(𝑆𝑎𝑙𝑒𝑠)/(𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠) is a standard measure for total asset turnover. It is one of the measures of management’s capability in dealing with competitive conditions.
Empirical evidence from Altman's Study (1968) shows that if the Z-score if greater than 3, then the company is most likely safe based on financial data; if the score is between 2.7 and 3.0, then the company is probably safe from bankruptcy, but this is in the grey area and caution should be taken; when Z-Score is between 1.8 to 2.7, then the company can be bankrupt within the next 2 years, while with a Z-Score of less than 1.8, the company has the highest probability of defaulting in next 2 years.
3.5 Ohlson O-‐Score
Ohlson (1980) postulated O-score as an alternative to Altman Z-score. It is a 9-factor linear combination of coefficient-weighted business ratios readily available from accounting disclosures of public companies. The O-score is measured as:
𝑂 − 𝑠𝑐𝑜𝑟𝑒 = −1.32 − 0.407 log 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠 + 6.03 𝑡𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 − 1.43 𝑤𝑜𝑟𝑘𝑖𝑛𝑔 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 + 0.076 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑎𝑠𝑠𝑒𝑡𝑠 − 1.72 1 𝑖𝑓 𝑡𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 > 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠𝑠, 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 − 2.37 𝑛𝑒𝑡 𝑖𝑛𝑐𝑜𝑚𝑒 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠− 1.83 𝑓𝑢𝑛𝑑𝑠 𝑓𝑟𝑜𝑚 𝑜𝑝𝑒𝑟𝑡𝑖𝑜𝑛𝑠 𝑡𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 + 0.285 1 𝑖𝑓 𝑛𝑒𝑡 𝑙𝑜𝑠𝑠 𝑓𝑜𝑟 𝑙𝑎𝑠𝑡 𝑡𝑤𝑜 𝑦𝑒𝑎𝑟𝑠, 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 − 0.521 𝑛𝑒𝑡 𝑖𝑛𝑐𝑜𝑚𝑒!− 𝑛𝑒𝑡 𝑖𝑛𝑐𝑜𝑚𝑒!!! 𝑛𝑒𝑡 𝑖𝑛𝑐𝑜𝑚𝑒! + |𝑛𝑒𝑡 𝑖𝑛𝑐𝑜𝑚𝑒!!!|
The O-score breaks itself down into 9 different approximate measures of a firm’s default risk. These 9-variables are used to determine firm-size, leverage, working capital, liquidity, profitability, change in net income, and debt financing. Two of the 9 measures are dummies.
Size: log 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠 − Ohlson measures a company’s size as its total assets. Smaller companies are deemed to be more at risk of failure.
Leverage Measure: !"!#$ !"#$"!"%"&'
!"!#$ !""#$" – is designed to capture the indebtedness of a company, the more leveraged the more at risk the company is to shocks.
Working Capital Measure: !"#$%&' !"#$%"&!"!#$ !""#$" − Even if a company is endowed with assets and profitability, it must have sufficient liquidity to service short-term debt and
Inverse Current Ratio: !!""#$% !"#$"!"%"&'!"##$%& !""#$" - This is another measure of a company’s liquidity.
Discontinuity Correction for Leverage Measure:
1 𝑖𝑓 𝑡𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 > 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠𝑠, 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 − Dummy variable equaling one if total liabilities exceeds total assets, zero otherwise. Negative book value in a corporation is a very special case and hence Ohlson felt the extreme leverage position needed to be corrected through this additional measure.
Return on Assets: !"# !"#$%&
!"!#$ !""#$"− An indicator of how profitable a company is, assumed to be negative for a close to default company.
Funds to Debt Ratio: !"#$% !"#$ !"#$%&!'(!"!#$ !"#$"!"%"&' − A measure of a company’s ability to finance its debt using its operational income alone, a conservative ratio because it does not include other sources of cash. If the ratio of funds from operations to short-term debt is less than one, the company may have an immediate problem.
Discontinuity Correction for Return on Assets:
1 𝑖𝑓 𝑛𝑒𝑡 𝑙𝑜𝑠𝑠 𝑓𝑜𝑟 𝑙𝑎𝑠𝑡 𝑡𝑤𝑜 𝑦𝑒𝑎𝑟𝑠, 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 − Dummy variable equaling one if income was negative for the last two years, zero otherwise.
Change in Net Income: !"# !"#$%&!! !"# !"#$%&!!!
!"# !"#$%&!!|!"# !"#$%&!!!|− Designed to take into account any potential progressive losses over the two most recent periods in a company’s history.
Together these nine variables build an O-score where the probability of failure is : 𝑒!!!"#$%
1 + 𝑒!!!"#$%
Results greater than 0.5 indicate a firm with high chance of default.
3.6 Which measure is used?
It has been argued [see Mansi, Maxwell, and Zhang (2010); Begley, Ming, and Watts (1996); Kumar and Kumar (2012)] that Ohlson O-score is better predictor of
bankruptcy than Altman Z-score. However there maybe merits in using both the measure, and they act as a good source to compare the empirical results against two popular models. So I use both the measures in my study, with O-Score results
providing the baseline measure, and Z-Score results used as an alternative measure.
Also as mentioned before, the two measures do not work well for financial
companies. This is because financials tend to be highly levered and their operating risks and exposures are not well disclosed – so I filter them out in my regression tests.
3.7 Construction of Portfolios for the two measures
The O and Z scores are calculated for the companies using the annual financial data of the company from a year before. For example, when classifying companies in year 2000, I use the annual financial data of 1999 to calculate the two scores. Based on each year’s O and Z scores, the companies are split into 4 portfolios based on their probability of default. These 4 portfolios are then used for the 12 monthly MPC meeting dates in the regression explained in Section 4. On moving to the next year, the portfolios are then again redefined using the new annual financial data of the constituents of FTSE All-Share Index for the year.
In case of O-score, it is transformed into a probability of failure, P, using a logistic transformation where P > 0.5 indicates a company at risk, and P < 0.5 indicates a safe company.7 I categorize all the companies with a probability of default greater
than 0.5 into portfolio P1 – which contains all the companies at highest risk of financial distress, and divide all the companies with probability less than 0.5 into tercile portfolios. For Z-Score, the companies are split into 4 portfolios using the range suggested by Altman – with portfolio P1 containing all the companies with a Z less than 1.8 (with highest probability of bankruptcy), P2 containing companies with a Z in the range 1.8 to 2.7, P3 containing companies with a score of 2.7 to 3, and P4 containing companies with a Z-Score of greater than 3 – which are considered to have least probability of bankruptcy.
Table 2 presents the descriptive statistics of the 4 portfolios using the above classification. One of the observations from the statistics is that O-Score is quite conservative in labeling a firm as having high probability of default, while Z-score has labeled significantly more companies as having significant probability of default.
7
P > 0.5 threshold was suggested by Ohlson (1980) in his study. However, since
50% probability of default is already quite high, I ran my analysis (in unreported results) with a threshold of 0.4 and 0.3 – which did not give statistically different results.
Table 2
Distribution of O-probability of default based on O-scores and Z-scores for the
quartile portfolios
O-‐score distribution
Mean Median Min Max St.Dev. Count P1 -‐ Most Distressed 0.79 0.79 0.5 1 0.17 1,505 P2 0.12 0.079 0.038 0.5 0.091 23,143 P3 0.022 0.021 0.01 0.038 0.0079 23,426 P4 -‐ Least Distressed 0.0043 0.0038 0 0.01 0.003 23,686 Z-‐score distribution
Mean Median Min Max St.Dev. Count P1 -‐ Most Distressed 0.81 1.22 -‐89.1 1.8 2.81 16,268 P2 2.26 2.26 1.8 2.7 0.26 17,101 P3 2.85 2.85 2.7 3 0.086 5,396 P4 -‐ Least Distressed 6.48 4.32 3 261 10.4 32,260
Next I calculate the returns for each of the portfolios on the BoE MPC meeting dates, using the closing prices on the day of meeting, and the closing prices a day before the meeting for all the companies of the portfolio. This is based on the hypothesis that the price change of the companies on the MPC meeting date can largely be attributed to the unexpected rate change decision of the committee. The returns are calculated for both value-weighted and equal-weighted portfolios for analysis. The value-weighted portfolios should in general be more representative of the market reaction to rate changes, as they take the market capitalization of the companies in the portfolio into account, when calculating returns. However, I also use equal-weighted portfolios in my analysis, as it is quite likely that smaller companies react more to rate changes, compared to larger ones – even within the same portfolios.
Table 3 presents the summary statistics for the returns of the 4 portfolios for both O and Z scores. The returns are presented for both Equal Weighted and Value
Weighted Portfolios. For the O portfolios, the mean return is 0.37% for portfolio P1, while it is close to 0% for portfolio P4, in the value-weighted case. Also, the standard deviation of returns is much higher for P1, when compared to other portfolios. In the case of Z-Score sorted portfolios, the mean return for portfolio P1 is 0.08%, while it is 0.05% for portfolio P4. The standard deviations are also of similar magnitude for all the portfolios.
Table 3
Returns Summary Statistics for the quartile portfolios sorted based on Altman-Z and Ohlson-O scores
O-‐score sorted portfolios -‐ Returns Statistics
Mean Median Min Max St.Dev. P1 -‐ Most Distressed Equal Weighted 0.33% -‐0.01% -‐8.50% 16.00% 3.10% Value Weighted 0.37% -‐0.03% -‐12.00% 35.00% 4.30% P2 Equal Weighted 0.14% 0.07% -‐5.20% 11.00% 1.50% Value Weighted 0.06% 0.03% -‐5.40% 4.70% 1.50% P3 Equal Weighted 0.12% 0.14% -‐6.10% 4.80% 1.20% Value Weighted 0.05% 0.08% -‐6.10% 4.20% 1.20% P4 -‐ Least Distressed Equal Weighted 0.11% 0.09% -‐6.00% 4.80% 1.40% Value Weighted 0.02% 0.04% -‐5.90% 3.70% 1.50%
Z-‐score sorted portfolios -‐ Return Statistics
Mean Median Min Max St.Dev. P1 -‐ Most Distressed Equal Weighted 0.12% 0.09% -‐5.10% 4.60% 1.40% Value Weighted 0.08% 0.14% -‐7.10% 4.50% 1.40% P2 Equal Weighted 0.13% 0.14% -‐6.40% 4.60% 1.30% Value Weighted 0.01% -‐0.01% -‐6.10% 3.10% 1.30% P3 Equal Weighted 0.31% 0.16% -‐6.40% 41.00% 3.20% Value Weighted 0.07% 0.00% -‐6.60% 6.60% 1.50% P4 -‐ Least Distressed Equal Weighted 0.11% 0.10% -‐5.80% 4.10% 1.30% Value Weighted 0.05% 0.12% -‐5.90% 4.20% 1.50%
4. Response of Stock Market Returns to Interest Rate Changes
For the event-study, I begin my analysis with the relationship between returns of the default-risk based sorted portfolios versus the unexpected and expected components of the interest rate change. This was the benchmark model employed by Bernanke and Kuttner (2005) :
𝑟!,! = 𝑎 + 𝑏!∆𝑖!! + 𝑏!∆𝑖!!+ 𝜀!
where rp,d is the return of portfolio p on MPC meeting day d. In the Appendix 1, I
show the results of the above test for both O and Z score sorted portfolios. As can be seen, the results do not look either economically or statistically significant. This result is in line with the similar results obtained for Florackis, Kontonikas, and Kostakis (2014) in their study of UK stock market.
Next I use the model proposed by Florackis, Kontonikas, and Kostakis (2014) on these default risk-sorted portfolios. Their model proposes a structural break in the relationship between the stock returns and interest rate change during the most recent credit crisis. In general a positive interest rate shock (a interest rate cut) is
associated with an increase in stock returns, while a negative interest rate shock is associated with decrease in stock returns. However during the credit crisis interest rate cut, more than that expected by the markets, was viewed as a sign of worsening economy. For example, when BoE MPC cut the interest rate by whopping 150 basis points (44 basis points in this cut was unexpected component) on 6th November 2008, the FTSE All-Share Index plummeted by 5.5%. Negative monetary policy shocks would normally be surprise interest-rate increase. However during the financial crisis, an interest rate cut would be perceived as a bad news, and hence a negative shock. So, a dummy variable is introduced which takes a value of 1 during the crisis and 0 otherwise. The final regression model to be tested is:
𝑟!,! =
𝛼 + 𝛽!! 1 − 𝐷
!!"#$#$ ∆𝑖!!+ 𝛽!! 𝐷!!"#$#$ ∆𝑖!!+ 𝛽!! 1 − 𝐷!!"#$#$ ∆𝑖!!+ 𝛽!! 𝐷!!"#$#$ ∆𝑖!!+ 𝜀!
where Dcrisis is crisis-period dummy variable that takes a value of 1 from August 2007
to March 2009, and 0 otherwise and Δie
d andΔiud are expected and unexpected
component of interest rate change on BoE rate decision days, rp,d is the return of
portfolio p on MPC meeting day d.
August 2007 is generally considered as start of the financial crisis, when major questions were raised about the global economy. American Home Mortgage Association filed for Chapter 11 in August 2007. In the UK, Chancellor of the Exchequer authorized the Bank of England in September 2007 to provide liquidity support for Northern Rock, the United Kingdom’s fifth-largest mortgage lender. Similarly, the stock markets reached their lowest levels by March of 2009 and started their recovery soon after. The Bank of England also started it Quantitative Easing program in March 2009 in an attempt to boost private spending.
With the description of portfolios P1 to P4 for Altman Z-Score and Ohlson O-Score in Section 3.7, and the credit crisis defined above, I run the equation above, and test for empirical evidence to test the hypothesis that the effect of monetary policy shocks on companies in financial distress is more, when compared to companies not in financial distress.
4.1 Results
Table 4 presents the results of the regression for O-Score based sorted portfolios – both for equal-weighted and value-weighted portfolios. Table 5 presents the
corresponding results for Z-Score based sorted portfolios. For estimation, I use ordinary least squares, where t-values are calculated using Newey and West (1987) standard errors.
First, I discuss the results of O-score based sorted portfolios. In Table 4, I present the results for both equal-weighted and value-weighted portfolios. For both the
categories, outside the crisis, on expected lines we see the inverse relationship between returns and unexpected interest rate. The relationship is both economically and statistically significant. Similarly, during the crisis, the direction of relationship is reversed and a positive relationship is observed between returns and unexpected rate change. The result is again both economically and statistically significant. These results are inline with the similar studies done before [see Bernanke and Kuttner (2005) ; Florackis, Kontonikas, and Kostakis (2014)]. An explanation for the reversal of shock-return relationship during the crisis is that a decrease in interest rate signaled worsening economic prospects to the market participants – leading them to liquidate their positions and move to safer assets like treasuries and gold [see e.g. Longstaff (2002); Chordia, Sarkar, and Subrahmanyam (2005); Goyenko and Ukhov (2009)].
More interestingly, the coefficient on unexpected rate change for portfolio P1 is highest in magnitude both during and after the crisis, and decreases progressively to portfolio P4 – showing the differential effect of interest rate change on companies in financial distress vis-à-vis companies not in financial distress. In my results in Table 4, for value-weighted portfolio, an unexpected 25 basis points increase is associated with a -3.56 % return for the most distressed portfolio, while the return is only -1.75% for the least distressed portfolio. Similarly, during the crisis, a 25 basis point decrease is associated with a return of -4.94% for the most distressed portfolio, while for the least distressed portfolio the return is -3.7%.
One conclusion of this observation is that a rate increase is much more costly for the companies in financial distress by, for example, affecting their access to new capital, or by affecting their existing debt servicing. Second, increasing rate environment may not only make the debt expensive – but also decrease the revenue and income because consumers are less likely to spend when rates are rising. This effect is more likely to affect the profit margins of companies based on their level of distress. Third, the expected decrease in profitability is further like to make new debt more expensive by increasing the credit spread of the companies. This can be because of couple of
reasons – a decreasing profitability makes default more likely and may affect the credit ratings of the companies, and second, increasing rate environments are in general associated with increase in absolute credit spreads.
Table 4
Benchmark result based on O-scores. The companies of FTSE All Share Index for a particular year are split into 4 portfolios ranging from P1 to P4 – P1 having maximum
probability of default and P4 having least probability of default. P1 contains all the companies for which the O-probability of default is higher than 0.5 (a value of O-probability of default higher than 0.5 shows significant chance of default as per Ohlson score). Portfolios P2 to P4 contains companies whose O-score ranges between 0 and 0.5, with each portfolio containing same number of companies. Portfolio P1-P4 is the spread portfolio between P1 and P4. Crisis is period between Aug-2007 and Mar-2009 – the period considered active phase of the credit crisis
.
The regression model being tested is:𝑟!,!= 𝛼 + 𝛽!! 1 − 𝐷
!!"#$#$ ∆𝑖!!+ 𝛽!! 𝐷!!"#$#$ ∆𝑖!!+ 𝛽!! 1 − 𝐷!!"#$#$ ∆𝑖!!+ 𝛽!! 𝐷!!!"#"# ∆𝑖!!
+ 𝜀!
Response of O-‐score based sorted portfolio returns (equal and value weighted) to rate changes -‐ Robust Estimates
Panel A: Equal Weighted Returns
P1-‐Most Distress P2 P3 P4-‐Least Distress P1-‐P4 spread Regular Unexpected -‐15.67*** -‐12.86*** -‐9.285*** -‐9.036*** -‐6.856* (-‐3.786) (-‐3.192) (-‐4.469) (-‐4.489) (-‐1.902) Crisis Unexpected 16.74*** 15.36*** 14.32*** 16.07*** 0.666 (3.815) (6.185) (5.171) (6.487) (0.286) Regular Expected -‐2.051 -‐0.597 -‐0.423 -‐0.776 -‐1.515 (-‐0.552) (-‐0.356) (-‐0.342) (-‐0.598) (-‐0.576) Crisis Expected -‐2.296* -‐1.311* -‐0.382 -‐0.212 -‐2.071** (-‐1.945) (-‐1.736) (-‐0.549) (-‐0.356) (-‐2.454) R-‐squared 9.40% 29.50% 30.90% 29.50% 2.30%
Panel B: Value Weighted Returns Regular Unexpected -‐14.25*** -‐8.437*** -‐8.046*** -‐7.014*** -‐7.627 (-‐2.681) (-‐3.606) (-‐4.198) (-‐3.410) (-‐1.607) Crisis Unexpected 19.74*** 15.50*** 14.96*** 14.81*** 4.914 (3.386) (6.096) (6.833) (6.849) (1.254) Regular Expected -‐2.301 -‐1.372 -‐0.761 -‐0.679 -‐2.056 (-‐0.643) (-‐0.948) (-‐0.613) (-‐0.460) (-‐0.755) Crisis Expected -‐2.171 -‐1.179 0.409 0.231 -‐2.378** (-‐1.558) (-‐1.300) (0.744) (0.399) (-‐2.200) R-‐squared 5.40% 22.60% 31.50% 20.40% 1.50%
Second observation from Table 4 is that the economic and statistical significance of the differential returns across portfolio P1 to P4 is much more pronounced for value-weighted portfolios when compared to equal-value-weighted portfolios. One interpretation of this observation is that size (represented by market capitalization of the company) plays an important role in determining the effect of change in the interest rate. This result is also inline with the P1-P4 spread portfolio in Table 4, where we see that the
coefficient on unexpected return is economically significant both outside and during the crisis (though not statistically significant for crisis period).
Third, the R2 for the portfolios is quite significant (higher than 20%) for most of the portfolios leading to the conclusion that the interest rate change has significant explanatory power to explain the returns on BoE MPC meeting days. One odd observation here is the significantly less R2 for portfolio P1 compared to other portfolios. However, I attribute this to the significantly less number of companies in portfolio P1 for the O-Score based definition of portfolios. This is because O-Score is quite conservative in labeling a firm as having significant probability of default – as also shown in Table 2 and discussed in Section 3.7. This observation is confirmed by Z-Score based results in Table 5 (see below), which has comparable number of companies in each portfolio (See Table 2).
Finally, the expected component of the interest rate change is not statistically significant for most of the cases, confirming the results from almost all the studies done since Berkman (1978).
Table 5 presents the corresponding results of the event-study for the Z-Score based sorted portfolios – for both equal-weighted and value-weighted portfolios. Again, for both the categories of return, outside the crisis the inverse relationship between returns and unexpected rate change is seen, while the relationship reverses it’s sign during the crisis. The coefficient for unexpected rate change is highest in magnitude for the portfolio P1 and decreases as we go to portfolio P4. The results are again much more pronounced for the value-weighted returns. For the value-weighted case, an increase of 25 basis points in unexpected rate change outside the crisis is
associated with a return of -2.34% for portfolio P1, while it is -1.89% for the portfolio P4. During the crisis, an unexpected 25 basis point cut leads to a return of -4.43% for portfolio P1, and a returns of -3.64% for portfolio P4. For the equal-weighted portfolio, while the pattern is, in general, similar, portfolio P3 statistics look out of line. Also, the P1-P4 spread portfolio values look economically significant only for the
value-weighted case – showing that the difference in returns across the portfolios is not significant. The coefficient for spread portfolio for value-weighted case, even though significant, is much less when compared to O-Score based results. This shows that for the dataset, O-Score may be doing a better job at sorting companies based on their probability of default.
Table 5
Benchmark result based on Z-scores. The companies of FTSE All Share Index are split into 4 portfolios – P1 to P4, based on the Z-score ranges proposed by Altman. P1 represent
companies with maximum financial distress, while P4 represents companies experiencing minimum financial distress, based on Z-scores. Portfolio P1-P4 is the spread portfolio between P1 and P4. Crisis is period between Aug-2007 and Mar-2009 – the period considered active phase of the credit crisis. The regression model being tested is:
𝑟!,!= 𝛼 + 𝛽!! 1 − 𝐷
!!"#$#$ ∆𝑖!!+ 𝛽!! 𝐷!!"#$#$ ∆𝑖!!+ 𝛽!! 1 − 𝐷!!"#$#$ ∆𝑖!!+ 𝛽!! 𝐷!!"#$#$ ∆𝑖!!
+ 𝜀!
Response of Z-‐score based sorted portfolio returns (equal and value weighted) to rate changes -‐ Robust Estimates
Panel A: Equal Weighted Returns
P1-‐Most Distress P2 P3 P4-‐Least Distress P1-‐P4 spread Regular Unexpected -‐10.35*** -‐8.672*** -‐24.15 -‐9.710*** -‐0.639 (-‐4.920) (-‐4.363) (-‐1.556) (-‐4.760) (-‐0.784) Crisis Unexpected 15.96*** 14.79*** 13.21*** 15.35*** 0.605 (6.705) (5.107) (4.623) (5.967) (0.878) Regular Expected -‐0.210 -‐0.801 1.122 -‐0.911 0.701 (-‐0.149) (-‐0.594) (0.365) (-‐0.592) (1.220) Crisis Expected -‐1.067 -‐0.697 -‐0.00209 -‐0.534 -‐0.533** (-‐1.378) (-‐1.124) (-‐0.00256) (-‐0.801) (-‐2.390) R-‐squared 30.40% 28.60% 14.50% 30.20% 3.90%
Panel B: Value Weighted Returns Regular Unexpected -‐9.370*** -‐5.798*** -‐8.013*** -‐7.573*** -‐1.797 (-‐4.033) (-‐3.436) (-‐2.887) (-‐3.614) (-‐1.173) Crisis Unexpected 17.70*** 15.08*** 14.47*** 14.56*** 3.136*** (10.55) (6.160) (5.331) (6.105) (3.190) Regular Expected -‐1.432 -‐0.495 -‐1.481 -‐0.624 -‐0.808 (-‐1.160) (-‐0.437) (-‐1.198) (-‐0.410) (-‐1.114) Crisis Expected -‐0.167 0.293 -‐0.509 0.00847 -‐0.176 (-‐0.312) (0.477) (-‐0.569) (0.0125) (-‐0.561) R-‐squared 33.90% 23.50% 19.00% 19.80% 3.40%
Next, similar to results for O-score, expected rate changes do not explain the returns on MPC meeting dates, suggesting that their effect has already been incorporated into the stock prices. Finally, R2 for the four portfolios is much evenly distributed and consistently above 20% for all the portfolios. However, for the spread portfolio it is much less. This along with insignificant estimates for the spread portfolio for Z-Score portfolios might point that while the relationship for returns versus surprise rate changes holds, and distress level of companies matter in explaining the returns, O-Score maybe much better predictor of the distress levels of companies.
