The impact of interest rates on the house prices through the
risk-taking channel prior the financial crisis in 2008.
Name: Milco Breed
Student number: 10059210 Supervisor: Ron van Maurik
Introduction
The United States house prices peaked around 2005. After this peak house prices started to decline and some people began to experience huge losses (Greenlaw, Hatzius, Kashyap and Shin, 2008). In 2007 millions of homeowners default simultaneously. The effect of the defaults is a reduction in house prices and a crash in the subprime mortgage market.
Analysts suggest that the monetary policy of the Federal Reserve System (Fed) is responsible for the crash on the housing market (Hume and Sentence, 2009). Low interest rates during the period 2001-2007 affected risk-taking behavior. Recent studies found a link between low interest rates and risk-taking behavior. They found empirical results for the existence of a risk-taking channel. The housing market is also affected by the risk-taking channel. This paper conducts research on the effect of risk-taking behavior on the housing market. The research question is: “Does the risk-taking behavior during the period of low interest rates affected the housing market?”. The research can add new information about a link between low interest rates and risk behavior on the housing market prior the crisis of 2008 in the United States.
To find the link between low interest rates and risk behavior in the housing sector, We will first discuss the literature about the causes of the housing bubble. To understand the link between monetary policy and the risk behavior we are discussing the risk-taking channel (Borio and Zhu, 2008). For the existence of this risk-taking channel we look to recent studies about this topic. Beside the literature study, an empirical research to the effect of interest rates on the housing market is made.
The goal of the empirical research is to estimate the link between monetary policy and the housing market in the period 2001-2008. This link could exist due the risk-taking channel. To test the existence of the risk-taking channel in the housing market in the United States during the period 2001-2008, the type of regression is explained first. Second, a cross correlation between house prices and stock prices is calculated. As third is the value of the real estate investments relative to total investments compared between periods with different monetary policy. To see if a trend is responsible for an increase of the real estate investments relative to the total investments an augmented Dickey-Fuller test is used. The augmented Dickey-Fuller test tests also for stationary. A Chow test is used to test if there is a break in the house price index after changing monetary policy in 2001. Finally a regression is made for estimating the size of the effect of decreasing interest rates on house prices.
In part I the literature is discussed. Part II will focus on how to test for the effect of the risk-taking channel on the housing market. Part III will present an analysis of this effect and the conclusion will summarize this paper.
Part I: Describing the literature
A Housing Market Crisis
Since 2001 the Federal Reserve System (Fed) decided to follow a low interest rates monetary policy. The economy of the United States was still recovering from the dotcom bubble in 2000 while the economy was hit by an enormous shock in aggregate demand. On 9/11/2001 the United States was attacked. A massive exogenous shock hit in the aggregate demand. The Fed wanted to boost confidence by flooding the economy with liquidity. They began with a series of cuts in interest rates. After 9/11 the US economy was slammed by more shocks. The scandal of Enron, the space shuttle Columbia disaster, the war in Iraq and the hurricane Katrina. The Fed lowered the interest rates to counter each shock (Langdana, 2009).
The main question is how have the Fed involved in the housing bubble? There are numerous of causes for the housing bubble, but some of the most important causes are the growing GDP, a sharp increase in the subprime mortgage market and the low interest rates of the Fed. Till 2007 house prices continued to grow, but in 2007 millions of homeowners default simultaneously. According to Reinhart and Rogoff (2008) the huge run-up in equity prices was sustainable thanks to economic growth. Equity prices rises because of a growing GDP. Angell and Rowley (2006) and Kiff and Mills (2007) argue that the sharp increase in the subprime mortgage market (from 8% in 2001 to 20% in 2006) was facilitated by the
development of so-called private label mortgage backed securities. Those private label mortgage backed securities made it possible to invest in the mortgage market in the short term. Due the low interest rates, investors were searching for higher yields and increased their demand for the private label mortgage backed securities.
Hume and Sentence (2009) argue also that the low interest rates of the Fed were responsible for the house prices crash. The Fed did not anticipate to avoid the housing bubble, because they underestimated the effect of risk-taking behavior. Because the lack of solid empirical evidence on this topic, the Fed was not focusing on the effects of low interest rates. Only recently some studies tried to test for the existence of a risk-taking channel. The
risk-taking channel (Borio and Zhu, 2008) gives an explanation how monetary policy could affect house prices.
The theory of the risk-taking channel
Low interest rates could influence the behavior of the people towards risk-taking. The idea behind this behavior is the risk-taking channel found by Borio and Zhu (2008). Low interest rates encourage people to invest in more risky assets and this is the main idea behind the risk-taking channel. Borio and Zhu (2008) found that there are at least three different channels, which makes the risk-taking channel operative.
The first set of effects why banks take more risk operates through the impact of interest rates on valuations, incomes and cash flows. A reduction in the policy rate can boost asset and collateral value as well as incomes and profits. Increasing asset prices and low interest rates tend to reduce price volatility. Risk perception reduces and risk tolerance increases. For example, in a rising market where stock prices increases and so the value of equity relative to corporate debt increases as well. This reduces corporate leverage and could thus decrease the risk of holding stocks (Gambacorta, 2009). Financial firms will release risk budgets and encourage the behavior to take more risk. This is an example of the widespread use of value-at-risk methodologies for economic and regulatory capital purposes (Danielsson et al, 2004). This has an impact on house prices. A reduction in interest rates increases the value of housing collateral. The net worth of households is more affected by fluctuation in the value of collateral when they are more leveraged (Kiyotaki and Moore, 1997, 243-244). In more developed mortgage markets consumers tend to be more
leveraged, because they have easier access to credit. Moreover, the impact of a change in interest rates on the housing market tends to be stronger when leverage is higher (Bernake and Gertler, 1989, 14-31).
The second way that low interest rates could affect risk behavior of banks is through the search of yield (Rajan, 2005). Asset managers tend to take more risk for contractual, behavioral or institutional reasons when interest rates are low. In a period with low interest rates the asset managers have a low certain rates of return. Managers will seek to meet the nominal returns which they had also been able to achieve in a period when interest rates were higher (Bank of international settlements, 2004). Some reasons why the managers target the same nominal interest rates during periods with low interest rates are
psychological like money illusion. Managers may ignore the fact that nominal interest rates may decline to composite for lower inflation. According to Borio and Zhu (2008) other
reasons are institutional or contractual. Life insurance companies and pension funds manage their assets with reference to their liabilities. The liabilities are linked to a minimum
guaranteed nominal rate of returns reflecting long-term actuarial assumptions rather than the current level of yields. So life insurance companies and pension funds need to take more risk during a period with low interest rates. The nominal rate of returns reflecting long-term actuarial assumptions may exceed the yields available on highly rated government bonds in a period with declining interest rates. The resulting gap can lead to higher investments in risk products by institutions. The impact of this channel may be stronger when the gap between market and target interest rate is unusually large. This set of effects is connecting with the findings of Angell and Rowley (2006) and Kiff and Mills (2007).
The third way operates through the communications policies and reaction function of the central bank. For example, the degree of transparency of the central bank about future policy could influence the risk behavior. By increasing the degree of transparency the central bank is removing uncertainty about the future. This cuts off large downside risks and could encourage people to take more risk.
Summarize studies about the risk-taking channel
For the existence of the risk-taking channel some empirical studies are done recently. Those studies tried to test the existence of the risk-taking channel. These studies are based on country bank data in Spain (Jimenez et al, (2009) and Boliva (Ioannidou et al, 2009). Altunbas et al (2009) used data from banks in developed countries, while Gambacorta used the
econometric framework of Altunbas et al (2009).
Jimenez et al (2009) have done a study about the impact of short-term rates on credit risk-taking by analyzing a comprehensive credit register from Spain. Their main question was whether monetary policy, which was mostly decided abroad, has an impact on the level of risk of individual bank loans. They found that in the short term low interest rates reduce the probability of default of outstanding variable rate loans, by reducing interest burdens of existing borrowers. In the medium and long term banks tend to grant more risky loans and the lending standards went down. This was because the higher collateral values and the
search for yield of the banks. Overall they concluded that in the short term the risk-taking channel does not hold, but that it holds in the medium and long term.
Ioannidou et al (2009) have analyzed the impact of monetary policy on bank risk-taking based on Bolivian data from 1999-2003. Their research was about investigating the impact of changing interest rates on loan pricing. They concluded that relaxing monetary conditions increases the risk-appetite of banks. The Bolivian banks increase the number of new risky loans when interest rates are low, but also reduce the rates they charge for risky borrowers relative to the rates they charge for less risky borrowers.
Altunbas et al (2009) analyzed the link between monetary policy and the expected frequency of banks in a more international perspective. The data they used was from listed banks operating in 16 developed countries from 1999-2008. Their paper analyzed the first two effects of the risk-taking channel of Borio and Zhu (2008). They found evidence that short-term low interest rates over an extended period of time increased risk-taking behavior of banks over the last decade.
These studies tried to test for the existence of risk-taking channel and all three researches found empirical evidence for the existence of a risk-taking channel. Gambacorta (2009) analyzed in his paper which factors have influenced the evolution of bank risk in the current crisis. His research based on a comprehensive database of listed banks operating in the European Union and the United States. Gambacorta (2009) was building on the
econometric work by Antunbas et al (2009) and found evidence of a link between an extended period of low interest rates prior to the crisis and banks risk-taking. Monetary policy is not fully neutral from a financial stability perspective. More interesting, Gambacorta (2009) found a positive relation for the change in the house price index corrected for
inflation and the change in the riskiness of a given bank. Part II: Testing the risk-taking channel on the housing market
In this part the existence of the risk-taking channel in the housing market is tested. The type of regression models and there validity is discussed. A cross correlation between house price index and stock price index is calculated. A growth of invested dollars in the real estate market after 2000 is tested. Also trends and a break of the house price index after 2001 are tested and finally a regression model is made for estimating the effect of the risk-taking channel on the housing market.
Type of regression models
The data consists monthly observations of the period 1 January 1991 till 1 January 2009: • The house price index, provided by Freddie Mac;
• stock price index of the S&P500; • Moody’s AAA bonds interest rates;
• real estate value of owned and managed receivables of financial firms; • total value of owned and managed receivables of financial firms; • Gross Domestic Product (GDP) of the United States;
• capital to asset ratio of banks in the United States; • inflation rates of the United States;
• Unemployment rates in the United States.
The variables used in the data are not independent and identically distributed. Values in the data are dependent from the values in the previous period. This causes autocorrelation or serial correlation (Stock and Watson, 2011, 522-523). To estimate coefficients with ordinary least squares the assumption of independently and identically distributed observations fails. Using time series data can solve the problem of autocorrelation. The assumption for
independent and identically distributed observations is replaced for random variables have a stationary distribution and �𝑌𝑌𝑃𝑃,𝐾𝐾1𝑡𝑡. . , 𝐾𝐾𝐾𝐾𝑡𝑡� and �𝑌𝑌𝑡𝑡−𝑗𝑗, 𝑋𝑋1𝑡𝑡−𝑗𝑗, . . , 𝑋𝑋𝑘𝑘𝑡𝑡−𝑗𝑗� become independent as j gets large (Stock and Watson, 2011, 537). For testing stationary an augmented Dickey-Fuller test can be used. According to Stock and Watson (2011, 537) other assumptions for a time series regression with multiple predictors are𝐸𝐸�𝑢𝑢𝑡𝑡�𝑌𝑌𝑡𝑡−1,𝑌𝑌𝑡𝑡−2, . . , 𝑋𝑋1𝑡𝑡−1, 𝑋𝑋1𝑡𝑡−2, � = 0, large outliner are unlikely and there is no perfect multicollinearity. For time series a number of lags need to be used. The number of lags used in the estimating models is based on the Bayes information criterion (Stock and Watson, 2011, 543).
Testing the connection between house price index and stock price index
According to the theory of Borio and Zhu (2008) house prices increase in a period with low interest rates, because the demand for riskier assets increases due the risk-taking channel. To test the effect of the risk-taking channel on house prices, a cross correlation of the sample between house prices and other risky assets like stocks provides information about
the connection of some of the risky assets discussed by Borio and Zhu (2008). To explore the relationship between the house price index and the stock price index a cross correlations can be used. The cross correlation gives a connection between the independent variable: house price index, and the dependent variable: stock price index.
The data consists information of the S&P 500 index from 1 January 1991 till 1 January 2009 and information of the house price index from 1 January 1991 till 1 January 2009. This cross correlation coefficient gives information about the connection between the growth of house prices and the growth of stock prices. A disadvantage of using a cross correlation coefficient is that it does not measure for causation. The cross correlation coefficient only gives information about how the variables are correlated.
Testing the proportion of real estate value relative to the total value
Since 2001 the Fed followed a low interest rate monetary policy. But because the lack of empirical research on the taking channel effect they were not anticipating on this risk-taking channel effect. In 2007 the housing market was in a bubble (Langdana, 2009). To test a possible existence of the risk-taking channel in the housing market, the period with low interest rates should be compared with a period with higher interest rates. According to the literature, the Fed decided in 2001 to follow a low interest rate monetary policy till 2007.
Before 2001 the Fed followed another monetary policy and the interest rates were higher. For testing the existence of the risk-taking channel, the proportion of real estate value of owned and managed receivables (in millions of dollars, seasonally adjusted) of financial firms (banks and finance company subsidiaries of bank holding companies, but not retailers of banks) divided to the total value (consumers market, business market en real estate market together) of owned and managed receivables of financial firms from 1 January 1991 till 1 January 2001 will be test against data from 1 January 2001 till 1 January 2008 with a comparing two proportions parametric test . The proportion of the real estate value
relative to the total value is chosen, because the total value consolidates a part of other economic variables that could influence the growth of the real estate value. For formula for the two proportion parametric test is:
𝑍𝑍 = 𝑝𝑝1 − 𝑝𝑝2
√(𝑝𝑝(1 − 𝑝𝑝)( 1n1 +n2)1 ~𝑁𝑁(0,1) 7
With p, p1 and p2:
𝑝𝑝 =n1𝑝𝑝1 + n2𝑝𝑝2n1 + n2 ,
𝑝𝑝1 = 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝐸𝐸𝐸𝐸𝐸𝐸𝑅𝑅𝐸𝐸𝑅𝑅 𝑣𝑣𝑅𝑅𝑅𝑅𝑢𝑢𝑅𝑅 2001 − 2007𝑇𝑇𝑇𝑇𝐸𝐸𝑅𝑅𝑅𝑅 𝑉𝑉𝑅𝑅𝑅𝑅𝑢𝑢𝑅𝑅 2001 − 2007 , 𝑝𝑝2 = 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝐸𝐸𝐸𝐸𝐸𝐸𝑅𝑅𝐸𝐸𝑅𝑅 𝑉𝑉𝑅𝑅𝑅𝑅𝑢𝑢𝑅𝑅 1991 − 2000𝑇𝑇𝑇𝑇𝐸𝐸𝑅𝑅𝑅𝑅 𝑉𝑉𝑅𝑅𝑅𝑅𝑢𝑢𝑅𝑅 1991 − 2000 This test is valid, because n1p = 19.19 and n1(1-p) = 76.81, n2p = 23.98 , n2(1-p) = 96.02 All values are above 5.
Augmented Dickey Fuller test
An augmented Dickey Fuller test is testing if the variables of the dataset have a stationary distribution. The findings of the previous part, the proportion of real estate value relative to the total value, could be influenced by a trend. A trend is a persistent long term movement of a variable over time (Stock and Watson, 2008, 546). The ratio between real estate value and the total value could have a stochastic trend. A stochastic trend is random and varies over time (Stock and Watson, 2008, 546-551). Problems caused by stochastic trends are autoregressive coefficients that are biased toward zero, nonnormal distributions of t-statistics and spurious regression. To test for the existence of a trend in the ratio between real estate value and the total value an augmented Dickey-Fuller test is used. The
augmented Dickey-Fuller test will look as followed with p lags:
∆𝑌𝑌𝑡𝑡 = 𝛽𝛽0+ 𝛿𝛿𝑌𝑌𝑡𝑡−1+ 𝛾𝛾1∆𝑌𝑌𝑡𝑡−1+ 𝛾𝛾2∆𝑌𝑌𝑡𝑡−2+. . +𝛾𝛾1𝑝𝑝∆𝑌𝑌𝑡𝑡−𝑝𝑝+ 𝑢𝑢𝑡𝑡 𝐻𝐻0: 𝛿𝛿 = 0 𝑣𝑣𝐸𝐸. 𝐻𝐻1: 𝛿𝛿 < 0
But when the time series has a trend in it and this trend is potentially slow-turning around a linear trend line, then this trend, “t” (the observation number), must be added as an
additional regressor. In this case the Dickey-Fuller regression becomes:
∆𝑌𝑌𝑡𝑡= 𝛽𝛽0+ 𝛼𝛼𝐸𝐸 + 𝛿𝛿𝑌𝑌𝑡𝑡−1+ 𝛾𝛾1∆𝑌𝑌𝑡𝑡−1+ 𝛾𝛾2∆𝑌𝑌𝑡𝑡−2+. . +𝛾𝛾1𝑝𝑝∆𝑌𝑌𝑡𝑡−𝑝𝑝+ 𝑢𝑢𝑡𝑡 𝐻𝐻0: 𝛿𝛿 = 0 𝑣𝑣𝐸𝐸. 𝐻𝐻1: 𝛿𝛿 < 0
Under the null hypothesis, 𝑌𝑌𝑡𝑡 has a stochastic trend (unit root). Under the alternative hypothesis 𝑌𝑌𝑡𝑡 is stationary. To avoid problems caused by stochastic trends, take the first difference of the time series. The first difference can also be tested if it still has a trend with the augmented Dickey Fuller test.
Testing for a break
A Chow test (1960) allows for a break whether a particular date causes a break in the regression coefficients. According to the first channel of the risk-taking channel (Borio and Zhu, 2008) collateral value is boosted by low interest rates. The interest rates changed in 2001 by the Fed. The house price index is the dependent variable and interest rates on Moody’s AAA bonds, the most save bonds where investor can invest in (Ratings Symbols and Definitons, 2014, 5), is the independent variable.
A time series regression on the house price index
Next, an econometric analysis is made. The econometric model estimates the effect of interest rates on the house price index. According to the risk-taking channel of Borio and Zhu (2008), declining interest rates should increase the amount of money invested in riskier assets like stocks and houses. To estimate the effect of decreasing interest rates during the period 2001-2007 on the housing market, a time series regression estimates the coefficient about the power of decreasing interest on the house price index. The number of lags (q) for each variable is based on the Bayes information criterion (Stock and Watson, 2011). Robust standard errors are used to correct standard errors for model misspecification (Stock and Watson, 2011).
There is a lack of data of risk-perceptions of banks or financial firms on the housing market. To estimate an effect of decreasing interest rates through the risk-taking channel on the housing market, data of the logarithm of the house price index is used to estimate an effect on the housing market. The logarithm of the house price index is the dependent variable. Lag values are used to correct for autocorrelation. The logarithm is used, because the house price index grows exponentially. A reduction in interest rates boosts the collateral value. The risk-taking channel encourages people to invest in more risky assets, for example
houses. The hypothesis is that decreasing interest rates should increase the house price index through the risk-taking channel.
One of the independent and explaining variables is the interest rates on Moody’s AAA bonds, the most save bonds where investors can invest in (Ratings Symbols and Definitons, 2014, 5). To get correct estimated coefficients the model need some control variables to filter the effect of decreasing interest rates. To conclude that the risk-taking channel have affected the housing market other effects on the housing market have to be eliminated. The control variables in the model are the logarithm of the stock price index, capital to asset ratio of banks, logarithm of GDP, inflation rate and the unemployment rate all of the US. These are the other independent variables. The logarithm of the stock price index is based on the percentage changes on historical prices of the S&P 500. The log is used, because the stock market grows exponentially. The stock market is connected with the housing market due the subprime mortgage market (Longstaff, 2010). A rising stock market tend people to invest more in stocks instead of houses and so lowers the value invested in houses (David Blitzer, 2013). On the other hand, a declining stock market tends people to invest more in real estate market as a substitute of the stock market.
The capital to asset ratio gives information about leverage of banks in the United States. Higher leverage makes the impact of a change in interest on the housing market stronger (Bernake and Gertler, 1989). The logarithm of GDP, anindicator of the real
aggregate output, is used to eliminate the effect of the huge-run up in equity prices caused by economic growth (Reinhart and Rogoff, 2008). The logarithm convert changes in variables into percentage changes and used for exponentially growth. Inflation and unemployment are macroeconomics variables that affect economic growth. To get reliable estimators on the impact on the house price index, the effect of interest rates on house price index needs to be consolidate from the control variable that could influence the housing market. Part III: Analyze the testing results
The connection between the house price index and the stock price index
A cross correlation coefficient is calculated to estimate the connection between house prices and stock prices, according to the data of the S&P 500 from 1991-2008 and data of the house price index from 1991-2008 provided by Freddie Mac. The cross correlation coefficients are shown in figure 1. The result suggests a connection between the housing
market and the stock market. The cross correlation is needed to conclude that the risk-taking channel (Borio and Zhu, 2008) could exist in the stock market as well as in the housing market. Figure 2 shows the logarithm of the house price index against the logarithm of the stock price index between 1991 and 2008. The stock market crashed in 2001 through the dotcom bubble and 9/11, while the housing market grew almost linear. Disadvantage of the cross correlation is that it not explains causation.
Figure 1: cross correlation. Source: Own calculations based on data of Freddie Mac
Figure 2: Log of the House Price Index versus log of the Stock Price Index. Source: Own calculations based on data of Freddie Mac and Yahoo finance.
Analyze the proportion of the real estate value relative to the total value
The two proportions parametric test is testing if the proportion of real estate value relative to the total value is significantly increased over the period 2001-2007 compared to the period 1991-2000. The outcome of the test has a z-value of 2.02029 and a p-value of
0.02168 or an alpha of 2.17%. Given an alpha of 5% the proportion of the real estate value of
owned and managed receivables (in millions of dollars, seasonally adjusted) to the total value of owned and managed receivables is significantly increased in the period 2001-2007.
In the period with low interest rates, the value of owned and managed receivables in the real estate market relative to the total value of owned and managed receivables is increased. Comparing this result to the relevant literature it is plausible to say that the part of invested dollars in the real estate market compared to the total value (consumer market, business market and real estate market) is increased. Financial firms invested relative more money in the real estate market. One possible reason why this happened is through the risk-taking channel, which explains the behavior towards risk during a period of low interest rates. Low interest rates have encouraged people to take more risk and invest in more risky assets. One example of a riskier asset is investing in the real estate market. This effect became stronger due the possibility to invest in the short term (Angell and Rowley, 2006) in the real estate market. An increasing amount of money invested in the real estate market increases the demand for real estate assets. This is one of the explanations why house prices have increased in the period 2001-2007. But on the other hand this result is likely to be affected by a trend. Moreover, decreasing returns on the stock market could be an explanation why people invested more money in the real estate market. In that case the risk-taking channel has not affected the real estate market.
The augmented Dickey Fuller test outcomes
Inflation and unemployment are flat and potentially slow-turning around a non-zero value. For inflation and unemployment the augmented Dickey-Fuller test without a time trend is used. For the ratio of the real estate value relative to the total value, logarithm of the house price index, Moody’s AAA rates, logarithm of GDP, capital to asset ratio, and the logarithm of the stock price, an augmented Dickey-Fuller test with a time trend is used. For the variables with an outcome below 5% significance value, the first differences are tested.
The outcomes of the augmented Dickey fuller test are shown in table 1. The ratio of the real estate value relative to the total value, AAA rates, logarithm of GDP, inflation, capital to asset ratio, unemployment and the logarithm of the stock price index are not stationary and have a stochastic trend. The logarithm of the house price index does not have a stochastic trend. Using the first difference eliminates random walk trends in a series (Stock and Watson, 2011, 555). For a time series regression all variables need to be stationary.
ADF Test R/T
ratio Log(HPI) AAA rates Log(GDP) inflation unemployment C/A ratio Log(SPI)
Outcome -2.506 -1.784 Outcome with time trend -1.154 -4.400 -3.171 1.267 -0.562 -1.986 First dif. -8.717 -6.283 -6.698 -10.248 -3.984 -6.98 -9.872
Table 1: Augmented Dickey Fuller Test values
Testing for a break
The outcome of the Chow test gives a F-value of 18.99 (p-value of 0.000). A significant break is found after 2001. Based on this test it is likely to say the change of interest rates in 2001 of the Fed have led to an increase of the house price index.
A time series regression on the housing market
The time series regression model is estimated as followed: 𝐿𝐿𝑇𝑇𝐿𝐿𝑅𝑅𝐿𝐿𝐿𝐿𝐸𝐸ℎ𝑚𝑚 𝑇𝑇𝑜𝑜 𝐸𝐸ℎ𝑅𝑅 𝐻𝐻𝑇𝑇𝑢𝑢𝐸𝐸𝑅𝑅 𝑃𝑃𝐿𝐿𝐿𝐿𝑃𝑃𝑅𝑅 𝐼𝐼𝐼𝐼𝐼𝐼𝑅𝑅𝐼𝐼 = 𝛽𝛽0+ 𝛽𝛽1 𝐿𝐿𝑇𝑇𝐿𝐿𝑅𝑅𝐿𝐿𝐿𝐿𝐸𝐸ℎ𝑚𝑚 𝑇𝑇𝑜𝑜 𝐸𝐸ℎ𝑅𝑅 𝐻𝐻𝑇𝑇𝑢𝑢𝐸𝐸𝑅𝑅 𝑃𝑃𝐿𝐿𝐿𝐿𝑃𝑃𝑅𝑅 𝐼𝐼𝐼𝐼𝐼𝐼𝑅𝑅𝐼𝐼𝑡𝑡−1+ ⋯ + 𝛽𝛽13 𝐿𝐿𝑇𝑇𝐿𝐿𝑅𝑅𝐿𝐿𝐿𝐿𝐸𝐸ℎ𝑚𝑚 𝑇𝑇𝑜𝑜 𝐸𝐸ℎ𝑅𝑅 𝐻𝐻𝑇𝑇𝑢𝑢𝐸𝐸𝑅𝑅 𝑃𝑃𝐿𝐿𝐿𝐿𝑃𝑃𝑅𝑅 𝐼𝐼𝐼𝐼𝐼𝐼𝑅𝑅𝐼𝐼𝑡𝑡−13+ ∆𝛿𝛿01 𝐴𝐴𝐴𝐴𝐴𝐴 𝑏𝑏𝑇𝑇𝐼𝐼𝐼𝐼𝐸𝐸𝑡𝑡 + ∆𝛿𝛿11 𝑅𝑅𝑇𝑇𝐿𝐿𝑅𝑅𝐿𝐿𝐿𝐿𝐸𝐸ℎ𝑚𝑚 𝑇𝑇𝑜𝑜 𝐺𝐺𝐺𝐺𝑃𝑃𝑡𝑡+ ∆𝛿𝛿21 𝑅𝑅𝑇𝑇𝐿𝐿𝑇𝑇𝐿𝐿𝐿𝐿𝐸𝐸ℎ𝑚𝑚 𝑇𝑇𝑜𝑜 𝐸𝐸ℎ𝑅𝑅 𝑆𝑆𝐸𝐸𝑇𝑇𝑃𝑃𝑆𝑆 𝑝𝑝𝐿𝐿𝐿𝐿𝑃𝑃𝑅𝑅 𝐿𝐿𝐼𝐼𝐼𝐼𝑅𝑅𝐼𝐼𝑡𝑡 + ∆𝛿𝛿31 𝑃𝑃𝑅𝑅𝑝𝑝𝐿𝐿𝐸𝐸𝑅𝑅𝑅𝑅 𝐸𝐸𝑇𝑇 𝑅𝑅𝐸𝐸𝐸𝐸𝑅𝑅𝐸𝐸 𝐿𝐿𝑅𝑅𝐸𝐸𝐿𝐿𝑇𝑇𝑡𝑡+ ∆𝛿𝛿41 𝑢𝑢𝐼𝐼𝑅𝑅𝑚𝑚𝑝𝑝𝑅𝑅𝑇𝑇𝑢𝑢𝑚𝑚𝑅𝑅𝐼𝐼𝐸𝐸𝑡𝑡+ ∆𝛿𝛿51 𝐿𝐿𝐼𝐼𝑜𝑜𝑅𝑅𝑅𝑅𝐸𝐸𝐿𝐿𝑇𝑇𝐼𝐼𝑡𝑡 + 𝜀𝜀
The time series regression has the logarithm of the house price index as dependent variable. The first 13 lags, estimated by the Bayes information criterion, correct the model for
autocorrelation. The difference of the rate on AAA bonds; difference of the logarithm of the stock price index; difference of the capital to asset ratio; difference of the logarithm of the GDP; difference of the inflation and the difference of the unemployment rate are the independent variables. The differences are used for stationary. The outcome of the
regression model is shown in table 2. The regression model has a 𝑅𝑅2 of 0.9998. This means that 99.98% of the house price index fits with the estimated variables. The first difference of the logarithm of Moody’s AAA bonds has a coefficient of -0.0005259, and a p-value of 0.692. This coefficient is not significant. Consistent with the hypothesis of the risk-taking channel
there is no empirical evidence that the risk-taking channel has affected the housing market. The estimated result shows that the effect of the first difference of the interest rates has no significant effect on the house price index.
The effect of changing interest rates on the house price index is consolidated from changes in the stock market, the leverage rate of banks and the macroeconomic variables such as GDP, inflation and unemployment. The most uncertain variables in the model are: the logarithm of the GDP, unemployment and the logarithm of the stock price index. There is less uncertainty about inflation and the capital to asset ratio.
This first difference of the logarithm GDP is the growth rate of the GDP. The effect of the growth rate of the GDP is not significant. A possible explanation is that house prices could not immediately react if numbers of the GDP are released. The stock market is a substitute for the housing market, but is related with the housing market as well. The regression does not consider the fact that the subprime market is traded on the stock market so there is a lack of connection between the stock prices and the house prices. This could lead to uncertain effects on the house prices.
A significant link between interest rates of Moody’s AAA bonds and the house prices index during the period 2001-2008 is not found. The explanation according to the risk-taking channel of Borio and Zhu (2008) is that investors are more willing to invest in more risky assets like, for example houses. Decreasing interest rates should increase the house price index. This link is not found between the first difference of the Moody’s AAA bonds and the logarithm of the house price index. Possible explanation for this effect is the inability of house prices to adapt on a change in interest rates. House prices need some months to react on a change in interest rates.
Figure 2: Time series regression
Conclusion
The financial crisis in 2008 has drawn a link between monetary policy and risk behavior for researchers and policymakers. This link is the risk-taking channel (Borio and Zhu, 2008) and this channel also affected the housing market. Since 2001, the year that the economy of the United States was still recovering from the dotcom bubble and they were attacked on 9/11, the Federal Reserve System decided to follow a low interest rates monetary policy. Some
Time series regression Number of ob 70
F( 19, 50) 19328.88
Prob > F 0
R-squared 0.9998 Root MSE 0.00226 Logarithm of the House Price Index
Robust
Coef. Std. Err. t P>t [95% Conf.Interval] Logarithm of the House Price Index
Log(HPI) t-1 1.192632 0.165064 7.23 0 0.8610924 1.524172 Log(HPI) t-2 -0.00114 0.279908 0 0.997 -0.5633528 0.5610686 Log(HPI) t-3 0.075916 0.193296 0.39 0.696 -0.3123296 0.4641625 Log(HPI) t-4 -0.42135 0.193744 -2.17 0.034 -0.8104982 -0.032206 Log(HPI) t-5 -0.02511 0.216291 -0.12 0.908 -0.4595413 0.4093254 Log(HPI) t-6 0.333035 0.2011 1.66 0.104 -0.0708855 0.7369553 Log(HPI) t-7 -0.31653 0.217886 -1.45 0.153 -0.7541709 0.1211017 Log(HPI) t-8 0.186156 0.177441 1.05 0.299 -0.1702449 0.5425571 Log(HPI) t-9 -0.26262 0.180113 -1.46 0.151 -0.6243918 0.0991447 Log(HPI) t-10 0.942442 0.227096 4.15 0 0.4863053 1.398578 Log(HPI) t-11 -0.71127 0.237597 -2.99 0.004 -1.188499 -0.2340433 Log(HPI) t-12 0.157656 0.317203 0.5 0.621 -0.4794648 0.7947768 Log(HPI) t-13 -0.16047 0.225512 -0.71 0.48 -0.6134191 0.2924879 Moody's AAA bonds
∆ AAA Bonds -0.00053 0.001321 -0.4 0.692 -0.003179 0.0021271 Logarithm of GDP ∆ Log(GDP) 0.006093 0.07113 0.09 0.932 -0.1367753 0.1489621 Logarithm of S&P 500 ∆ Log(S&P500) -0.00607 0.005218 -1.16 0.25 -0.01655 0.0044114
Capital to asset ratio
∆C/A ratio 0.00564 0.002824 2 0.051 -0.0000324 0.0113129 Unemployment ∆ unemployment -0.00269 0.003267 -0.82 0.414 -0.0092537 0.0038684 Inflation ∆ inflation -0.00459 0.002255 -2.03 0.047 -0.0091191 -0.0000595 Constant 0.054936 0.012981 4.23 0 0.0288615 0.0810094 15
analysts suggest that this monetary policy is responsible for the housing bubble in 2007 due the risk-taking channel. The main idea of the risk-taking channel is that low interest rates encourage people to take more risk. This works through three separate channels of the risk-taking channel. First, through the impact of interest rates on valuations, incomes and cash flows. Second, through the search for yield and the third one is through the communications policies and reaction function of the central bank (Borio and Zhu, 2008).
The effect of risk behavior was underestimated by the Fed, because there was a lack of empirical research on this topic. Only recent studies tried to test the existence of the risk-taking channel. Jimenez et al(2009) and Ioannidou et al(2009) found a link between low interest rates and risk-taking of banks in Spain and Bolivia. Altunbas et al (2009) found evidence that short-term low interest rates have increased risk-taking behavior of banks over the last decade. Gambacorta (2009) was building on the econometric work by Antunbas et al (2009) and found evidence of a link between an extended period of low interest rates prior to the financial crisis and banks risk-taking.
This paper does research to the influence of the risk-taking channel on the housing market prior the financial crisis. The research is based on time series. The main implication of the findings: there is not found a direct link between decreasing interest rates and increasing house prices in the period 2001-2008. A time series regression shows a link between the growth rate of the house price index and interest rates of Moody’s AAA bonds, growth rate of GDP, inflation, capital to asset ratio, unemployment and the growth rate of the stock price index. The empirical research has some weak estimators. This resulting in uncertainty of the link between interest rates en house prices.
It is important that monetary authorities are aware of the effect of risk-taking behavior en the consequences of this behavior. Monetary policy affected the housing market and stock market due the risk-taking behavior, while the monetary authorities underestimated the effect of this behavior.
Bibliography
Altunbas, Yener, Leonardo Gambacorta and David Marques-Ibanez. "Does monetary policy affect bank risk?" Banger Business School, 2012.
Angell, C. and C. D. Rowley. "Breaking New Ground in U.S. Mortgage Lending. FDIC: Outlook Summer 2006." Fderal Deposit Insurance Corporation, 2006.
"Bank for International Settlements." Bank for International Settlements. 2004.
Bernake, Ben S and Mark Gertler. "Agency costs, net wort and business fluctuations." American
Economic Review 79(1), 1989: 14-31.
Borio, Claudio and Haibin Zhu. "Capital regulation, risk-taking and monetary policy: a missing link int he transmission mechanism?" BIS working papers 268, Bank for international settlements, 2008.
Chow, Gregory C. "Tests of Equality Between Sets of Coefficients in Two Linear Regressions."
Econometrica, Vol 28, No. 3., 1960: 591-605.
Gambacorta, Leonardo. "Monetary policy and the risk-taking channel." BIS Quartely Review, 2009. Greenlaw, David, Jan Hatzius, Anil K Kashyap and Hyon Song Shin. "Leveraged Losses: Lessons from
the mortgaga market meltdown." Proceedings of the U.S. monetary policy forum., 2008: 7-59.
Hume, Michael and ANdrew Sentace. "The global credit boom: Challenges for macro-economics and policy." Journal of international money and finance 28(8), 2009: 1426-1461.
Ioannidou, Vasso, Steven Ongena and Jose-Luis Peydro. "Monetary policy and subprime lending: a tall tale of low federal funds, hazardous loans and reduced loan spreads." European Banking
Centre Dscussion Paper, no 2009-04s, 2009.
Jimenez, Gabriel, Steven Ongena, Jose-Luis Peydr and Jesus Saurina. "Hazardus times for monetary policy: what do twenty-three millon bank loans say about the effects of monetary policy on risk-taking?" Bank of Spain, Working papers, no 833, 2009.
Kiff, J. and P. Mills. "Money for Nothing and Checks for Free: Recent Developments in U.S. Subprime Mortgage Markets." Working Paper 07/188, International Monetary Fund, 2007.
Kiyotaki, Nobuhiro and John Moore. "Credit Cycles." Journal of political economy 105(2), 1997: 243-244.
Longstaff, Francis A. "The subprime credit crisis and contagion in financial markets." Journal of
Financial Economics 97, 2010.
Rajan, Raghuram G. "Has financial development made the world riskier?" National Bureau of
Economic Research Working Paper Series, no 11728., 2005.
"Ratings Symbols and Definitons." Moody's investors service, April 2014.
Reinhart, Carmen M. and Kenneth S. Rogoff. "Is the 2007 U.S. sub-prime financial crisis so different? An international comparison." National bureau of economic research, 2008.
Stock, H. James, Mark W. Watson. Introduction to Econometrics. Pearson, third edition, 2011.
