Value_Added_r9.DVI



Draft (Please do not quote, comments welcome)What is Prompting Australia to Intervene in the Residential Mortgage Market? A New Approach to the Analysis of Interest Rate Differentials on Single-family Residential Home LoansJames D. Shilling? and Piyush Tiwari ?center0 DePaul University University of Melbourne00 DePaul University University of MelbourneVersion: November 18, 2018AbstractThis paper focuses on understanding the observed differences in interest rates on single family residential mortgages during September 2008 to December 2017. Exploiting the conceptual difference in risks associated with fixed rate and variable rate mortgages for lenders, we construct a synthetic variable rate. Synthetic variables are obtained from 3 year fixed rate by adjusting them for interest rate risks premium and call options that are embedded in fixed rates. Estimated error correction model for the difference between actual and synthetic mortgage rate reveals that the unbiasedness hypothesis is rejected and that the lenders in pricing actual variable rates have attached a risk premia of 73 to 97 basis points over synthetic rates. This requires further investigation into institutional arrangements, market structures, underwriting and lending practices of banks as these remain unexplained. Keywords: mortgage rate differences, swaps, swaptions, error correction modelJEL Classifications: G21?DePaul University, 1 East Jackson Boulevard, Chicago, IL 60604, Email: shilling@depaul.edu. ? University of Melbourne, Parkville, VIC 3010, Australia, Email: piyush.tiwari@unimelb.edu.auIntroductionThe Australian banking royal commission is kicking off with an inquiry into the pricing of single-family residential home loans in Australia. The current outstanding mortgage debt in Australia is around $1.6 trillion spread across six million loan accounts (ACCC, 2018). Between June 2007 and December 2017, the Australian banking sector originated more than $0.9 trillion in residential home loans (ABS, 2018). That is more than 1.7 times the money lent on personal loans, credit cards and business borrowing (ibid). It is now clear that Australian households are dangerously overleveraged, and as the ratio of mortgage debt to GDP has reached an all-time high of about 1.9, which is among the highest in advanced countries (Pascoe, 2018). Concerns have risen over the past several years that the highly concentrated retail banking sector in Australia, with the four major banks having more than 80% of the residential mortgage market and household deposits, has been overcharging fees and interest rates on single-family residential home loans since 2011 (ACCC, 2018). Given this belief, this paper, improving significantly over concepts and approaches followed by similar studies in the finance literature that has examined the pricing of single-family residential home loans in the 1980s and 1990s, is intended to analyze objectively the interest rate differentials on single-family residential home loans in Australia over the September 2008 to December 2017 period. The paper is the first in estimating rate differences on actual versus synthetic single-family residential home loans. A debtor with a floating-rate loan can convert this loan into a fixed-rate loan with an option to repay, in an attempt to hedge against foreseen increases in interest rates using interest rate swaps and swaptions. Similarly, fixed-rate loan can be converted into a synthetic variable-rate loan by simply reversing this process, which is the approach taken in this paper. We take a fixed rate single-family residential home loan and then use interest rate swaps and swaptions to convert this loan into a variable-rate loan to determine the existence of differences between actual and synthetic mortgage financing costs in Australia. Any resulting difference between actual and synthetic mortgage financing costs can be attributed to one of two factors: imperfections in the financial market, barriers to the efficient flow of funds, and less than perfectly competitive lending conditions versus differences in credit risk, loan characteristics (e.g., loan-to-value ratio), funding and servicing costs, and other characteristics of the lender-borrower relationship. This is a new insight in the interest rate differential literature on single-family residential home loans. This paper will ask the question: Is the royal banking commission justified in intervening into the single-family residential home loan market at this time? As shown in Figure 1, the Actual versus synthetic single-family residential home loans changed greatly during the period between September 2008-December 2017. The gap between actual and synthetic widened after September 2008 and it has persisted since then.Figure 1: Actual versus synthetic mortgage ratesInstitutional BackgroundThe Australian mortgage market is dominated by four large banks of national character, and the loans they make are funded primarily through deposits with short duration. Hence, most mortgage loans in Australia – approximately 75% of new owner occupied residential mortgages– are variable-rate loans (ACCC, 2018) that track the official cash rate, set by the Reserve Bank of Australia.There is a modest amount of fixed-rate lending in the Australian mortgage market, owing to the use of securitization. However, unlike in the U.S., the loan terms are fixed generally for only 3 years or less (though fixed terms up to 10 years are available, these are generally not offered to customers), after which time the loan converts automatically to a variable-rate loan. Mortgage loans in Australia are classified as follows: first there are basic loans with limited options. Next, there are flexible loans with facilities such as the option to redraw, and the option to make early repayments or convert from a variable- to a fixed-rate, or vice versa. Then there are discounted loans, where the loan has a discounted or “honeymoon” rate of interest for the first year of the life of the loan. ACCC (2018) found that new borrowers on variable rate contracts paid significantly lower interest rates than the existing borrowers. It may be noted that fixed rate mortgages lack add-on features compared to variable rate mortgages. Features such as redraw, offset and progressive drawdown are only available on variable rate mortgages (ACCC, 2018). Almost all home mortgages in Australia are underwritten solely based on the ability to pay back the loan. For example, most lenders set maximums for these ratios, such as, e.g., the monthly mortgage repayment cannot exceed 25 to 35% of the borrower’s monthly income. However, most lenders will allow a higher payment-to-income ratio for borrowers who are well-qualified.Loan-to-value ratios (LVR) in Australia can range in size from 80 to 95%. Lenders have reduced their risks arising from high LVR since 2009 by consistently lowering their LVRs but still nearly one fifth of approvals have LVRs that exceed 80% (Somasundaram, 2017). These low loan-to-value ratios offset a high debt-to-income ratio. The typical loan term is 25 years but could range from 20 to 30 years. In practice, most mortgages are often paid off well before stated maturity with households choosing to make excess repayments on standard loans. The annual limit on excess payment of principal in a fixed rate mortgage is restricted.The concentrated industry sector has led to high returns for retail banks. A report by Morgan Stanley (2018) estimates that Australia’s four largest retail banks have returns on capital that are 15% higher than business banks and 20% higher than institutional banking and are higher than peers (including banks in the US, Sweden, Singapore, Hong Kong, Spain, France, the UK, and Japan). What is more, the source of these excess returns appears to be from housing loans and household deposits. Since deregulation in the 1980s, the major retail banks have increased their market power due in part to the acquisition of smaller banks and partly to the stagnant market shares of international banks. Another barrier to entry for smaller competitors has been the limited access to customer data. This limited access to customer data has made it difficult for new entrants to identify and target the major banks’ most valuable customers.Lenders in Australia have followed a complex and opaque mortgage pricing structure. The interest rates that the borrowers pay has two components: headline interest rate and any discount off the headline rate granted by the lender. Headline variable or fixed rates are the reference rates on which variable and fixed interest rates paid by new and existing customers are based. ACCC (2018) highlights that prior to 2015 lenders had single headline variable rate (referred as standard variable rate) but now they maintain several headline variable rates for various products and combination of borrower and repayment types. The residential mortgage price inquiry conducted by ACCC (2018) found that the discounts that the banks offer on headline variable rates are highly discretionary and non-transparent. These depend on factors such as perceived borrowers’ risk profile, geographic location of borrower or their property, borrower’s value to the bank and bank’s desire to write new business. Borrower’s ability to negotiate also influences the discount.The complexity in mortgage choice for borrowers has led to borrowers’ increased reliance on mortgage brokers. More than half of residential mortgages are originated through mortgage brokers and large number of these firms are owned by or affiliated with a major bank (ACCC, 2018).There are two major types of borrowers – one who borrow for owner-occupied housing and the two, who borrow for investment housing. The average growth in investment housing lending was higher than lending for owner occupied housing between June 2008 to June 2015 (Figure 2). The growth in lending for investment housing has slowed down since then compared to growth in lending for ownership housing. Motivation for investment housing is the ‘negative gearing’ policy which has kept the real cost of debt low and attractive compared to other investment opportunities as the loss between rent and interest payment can be offset against income for tax purposes. Generally, these loans are interest only variable rate loans.Table 2: Growth in lending for owner-occupied and investment-owned housing loan values Two recent prudential regulatory benchmarks with the objective of reducing pressure on house prices caused by investment housing that have impacted mortgage lenders are: introduction in December 2014 of an annual 10 per cent growth benchmark for residential mortgage lending to investors and in March 2017, introduction of a 30 per cent benchmark on the share of new residential mortgage lending to borrowers making interest only repayments by Australian Prudential Regulatory Authority (APRA). The consequence of these regulations has been that banks have reduced discretionary discounts, reduced incentives for investor borrowers, removed fee waivers on some residential mortgages for investor borrowers, tightened loan approval criteria and increased headline interest rates.During 2015-16, APRA introduced two new prudential requirements for measurement of liquidity risk. The first, Liquidity Coverage ratio, introduced in January 2015 requires banks to hold enough good quality liquid assets to cover expected cash outflows during a 30-day stress scenario of depositors being more likely to call on their funds, and when rolling over wholesale debt is difficult (ACCC, 2018). The second, introduced in December 2016 (effective January 2018), is the Net Stable Funding Ratio requiring banks to fund their lending activities through more stable sources of funding (ibid).On 24th June 2017 Major Bank Levy Act came into effect. The Act imposes a levy of 0.015 per cent of the applicable liabilities each quarter of banks whole liabilities exceed $100 billion in any quarter.The brief institutional discussion above illustrates key features of Australian mortgage industry – lack of competition, opaque pricing of mortgages, high debt to income ratio, a high share of investor housing. This forms an important backdrop to investigate pricing of mortgages in Australia.Construction of Equivalent Mortgage RatesBasic IdeaTo construct equivalent variable-rate mortgage rates for Australia, the following procedure is used. First, we obtain monthly data on fixed-rate mortgage yields in Australia. Then direct adjustments are made to these yield series to convert them into a variable-rate loan. These include the impact of call risk and interest rate risk. To make the adjustment for maturity, we use a 3-year interest rate swap. An interest rate swap is simply an exchange of one set of cash flows (e.g., a floating-rate payment) for another (e.g., a fixed-rate payment). We assume that the Australian homeowner pays the floating amount of interest in the swap agreement and receives at the same intervals a fixed amount of interest on some notional principal (i.e., the mortgage amount). The homeowner uses the fixed-rate payments to make the payment on his or her 3-year fixed-rate mortgage. Hence, on net the homeowner is left with a single obligation to pay a short-term variable interest rate.To make the adjustment for the call option in the 3-year fixed-rate mortgage, we use a 2-year call swaption (an option to receive fixed and pay floating on a swap). Economically, a 3-year, prepayable fixed-rate mortgage in Australia is equivalent to a variable-rate mortgage plus an interest rate swap and a swaption. Here the swaption is an option to receive the fixed rate specified in the swaption while paying a floating rate (“receiver” swaption). We choose year 2 because it is the middle year of the 3-year fixed-rate mortgage. If swap rates have fallen, the homeowner would exercise the swaption and elect to receive the agreed-upon fixed rate specified in the swaption and pay the floating rate on the swap. Simultaneously, the homeowner would enter into a short position (pay fixed, receive floating) in a plain vanilla interest-rate swap based on the market. The floating rates would cancel out and the homeowner would end up receiving an interest rate spread between the agreed-upon fixed rate specified in the swaption minus the fixed rate paid in the plain vanilla interest-rate swap based on the market. Hence, this call option essentially allows the homeowner to effectuate an interest rate savings equal to the difference between the fixed interest-rate on the 2-year call swaption and the fixed interest-rate based on the market, thereby replicating the option to refinance the 3-year fixed-rate mortgage at a lower interest rate, albeit only at time t = 2 years. Should, of course, swap rates rise over this time period, the swaption would expire unexercised, and the homeowner would simply continue to pay the contract rate specified in his or her 3-year fixed-rate mortgage. The price of this option to refinance is the premium paid by the holder of the call swaption. To illustrate the effect of these adjustments, let rm denote the actual interest rate on a 3-year fixed-rate mortgage in Australia. Similarly, let i3 denote the annual fixed swap rate on a 3-year contract and sw the price of a 2-year swaption. Using this notation, the yield on a synthetic variable-rate mortgage in Australia can be decomposed into three terms as follows:y=i1+sprd-sw(1)where sprd=rm-i3 is the interest rate spread on a 3-year fixed-rate mortgage in Australia. The term i1 is the opportunity cost of the lender’s money for 1-year at time t. The term sprd can be interpreted as a risk premium. This risk premium should vary, according to standard option pricing theory, with the borrower’s option to buy back or call the mortgage at par and the option to sell or put the hose to the lender at a price equal to value of the mortgage. The term sw is the added interest rate that a 3-year fixed-rate mortgage borrower pays for the privilege to refinance the loan at the end of year 2. To construct synthetic estimates of y, there must be estimates of rm, i3, i1, and sw. The sources of these data are as follows. The 3-year fixed mortgage rates for Australia are actual 3-year bank lending rates on owner-occupier housing loans as reported by the Reserve Bank of Australia from their monthly survey of rates on bank loan rates. The interest rate swap data i3 are taken from Bloomberg. The 3-year swap rate is equal to the 3-year government bond rate plus a swap spread. The opportunity cost of the lender’s money for 1-year is not available throughout the time period which we considered. It was therefore necessary to use a proxy for i1. The proxy we chose was the 2-year government bond rate. The swaption premia, sw, for Australia are collected from Bloomberg. The sample is restricted to the following time periods: June 2008 to December 2018. Illustrative CalculationsA simple example serves to illustrate the construction of equivalent variable mortgage rates. Suppose the following situation exists: An Australian household is considering taking out a prepayable mortgage on a fixed basis for 3 years. The mortgage has an amortization period of 30 years. The interest rate on this mortgage can be compared to the interest rate on an equivalent variable-rate mortgage.Here are the two steps that the Australian homeowner would need to undertake to construct the interest rate on an equivalent variable-rate mortgage.Step 1: The household enters into a 3-year swap agreement. The household agrees to pay a short-term rate of interest, i1, to a counterparty. In return, the counterparty agrees to pay a long-term rate (3-year fixed) of interest, i3, to the household.Step 2: The household writes a 2-year call swaption for sw. The swaption would allow the counterparty to realize an interest rate savings equal to the difference between the agreed-upon fixed interest-rate on the 2-year call swaption and the fixed interest-rate based on the market, thereby replicating the option to refinance the 3-year fixed-rate mortgage at a lower interest rate with the following proviso: The call swaption is exercisable only at time t = 2 years. In turn, the counterparty agrees to pay a swaption premium of sw to the household. This swaption premium allows the household to offset the price paid for the prepayment option in the 3-year fixed mortgage.In this case, the net obligation of the household would bey=rm-i3-i1-sw=i1+rm-i3-sw=i1+sprd-sw(2)which shows that y is the borrowing rate on a synthetic variable-rate mortgage. For rm=4.15%, i1=1.99% (as measured by the 2-year government bond rate), i3=2.18% (as measured by a 3-year swap rate, which is equal to the 3-year government bond rate plus a swap spread), and sw=0.63% (which are actual values for March 2018), the value of y is 3.33%. With an actual rate of r=5.20% on a variable-rate mortgage in March 2018 (as reported by the Reserve Bank of Australia), the result for r-y=1.87%. Summary Measures of Cost DifferencesFigure 1 shows that the relative picture for actual variable-rate borrowing costs in Australia compared with the borrowing rate on a synthetic variable-rate mortgage. There appears to be quite a difference in relative borrowing costs over the subsample period July 2011-March 2018, in that the actual variable-rate mortgage interest rate is only slightly higher than the synthetic mortgage rate (without adjusting for the call swaption premium) for the subsample period December 1998-June 2011, but contrasts sharply with the synthetic mortgage rate (with and without adjusting for the call swaption premium) for the subsample period July 2011-March 2018. Overall, the actual variable-rate mortgage interest rate compared to the synthetic mortgage rate is higher over the subsample period July 2011-March 2018 by 134 basis points compared to 71 basis points during December 2008-June 2011. The highest differential between actual and synthetic borrowing costs over this subsample period is 218 basis points in October 2008; and the lowest is -45 basis points in January 2010.Interestingly enough, since June 2011 actual variable-rate borrowing costs appear to be much higher than synthetic rates (with and without adjusting for the call swaption premium) by between 134 and 224 basis points. The highest differential between actual and synthetic borrowing costs (with adjusting for the call swaption premium) over this subsample period is 326 basis points in November 2011; and the lowest (without adjusting for the call swaption premium) is 95 basis points in June 2015. In the remainder of this paper, we examine these differences between actual and synthetic borrowing costs to determine whether these differences reflect our inability to fully specify all the nonlinearities and interactions in the pricing of variable-rate mortgages (including borrower self-selection effects), as well as to measure the swap rates and swaptions precisely, or whether these differences reflect predatory lending practices which have cost consumers dearly.Variation in Actual Mortgage RatesTo explain variation in actual mortgage rates implied by changes in our synthetic mortgage rates, we estimate regressions in error-correction form (see Green (2000) and Tiwari and Moriizumi (2003)), relating actual mortgage rates to synthetic mortgage rates in the long-run, and using this difference at the beginning of the period to explain the adjustment in the actual mortgage rate back toward its long-run relationship with our synthetic mortgage rate during the period. We also allow for transitory shocks to changes in the actual mortgage rate arising from short-term changes in credit availability and loan underwriting standards. Changes in any of the independent variables in the model or a change in the error term in the regression are transitory shocks that have an immediate effect on the dependent variable but will eventually be offset, unless they ultimately affect the long-run co-integrated relation between the dependent and independent variable (Case et al., 2013).Specifically, our basic functional form is:Δrt=ω+αΔrt-1+βΔyt+πΔXt+γrt-1-yt-1+εt(3)where Xt represents a vector of exogenous credit availability and loan underwriting variables, εt is a random error term, and we anticipate that γ<0. In this equation, we control for a number of covariates, including the change in the net deposit flows in the retail banking sector, which should display a negative coefficient, as rising deposit flows would indicate greater relative supply in the mortgage market and thus lower interest rates on mortgages. We also control for innovations in mortgage market by including proportion of mortgage loans securitized. Two variants have been used – change in total loans securitized (domestic and international) as a proportion of housing loan outstanding and change in residential loan securitized in international markets as a percent of total residential loans securitized. Transfer pricing policy regulations changed in July 2016, which negatively impacted the volume of international loan securitization. Securitization would reduce risk of lenders and lower the interest rate on mortgages. Reduced securitization would have precisely reverse impact on interest rates. Relative changes in default rates will affect risk associated with mortgage lending. In pricing loans, lenders would require higher risk premium with rising default rates and vice versa. A brief description of covariates and their construct is described below:Credit availability (Net deposit flows in retail banking sector): Following Jaffee and Rosen (1979), we measure credit availability as the flow of deposits into banks, ΔDEP, deflated by aggregate dollar value of single-family housing starts.?To construct a series on house price for a single family house, we use the gross household income from the Australian Bureau of Statistics for December 2017 and multiply this by a typical house-price-to-income ratio of 6.5 to obtain house price. Weighted average house price index for 8 cities in Australia as reported by Australian Bureau of Statistics was then used to rescale this to contract backwards the series of single family house prices for the period September 2008 to December 2017.? Aggregate dollar value of single-family housing starts is obtained by multiplying single family house price by the housing start. Credit availability series – defined as the ratio of (real) deposit flows (using data from Reserve Bank of Australia) divided by dollar value of single family housing starts.? The value of ΔDEP is actual minus expected growth in deposits, derived from using a linear procedure to actual deposits.? An alternative, the difference DEP and DEP-1 (lag) was also tried as a measure of ΔDEP.? However, the most compelling results are obtained when ΔDEP is actual minus expected growth in deposits. Hence the measure of net deposit flows in retail banking sector that is used in this paper is actual minus expected growth in deposits.Percent international securitization: A defining feature of securitization in Australia has been that a significant proportion of residential mortgage backed securities (RMBS) were issues to international investors. This allowed the banks to keep the cost of funding low. In September 2008, around the time when the global financial crisis unfolded, around 47 percent of Australian RMBS were sold to international investors. The share steadily declined to 3 percent in December 2017. At the time when changes to the transfer pricing policy in July 2016 were introduced, the international issuance of RMBS was merely 3 percent of the volume of RMBS issued. This paper uses Macquarie Bank’s series on RMBS outstanding to construct percent international securitization variable, which is defined as international RMBS outstanding to total outstanding RMBS in each period. In another variation, percent international securitization variable is interacted with dummies for before and after introduction of transfer pricing policy in July 2016. Delinquency rate: The delinquency rate used in this paper is the proportion of on RMBS loan pool that is delinquent. Three rates are reported in the data compiled by Macquarie Bank – 30-day delinquencies, 60-day delinquencies and 90-day delinquencies. This paper uses 90-day delinquencies as the covariate. With increasing delinquencies, the lending rates are expected to rise.Percent residential mortgages securitized: Another variable that can have a short-term impact on mortgage interest rates is percent of residential mortgages that are securitized. Higher percent of loans sold as RMBS is expected to lower the interest rate.Loans for investment housing as percent of total mortgage lending: A rising share of loans for investment housing in lenders’ residential loan portfolio is viewed as increase in risk. Generally, loans for investment housing are interest only mortgages and variable rate loans. With increase in share of investment housing loans, the interest rates would be expected to rise. There were two major regulatory directions by APRA, as discussed earlier, to discourage lending for investment housing, one in December 2014 and the other in March 2017. Our model uses three versions of share of loans for investment housing in total lending. One, a continuous variable, as it is, second distinguishing the effect of December 2014 guidelines on share of investment housing loans by interacting the share of investment housing loans with a dummy capturing the pre and post December 2014 period and third, distinguishing the effect of March 2017 effect by interacting share of investment housing loans with dummy for pre and post March 2017 period.Results and DiscussionThe first step is to conduct unit root tests on time series variables included in our model. Appendix 1 presents Augmented Dickey Fuller (ADF) test results for presence of unit roots under three forms – (i) intercept, no trend (ii) trend and intercept and (iii) no trend, no intercept. The results indicate that we can reject null hypothesis of ‘no unit root’ and the series are non-stationary. A Johansen cointegration test suggests (Appendix 1) that these series are first-order integrated. Given these test results, we can run error correction model. Results from error correction model for various alternative specifications are presented in Table 1. Estimated error correction model (equation 3) includes variables to control for credit rationing/credit availability, changes in underwriting standards, self-selection effects between fixed and variable rate mortgages etc. Table 1 presents four versions of estimated model. Model 1 includes exogenous shock variable – change in credit availability. Model 2, adds another variable – change in international RMBS outstanding to total RMBS outstanding. International RMBS issuance has served two purposes; one it has increased profitability of banks particularly before July 2016 when transfer pricing policy were tightened and second, in keeping the cost of funds low, which in turn helped in lowering mortgage interest rates. Model 2 also includes a variable for change in delinquency rate. In the absence of delinquency rate for the whole of mortgage pool, this papers uses the proportion of loan in RMBS pool that is delinquent for 90 days. A sudden rise in delinquency will cause lenders in increase the risk premium that they will include in the interest rates for lending for housing. Model 3 includes another variable – changes in the share of investment housing loan in total housing loans. One would expect that, given that investment housing loans are riskier as they are largely variable rate, interest only loans, lenders would charge a higher interest for such loans. In increase in these loans, is expected to increase the mortgage rate. In the final version of the model included in this paper, we separate the effect of changes in the proportion of international RMBS issuance before and after the imposition of new transfer pricing policy (TPP) in July 2016 by interacting the proportion of international RMBS with a dummy capturing pre and post 2016 TPP. In addition, to these models a number of other models with variables such as interactive variable of dummy for pre and post December 2014 APRA restrictions on banks for lending to investor housing with proportion of lending for investment housing; and interactive variable of dummy for March 2017 APRA restrictions on lending for investment housing with proportion of lending for investment housing, were tried. The results from these have not been included here as these variables were insignificant in error correction model. Two possible reasons for these variables to not show significant impact on interest rates – one, the effect of these restrictions on banks’ lending for investment housing was negligible, which, in fact, was confirmed during ACCC (2018) Royal Banking Commission inquiry and second, any such shocks are ultimately confirmed by changes in synthetic mortgage rates. We did not include a variable to capture the impact of mortgage levy as there has not much time lapsed to enable us to examine their impact. Table 2: Error correction modelDependent variable: Change in actual mortgage rateVariablesModel 1Model 2Model 3Model 4ΔSynthetic mortgage rate 0.62 (7.52)0.64 (7.78)0.62 (7.44)0.64 (7.67)ΔShare of international RMBS in total outstanding RMBS-3.60 (-1.66)-3.22 (-1.5)ΔDELIN900.116 (1.21)0.08(0.8)0.11(1.08)ΔCredit availability-2.42E-7 (-1.8)-3.67E-7 (-2.54)-3.07E-7 (-2.0)-3.60E-7 (-2.44)ΔShare of loans for investment housing in total loans-8.03 (-1.14)Dummy for After TPP x Δinternational RMBS in total outstanding RMBS -18.51(-0.68)Dummy for Before TPP x ΔInernational RMBS in total RMBS outstanding-3.65 (-1.7)Lagged actual minus synthetic mortgage rate-0.35 (-3.8)-0.30 (-3.23)-0.30 (-3.3)-0.31 (-3.2)Constant0.34 (3.14)0.23 (1.82)0.24 (1.9)0.23 (1.8)R squared0.760.790.800.79The coefficient on the lagged actual minus synthetic mortgage rate which corresponds to γ in equation (3), is highly significant in all variants of the estimated error correction model. The negative sign of γ is as expected. This suggests that ceteris paribus, the actual mortgage rate should converge to its equilibrium level barring temporary shocks. The estimated coefficient for change in synthetic mortgage rate is between 0.62-0.64 suggesting that if synthetic mortgage rates suddenly change, the actual mortgage rates change by 62-64 percent of that change instantaneously. The coefficient of lagged actual minus synthetic mortgage rate variable suggests that the gap between actual and equilibrium mortgage rate decreases each period by 0.30 to 0.35 across various model specifications. To illustrate the distinction, if synthetic mortgage rates were to increase by 100 basis points and then return to its previous level, then actual mortgage rates in the first period would increase by 62-64 basis points. After the second period actual mortgage rate would begin to decrease gradually and converge to its initial level. In contrast, if the shock to synthetic mortgage rate is permanent, then actual mortgage rate slowly converges to a value that exceeds the initial actual mortgage rate by 100 basis points, as given by the long run equilibrium relationship. The negative sign on lagged actual minus synthetic mortgage rate indicates that as the actual mortgage rates move away from equilibrium, there is an adjustment back towards the equilibrium.Error Correction Model provides insights into test for unbiasedness hypothesis (Fama, 1984). Unbiasedness would imply that the long-run relation is Actual mortgage rate = a + b Synthetic mortgage rate, where a = 0 and b = 1.0. If the lenders expect the synthetic mortgage rate to be the same as actual mortgage rate then they will be indifferent between the two under certain conditions. These conditions are (i) agents are risk neutral and (ii) there are identical transaction costs associated with fixed rate and variable rate mortgages. The only difference between the fixed rate and variable rate mortgages is the value of interest rate option. However, with Δ Actual mortgage rate= 0 and Δ Synthetic mortgage rate = 0, the above would suggest that the long-run relationship between Actual mortgage rate and synthetic mortgage rate (since actual and synthetic rates are cointegrated) is given by Actual mortgage rate = 0.34/0.35 + 1.0 Synthetic mortgage rate = 0.97 + 1.0 Synthetic mortgage rate (coefficient of a estimated from Model 1, Table 1). A strong form – unbiasedness – market efficiency hypothesis and no risk premium, implies a=0, b=1. A non-zero ‘a’ implies risk premium. Testing the unbiasedness-market efficiency hypothesis requires testing the null hypothesis that γ=1 in equation 3. The null would imply that there is no systematic time-varying component to prediction error. However, as in Table 1, γ is negative. This implies that risk premium separates the synthetic mortgage rate from actual mortgage rate and synthetic mortgage rate is not unbiased predictor of actual mortgage rate. Equation 3 includes other variables changes to which cause transitory shock to cointegrated relation between actual mortgage rates and synthetic mortgage rates. The results suggest that credit availability explains a portion of the shocks -- innovations -- in the actual mortgage interest rate. Rising deposit flows or declining housing starts indicate greater relative supply in the mortgage market and thus lower interest rates on mortgages. The negative sign for credit availability in Table 1 confirms that rising deposit flows explains a portion of the shock to the actual mortgage interest rate by having a negative impact.? Other shocks such as changes in the share of international RMBS to total RMBS outstanding, delinquency rates have expected negative signs but they are not significant. Even the interactive variable of TPP dummy and share of international RMBS to total RMBS outstanding is not significant, though the signs are as expected. Surprisingly the sign of changes in the share of investment housing loans in total is negative. However, if we dig deeper, it is not surprising. Between 2015 and December 2017, banks started to quote different headline variable rates for owner-occupied and investor-owned housing loans. This was further segmented by interest only and repayment type mortgages. The headline interest rates for repayment type owner-occupied mortgages have declined significantly, while the interest on other three (interest only and repayment type loans for investor-owned housing, and interest only loans for owner-occupied loans) have increased (ACCC, 2018). The complex interplay of these structures and banks discretion on offering discounts on headline rates to borrowers who are important for business and in whom banks have long-term interest. Consequently many of the investment housing borrowers are able to secure discounts which are better than first time owner occupied housing borrowers. ACCC (2018) also reveals that due to lack of competition among banks (or tacit collusion), banks have been reluctant to change rates that are out of line with other banks or make first move making APRA regulatory directions ineffective.The inclusion of these transitory shocks in the ECM reduces the risk premia in actual mortgage rate over synthetic mortgage rate slightly but not a lot. ‘a’ ranges between 0.73 (Model 4) to 0.97 (Model 1).Is Royal Commission justified to intervene?In summary, what these results suggest is that even after specifying all the nonlinearities and interactions in the pricing of variable-rate mortgages, as well as to measure the swap rates and swaptions precisely, a large part of the difference between actual and synthetic rate mortgages (aka risk premium that lenders charge) remain unexplained. The failure of unbiasedness hypothesis indicates that there is a premium that separates actual and synthetic mortgage rates. Shocks for default, liquidity and regulatory restrictions have been insignificant in explaining this premium. The results indicate that the mortgage interest rates are way above what perfect capital market would have warranted. This suggests possibilities of predatory institutional practices and anomalies in mortgage pricing which justifies Royal Commission intervention.ConclusionThis paper investigates the mortgage pricing practices of lenders in Australia, in particular, whether the interest rates that they charge on single family residential loans are in line with what perfect capital markets would have warranted. The institutional context with in which this question is being examined is the perception that the Australian mortgage industry lacks competition, pursues opaque pricing of mortgages, households are highly leveraged resulting in lenders market and there is a high share of lending for investment housing in total lending. There have also been regulatory changes (such as APRA restrictions on lending for investment housing and on interest only loans, imposition of a mortgage levy to limit oligopolistic behavior of large four banks, changes in transfer pricing policy which has constrained international RMBS issuance) that have rationed credit. There has also been growth in deposit flows, which has led to increase in credit availability.In an innovative method, by exploiting the conceptual difference in risks associated with fixed rate and variable rate mortgages for lenders, we construct a synthetic variable rate. Synthetic variables rate are equivalent variable rate mortgage rates obtained from 3 year fixed rate by adjusting these for interest rate risks premium and call options that are embedded in fixed rates. We use 3 year swaps and 2 year swaption rates to make these adjustments. The difference between actual and synthetic rates is higher over the subsample period July 2011-March 2018 by 134 basis points compared to 71 basis points during December 2008-June 2011. An error correction model has been estimated to explain the variation between actual and synthetic mortgage rates. The results reject unbiasedness hypothesis. Despite accounting for various variables for shocks in short term, the risk premia that lenders have charged over synthetic mortgage rate is between 73 to 97 basis point. These results reveal that the mortgage lenders with concentration of four large banks, who have collective share of more than 80 percent of mortgage lending have followed non-transparent pricing policies. ReferencesACCC (2018), Residential Mortgage Price Inquiry Interim Report, Australian Consumer and Competition Commission.Case, K., Quigley, J., and Shiller, R. (2013), Wealth Effect Revisited, NBER Working Paper 18667, National Bureau of Economic Research, Cambridge, MA, , E. (1984), Forward and Spot Exchange Rates, Journal of Monetary Economics, 14, pp. 319-338.Green, W. H. (2000), Econometric Analysis, 4th Edition, Englewood Cliffs, NJ, Prentice Hall.Jaffe, D. M. and Rosen, S. (1979), Mortgage Credit Availability and Residential Construction, Brookings Papers on Economic Activity 2, pp. 333-384.Pascoe, M. (2018), Forget the Price Drop, This is the Real Housing Crisis, The News Daily, 12 November 2018.Tiwari, P. and Moriizumi, Y. (2003), Efficiency in Housing Finance: A Comparative Study of Mortgage Instruments in Japan, European Journal of Housing Policy, 3(3), pp. 267-288. Appendix 1: Augmented Dickey Fuller unit root and Johansen cointegration testPanel 1: ADF Test statisticsVariableTest statisticCritical Value (5%)Actual mortgage rateIntercept, No trend-3.23-2.96Intercept, Trend-3.49-3.55No intercept, no trend-1.89-1.95Synthetic mortgage rateIntercept, No trend-2.14-2.96Intercept, Trend-2.77-3.55No intercept, no trend-1.57-1.95Credit availabilityIntercept, No trend-1.75-2.96Intercept, Trend-1.38-3.55No intercept, no trend1.49-1.95Delinquency rateIntercept, No trend-1.35-2.96Intercept, Trend-1.41-3.55No intercept, no trend0.13-1.95Panel 2: Johansen Cointegration Test (Actual meeting rate, synthetic meeting rate, credit availability, delinquency rate)RankTrace statisticsCritical Value (5%)068.1147.21126.5129.68Maximum statisticsCritical Value (5%)041.6027.07120.5820.97 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download