ArticlePDF Available

Mortgage Product Diversity: Responding to Consumer Demand or Protecting Lender Profit? An Asymmetric Panel Analysis

Authors:

Abstract

This paper explores determinants of mortgage product diversity for owner occupied and investment loans in the Australian housing mortgage market. From 2001-2012, 65 lenders introduced 1,220 mortgage products in Australia. We examine whether the product proliferation was a result of consumer demand, or a response to pressure to lower lending rates. We find that consumer demand for mortgages does not have a significant relationship with the number of mortgage products, but that decreases in the policy interest rate are highly significant as an explanatory variable for product proliferation. Such behaviour is consistent with information obfuscation, reducing the ease with which consumers can compare lending rates. Further, the relationship between mortgage products offered and the policy interest rate is asymmetric: decreases in the cash rate are associated with increased mortgage products offered, but increases in the cash rate have a more muted effect on decreasing the number of products.
1
Mortgage Product Diversity: Responding to Consumer Demand or Protecting Lender
Profit? An Asymmetric Panel Analysis
Jeremy Nguyen(a), Abbas Valadkhani(a),*, and Russell Smyth(b)
(a) Swinburne Business School, Swinburne University of Technology, Victoria 3122, Australia
(b) Department of Economics, Monash Business School, Monash University, Victoria 3800, Australia
This version: accepted for publication in Applied Economics, March 2018.
This paper explores determinants of mortgage product diversity for owner
occupied and investment loans in the Australian housing mortgage market. From
2001-2012, 65 lenders introduced 1,220 mortgage products in Australia. We
examine whether the product proliferation was a result of consumer demand, or
a response to pressure to lower lending rates. We find that consumer demand for
mortgages does not have a significant relationship with the number of mortgage
products, but that decreases in the policy interest rate are highly significant as an
explanatory variable for product proliferation. Such behaviour is consistent with
information obfuscation, reducing the ease with which consumers can compare
lending rates. Further, the relationship between mortgage products offered and
the policy interest rate is asymmetric: decreases in the cash rate are associated
with increased mortgage products offered, but increases in the cash rate have a
more muted effect on decreasing the number of products.
Keywords: Mortgages; interest rates; product differentiation; price competition;
Australia.
* Corresponding author
Professor Abbas Valadkhani
Swinburne Business School
Swinburne University of Technology
Victoria 3122 Australia
Tel: +61 3 9214 8791
E-mail: abbas@swin.edu.au
2
Mortgage Product Diversity: Responding to Consumer Demand or Protecting Lender
Profit? An Asymmetric Panel Analysis*
This paper explores determinants of mortgage product diversity for owner
occupied and investment loans in the Australian housing mortgage market. From
2001-2012, 65 lenders introduced 1,220 mortgage products in Australia. We
examine whether the product proliferation was a result of consumer demand, or
a response to pressure to lower lending rates. We find that consumer demand for
mortgages does not have a significant relationship with the number of mortgage
products, but that decreases in the policy interest rate are highly significant as an
explanatory variable for product proliferation. Such behaviour is consistent with
information obfuscation, reducing the ease with which consumers can compare
lending rates. Further, the relationship between mortgage products offered and
the policy interest rate is asymmetric: decreases in the cash rate are associated
with increased mortgage products offered, but increases in the cash rate have a
more muted effect on decreasing the number of products.
Keywords: Mortgages; interest rates; product differentiation; price competition;
Australia.
Introduction
Mortgages are typically complex product bundles with characteristics such as varying initial
discount rates, setup fees, redraw facilities and self-certification of income. As a result, it can
be difficult for borrowers to compare, and contrast, competing mortgages in making optimal,
transparent selections (Ward, 2009). The proliferation of a wide variety of mortgage products
adds to the complexity of the home loan market. Bucks and Pence (2006) suggest that due to
difficulties associated with collecting and processing information about different types of
mortgages, many borrowers do not fully understand all of the terms and conditions embedded
in loan contracts. They also posit that neither lenders nor savvy households have an incentive
to design or advocate for products that are more transparent, and competition does not drive
out confusing or misleading products (Bucks and Pence, 2006, p.232).
* We thank an anonymous referee, whose useful feedback considerably improved an earlier version of this
article. The usual caveat applies.
3
Our motivation for this paper comes from the proliferation of mortgage products in
Australia. Lenders in this market have certainly not been shy about adopting a product
proliferation strategy. During the sample period (2001-2012), 65 lenders in Australia
introduced 1,220 mortgage products with names such as Mortgage eliminator, Gold variable,
Standard variable, Investment variable, Homebuyer power, Come & Go, Portfolio loan,
Reward rate and Investment advantage, each with a different interest rate. These products also
often have other features to attract borrowers and to set themselves apart.
In principle, firms generally have incentives to create new products that ‘fill the gaps
between other firms existing products: this tends to add to informational asymmetry and
lessens the pressure to compete on price (Klemperer, 1992). As Brander and Eaton (1984)
note, in situations ‘when production of a range of potential products is limited to a group of
competing firms that are established in the market, each firm is likely to seek to develop
products that are close substitutes for what it currently produces, since joint production of
these products will lead to less intense price and output competition at a later stage’.
On the other hand, product proliferation may certainly be a response to consumer
demand. The notion that increasing returns to scale foster market concentration, while product
differentiation and heterogeneity in consumer preferences promotes competition, is well
established (Dixit and Stiglitz, 1977; Krugman, 1980).
In view of the divide in the literature, it is naturally of interest to investigate whether
Australian lenders have used greater product diversity to decrease price competition (i.e. to
avoid competing via lower interest rates), or whether greater product diversity has been driven
by consumer demand (or perhaps some combination of both). Our analysis of Australian data
suggests that lenders’ use of product proliferation is not significantly associated with changes
in consumer lending demand, and is significantly associated with pressures for lenders to
lower their lending rates (i.e. changes in the policy interest rate).
Australia’s mortgage market provides an interesting context in which to undertake a
study of this type, for several reasons. First, as outlined above, there has been a substantial
increase in product diversity in this market. Second, there is a perception that the level of
competition is low: the banking sector in Australia is highly concentrated, with the four
largest banks accounting for 60% of all owner-occupied home mortgage approvals (Davis,
2011). Following the Global Financial Crisis (GFC), these four banks were able to raise
interest rates on mortgage products by more than the increase in the policy (‘cash’) rate, and
4
these banks have been slow in passing on cuts in the cash rate to housing loan borrowers (Lim
et al., 2013). While this asymmetry is consistent with the response of other Australian interest
rates (Smales, 2012; Valadkhani et al, 2013), and of mortgage rates internationally (Månsson,
2013), it has fuelled widespread public anger and resulted in a parliamentary inquiry into
competition in the banking sector (Neal, 2011).
Third, in contrast to the United States where most borrowers take up mortgages with
rates fixed for long periods, the Australian mortgage market is dominated by variable-rate
mortgages (VRMs) (Davis, 2011; Stewart et al., 2013). VRMs are typically written for a term
of up to 30 years and, in contrast with most other countries, the interest rate on VRMs in
Australia is adjustable at the discretion of the banks. Borrowers face a choice between VRMs
and a range of other mortgage products, such as ‘honeymoon’ mortgages (which offer a
discount on the variable interest rate for an initial fixed period) and mortgages with fixed
interest rates for a short period (three-five years) (Stewart et al., 2013).
Home ownership in Australia has traditionally been very high and it constitutes the
largest component of household wealth (Windsor et al., 2015). However, Australia has
experienced a boom in housing prices and housing affordability is now at a record low.
Australians are among the most indebted households in the OECD, with high levels of debt
gearing and service (Valadkhani, 2014). Housing affordability (or lack thereof) was one of the
most important issues for voters in the 2013 general election (Yates and Berry, 2011). In such
a politically charged environment, the issue of competition and effective dissemination of
product information have become highly topical. The transparent pricing of mortgage
products can, thus, provide many tangible benefits to homebuyers, borrowers and investors,
financial intermediaries and marketers, the construction industry, the capital market and real
estate developers as well as promoting the efficient allocation of resources (Renaud, 2005;
Hillebrand and Koray, 2008; Valadkhani, 2014).
This study seeks to understand the relationship between, on one hand, the number and
diversity of mortgage products and, on the other, consumer lending demand and the policy
interest rate. We examine the effects of consumer demand and the policy interest rate on the
number of mortgage products from each lender, using asymmetric panel models. To
investigate further the issue of whether product proliferation has been used to reduce
pressures on lenders to reduce interest rates on their mortgage products, we also examine the
effects of consumer demand and the policy interest rate on average and maximum interest
5
rates charged by each lender. Our sample includes 65 individual bank and non-bank lenders in
Australia during the period 2001-2012.
We find no significant relationship between the variety of mortgage products offered and
consumer demand. We find a negative and significant relationship between the policy interest
rate and the variety of mortgage products offered by individual lenders. These findings
suggest that product proliferation has been driven not by increased demand from consumers,
but has been employed to reduce competitive pressure on lenders. We also find that the effect
of the policy interest rate on the number of mortgage products offered has an asymmetric
effect: cuts in the cash rate have a coefficient greater in magnitude than do increases in the
cash rate. Decreases in the cash rate are significantly associated with increased product
proliferation, while increases in the cash rate are associated with a smaller effect on reducing
proliferation.
The findings are of general importance to the wider public, given that there is
considerable interest in the implications of diversity in mortgage product choice. This interest
reflects, in part, the role that the mortgage market played in the United States in precipitating
the GFC (Crotty, 2009) and continues to exist in many developed economies in the
transmission of monetary policy (Calza et al., 2013; Lim et al., 2013).
This paper contributes to the literature on the relationship between product differentiation
and competition. As previously noted, there is a strand of the literature that notes that product
proliferation can be used to decrease the pressure to compete on price (Klemperer, 1992;
Brander and Eaton, 1984). Another strand focuses on product differentiation as a response to
consumer demand, and its role in promoting competition (Dixit and Stiglitz, 1977; Krugman,
1980). These ideas have been applied, for example, to explain why, and how, competition has
developed over time in highly concentrated markets, such as leading internet platforms
(Haucap and Heimeshoff, 2014). We extend this literature by examining the relationship
between the number, and range, of mortgage products and competition in the banking sector,
with the aim of assessing which strand of the literature is most applicable to the Australian
mortgage loan market. As such, we are the first to examine product differentiation as a
possible substitute for price competition in the context of mortgage lending.
A large literature examines the relationship between competition and interest rates for the
banking industry as a whole. This literature tests the structure conduct performance
hypothesis, which states that in more concentrated markets, banks engage in anticompetitive
6
behaviour, charging high interest rates and earning supranormal profits (Bain, 1956). Most
previous studies find a positive relationship between market concentration and interest rates in
the banking sector (see eg. Baquero et al., 2012 for a review). While we also examine the
relationship between competition and interest rates in the banking sector, we differ in that our
focus is on the number of mortgage products, rather than the number of banks.
Hypothesis development
We test three main hypotheses, which are related to the relationship between the number of
mortgage products offered by individual lenders and (1) demand for mortgage products, (2)
interest rates, and (3) asymmetric movements in interest rates.
We examine whether the increase in mortgage products offered resulted from presumed
heterogeneity in consumer preferences and increases in consumer demand. Our first
hypothesis is:
H1: The number of mortgage products offered by a lending institution is a function of
consumer demand for mortgages.
We also investigate an alternative strand of theory: if lenders perceive themselves to be
subject to increased competitive pressures, they will tend to respond by increasing product
diversity. Thus, a rise in any given lender’s product diversity would be expected to be
associated with downward pressure on that lender’s interest rates. Our second hypothesis is
accordingly:
H2: The number of mortgage products offered by a lending institution is a function of the
cost of funds (i.e. the policy interest rate), where a decrease in the cost of funds reflects
pressure on lenders to lower rates.
Further, the asymmetric response of banks to monetary policy tightening and easing is
well established (Valadkhani et al, 2013). We examine whether this asymmetry also holds
when investigating product proliferation:
H3: The quantity of mortgage products offered is asymmetric in its relationship with
changes in the policy interest rate. Falls in the policy interest rate have more of an effect on
product proliferation than do increases in the policy rate.
Empirical methodology
The order of integration of interest rates has important implications for financial modelling at
7
both theoretical and empirical levels. For example, if they contain unit roots, then unexpected
shocks will exert permanent effects on interest rates through time. On the other hand, if
interest rates are stationary, then the impacts of such shocks are more likely to be transitory.
Although the presence of a unit root in the nominal interest rate is well-documented in some
pure time series studies, DeJong et al. (1992) argue that such non-stationarity arises from the
existence of finite samples and the well-known low power of standard unit root tests, which
cannot distinguish between an I(1) and a near unit-root process.
To address the low power of standard time-series unit root tests in finite samples to reject the
null of a unit root against alternative hypotheses with highly persistent deviations from
equilibrium, we first perform five panel unit root tests proposed, respectively, by Levin, Lin
and Chu (LLC, 2002); Im, Pesaran and Shin (IPS, 2003); ADF-Fisher (Maddala and Wu,
1999; Choi, 2001); PP-Fisher (Maddala and Wu, 1999; Choi, 2001); and Pesaran (2007). The
LLC test assumes a common unit root process, whereas the IPS, ADF-Fisher and PP-Fisher
tests allow for individual unit root processes. A limitation of the LLC, IPS, ADF-Fisher and
PP-Fisher tests, however, is that they do not address cross-section dependence (CSD). Pesaran
(2007) takes account of the presence of CSD in testing the null of homogeneous non-
stationary series. As discussed below, the results support the view that all the three interest
rates are stationary; we thus proceed to discuss our dynamic panel models below.
Each of the M lenders (i=1,2,…, M) supplies Ni (the number of) different mortgage
products over time. We adopt the following three asymmetric panel regressions:
         
  
     (1)

         
  
  
   (2)

         
  
  
   (3)
where  represents the number of mortgage products offered by lender i at time t;
represents the policy interest rate set by the Reserve Bank of Australia (i.e. the cash rate) as a
proxy for the short-run funding cost with expected marginal negative effects (i.e. β and β+ <
0). It should be noted that the use of other alternative proxies for funding cost, such as the 90-
day bank bill interest rate, does not alter our results.; is a dummy variable equal to 1 if
has risen from the previous t (i.e. if the policy interest rate has risen);
represents the
constant dollar value of combined Owner Occupied and Investment mortgage loans approved
in time t, serving as a proxy for capturing the impacts of quantity of loans demanded. Note
8
that there is no subscript i for this variable, as
D
t
L
denotes the aggregate value of loans
demanded at time t from all banks, building societies, credit unions, mortgage companies and
other authorised deposit taking institutions. There is no consistently, and accurately, defined
data on
D
it
L
for each lender and each year during the sample period. We do not know exactly
the share of each of the 65 lenders in the total value of owner occupier and investor home
loans for each year. Had it been otherwise, we would estimate a separate demand function for
owner occupier and investor home loans for each lender.
The coefficient γ is assigned to the lagged dependent variable, capturing inertia
associated with changes in the number of mortgage products in equation (1), and inertia in the
relevant interest rates in equations (2) and (3), discussed below. Given the use of stationary
panel data and the fact that M is considerably larger than T, γ is expected to be positive and
below unity. The εit are the residual terms.
Modelling the interest rates for a large number of diverse products, designed for
borrowers of different characteristics and risk profiles, is not practical. Instead, we choose the
average and maximum interest rates of all mortgage products on offer from each lender,
providing a diverse pricing representation of their products. We define these representative
mortgage interest rates as follows:
A
it
R
=average interest rate of all mortgage products on offer by the ith lender at time t,
Max
it
R
=maximum interest rate of all mortgage products on offer by the ith lender at time t.
To estimate equations (1)-(3) in light of potential endogeneity and cross-sectional
dependence, we utilise Generalised Method of Moments (GMM), where the instruments are
specified as follows for each equation:
Equation (1):   
 
;
Equation (2):   
 


;
Equation (3):   
 
To test the validity of the above three sets of instruments, we report J-statistics for the null
hypothesis that the over-identifying restrictions are valid. We employ the GMM estimator:
GMM estimator=

 (4)
9
where X includes all explanatory variables on the right-hand side of equations 1-3, Z is a set
of K instruments, WT is the weighting matrix, and y represents the dependent variable. We
conduct two sets of tests. First, we examine if there are asymmetries in the response of the
number of mortgage products offered to increases in the cash rate, as opposed to decreases in
the cash rate:
=
Second, we test whether the pass through of the cash rate changes is equal to unity:
=1
=1
We accept that the ability, and willingness, of lenders to alter the number of mortgage
products and their interest rates are also influenced by their exposure to the overseas funding
mix, market power, reserves, the composition of on and off-balance sheet items and the extent
of securitization which could vary across individual lenders in each year. Obtaining such
consistent data is difficult without limiting N or T.
Data
According to the Australian Prudential Regulation Authority (APRA, 2015), as of November
2015 there were 159 authorised deposit-taking institutions (ADIs) in Australia with the
following composition: 26 Australian-owned banks, 7 foreign subsidiary banks, 44 branches
of foreign banks, 5 building societies, 73 credit unions and 7 other ADIs. The majority of
ADIs, plus a number of brokers and finance companies, offer home loans. For this study we
were able to extract mortgage interest rates for 65 lenders from the Canstar database, which
contains various types of mortgage interest rates, for each lender for the sample period 2001-
2012. The data extracted cover 24 banks, 5 building societies, 24 credit unions and 11
mortgage and finance companies.
1
The 65 lenders included in the sample constitute 95% of
the mortgage market in terms of mortgage loans over the studied period.
The chosen sample period contains years both prior, and subsequent to, the 2008 GFC,
yielding a total of 780 panel observations (12 years x 65 lenders). Table 1 presents the mean
and standard deviations of the variables employed in this study. During the sample period
1
The data were purchased from www.canstar.com.au. Canstar contain data for some lenders going back to 1992,
but the interest rate data are only available for the 65 lenders only from 2001. Starting the sample period prior
to 2001 would significantly curtail the number of lenders in the dataset.
10
(2001-2012) the RBA’s average cash rate was 5% and owner-occupier and investor mortgage
loans averaged $814 billion per annum. The 65 lenders offered home loan products with
varying diversity and interest rates. For example, ECU Australia offered, on average, only 2.6
mortgage products each year, with an annualized interest rate of 7.2%, (maximum 8.9%). By
contrast, Holiday Coast Credit Union offered, on average, 19 mortgage products, with an
average interest of 7.3% (maximum 8.8%). The lowest mortgage product variability is
observed for Unicredit-WA, BOQ, Bendigo Bank and Family First CU; and the highest
variability for Holiday Coast CU, Citibank, FCCS Credit Union and RAMS Home Loans.
TABLES 1-2 ABOUT HERE
As can be seen in Table 1, the following five lenders had the greatest average number of
products offered per year, over the period 2001-2012: Holiday Coast (18.8 products),
Citibank (15.8 products), FCCS Credit Union (12.5 products), RAMS Home Loans (11.8
products) and Aussie Home Loans (10.8 products). In contrast, the average number of
products offered in a particular year were lowest for the following five lenders: Bankstown
City CU (2.8 products), Hume Building Society (2.8 products), Police and Nurses Mutual
Bank (2.7 products), ECU Australia (2.6 products), and La Trobe Financial Services (2.3
products).
Prior to undertaking any empirical modelling, we ensure that there is a significant degree
of variability in the number of mortgage products across lenders. Comparing the first three
years (2001-2003) with the last three years (2010-2012) of the sample period shows that the
overwhelming majority of lenders increased the average number of mortgage products offered
(see Figure 1). Table 2 shows the descriptive statistics over time averaged across 65 lenders.
As can be seen from Tables 2-3, lenders offered 3.63 products on average during 2001-2003,
whereas in the post-GFC era, this figure more than doubled. Based on the four tests of
equality of means [i.e. t student, Satterthwaite (1946) t, Anova F, and Welch (1951) F tests]
reported in Table 3, such increases are statistically significant. Similarly, the four tests of
equality of variances [i.e. F, Bartlett, Levene F (Conover et al. 1981), and Brown and
Forsythe (1974) F tests] are consistently rejected, suggesting that the number of mortgage
products showed a greater degree of variability/diversity in the post- than in the pre-GFC era.
TABLE 3 AND FIGURE 1 ABOUT HERE
11
Results and policy implications
The results for the panel unit root tests are presented in Table 4. The LLC (2002) test, which
assumes a common unit root process in the null hypothesis, indicates that all four variables
are I(0). The same conclusion is reached when we assume an individual unit root process in
the following three tests: the W test (IPS, 2003), ADF-Fisher and PP-Fisher
2 tests (Maddala
and Wu, 1999). The last row of Table 4 reports the results of the Pesaran (2004) diagnostic
test for CSD. The results indicate that the null hypothesis of no CSD is rejected at the 1%
level of significance. Hence, we also report the Pesaran (2007) unit root test, which takes into
account the presence of CSD. The conclusions from the Pesaran (2007) test are consistent
with the first-generation tests, indicating that all four variables are I(0).
2
TABLE 4 ABOUT HERE.
Table 5 shows the results from estimating equations (1), (2) and (3) using GMM. The J-
statistics of over-identifying restrictions for all three models are not significant at any
conventional level, supporting the view that the chosen instruments are statistically valid. In
Model 1 in Table 5, neither
nor
1 are significant at conventional levels: the number of
mortgage products offered by lenders is not associated with the level of consumer demand for
loans. In Model 1, we also find that the estimated
+ and  coefficients are negative and
significant at 1%. The estimated
coefficients indicate that there is an inverse relationship
between the policy (cash) interest rate and the number of mortgage products that a lender
offers: when the official interest rate falls, lenders increase the number of products that they
offer. Further, the magnitude of the coefficient is larger when the changes in the cash rate are
negative, i.e. interest rates fall. A t-test rejects the null hypothesis that
+= at 1%: there is
asymmetry in the effect of cash rate decreases, compared to cash rate increases. The
significant
coefficient indicates that there is a significant degree of inertia in the number of
products offered and, at the same time, the coefficient still remains within the positive unity
circle. In short, product proliferation observed in the Australian mortgage market cannot be
attributed to consumer demand; rather, it is associated with increased pressure on lenders to
lower their interest rates.
TABLE 5 ABOUT HERE.
2
As to the cash rate and the aggregate demand for loans, one should note that 12 yearly observations are not
sufficient for conducting any type of unit root testing as they both have subscript t and not i. To address this
problem, we used all available monthly data (i.e. 1990m12-2015m9) and conducted the Lumsdaine and Papell
(1997) unit root test with two structural breaks for both variables. The results, available upon request, showed
that both macro-level variables were also stationary.
12
In Model 2 in Table 5, the estimated
+ and  coefficients are positive and significant at
1%. The cash rate is a highly significant explanatory variable for the average interest rate
offered across a lenders mortgage products. Hypothesis tests indicate that we can reject a
pass through rate of unity from the official cash rate to lenders average interest rates (i.e.
+=1 and =1) at 1%, and that we can reject the null hypothesis of symmetric responses to
increases and decreases in the cash rate (i.e.
+=) at 5%. As with Model 1, neither
nor
1
are significant at conventional levels: the level of consumer demand for loans is not
significant.
Results from Model 3 in Table 5, with a highly significant
+, but an insignificant
estimated , again are consistent with asymmetric interest rate behavior. Increases in the
cash rate are highly significant for the maximum interest rate that a lender offers, but
decreases in the cash rate are not significant for the maximum interest rate. As with Model 1,
we see significant inertia in the model (significant and positive
coefficient, within the unity
circle). It is not surprising to observe that the maximum interest rate (with relatively highest
risk) exhibits more resilience over time than the average interest rate.
Despite the fact that N=65 is larger than T=12, the three estimated models presented in
Table 5 perform quite well in terms of goodness-of-fit statistics with the minimum
(
=0.612) and maximum (
=0.733) being observed for Model 1 and Model 2, respectively.
Based on the estimated Durbins h, there is no evidence of serial correlation in any of the three
estimated models in Table 5.
Conclusion
In this paper, we examined whether the increase in the number of mortgage products in
Australia is associated with higher demand from Australian borrowers or increased pressure
on lenders to lower lending rates. We also examined if asymmetric relationships exist with
regards to official interest rate movements. To this end, we utilized publically unavailable
data (from Canstar) on various types of mortgage products offered by 65 major lenders during
the sample period 2001-2012, and the associated interest rates. We posited three hypotheses:
the number of mortgage products offered by lending institutions is not significantly associated
with increases in consumer demand; that there is a significant and negative relationship
between the number of mortgage products offered by a lending institution and the policy
interest rate; and that the relationship between the number of mortgage products and the
policy interest rate is asymmetric with respect to the direction of changes in the policy interest
13
rate. A multivariate, reduced-form VAR model was estimated for the number of mortgage
products offered, and two representative mortgage interest rates (i.e. average and maximum),
from each lender using GMM. As expected, the coefficient on the cash rate is negative and
statistically significant for both product proliferation and the average interest rate offered. One
can thus conclude that product proliferation in the mortgage loan market has not been demand
driven, but rather may have been employed by lenders to reduce price competition.
14
References
Australian Prudential Regulation Authority (APRA), (2015) List of Authorised Deposit-
taking Institutions, http://www.apra.gov.au/adi/Pages/adilist.aspx (accessed 4 October,
2015).
Bain, J. (1956) Barriers to New Competition. Harvard University Press, Cambridge, MA.
Baquero, G., Hamadi, M. & Heinen, A., (2012) Competition, loan rates and information
dispersion in microcredit markets. ESMT Working Paper.
Brander, J. A., & Eaton, J. (1984) Product line rivalry. American Economic Review, 74(3),
323-334.
Brown, M. B. & Forsythe, A. B. (1974) Robust tests for the equality of variances. Journal of
the American Statistical Association, 69(346), pp. 364-367.
Bucks, B. & Pence, K.M. (2006) Do homeowners know their house values and mortgage
terms? Divisions of Research & Statistics and Monetary Affairs, Federal Reserve Board.
Calza, A., Monacelli, T. & Stracca, L. (2013) Housing finance and monetary policy. Journal
of the European Economic Association, 11, pp. 101-122.
Choi, I. (2001) Unit root tests for panel data. Journal of International Money and Finance, 20,
pp. 249272.
Conover, W. J., Johnson, M. E. & Johnson, M.M. (1981) A comparative study of tests for
homogeneity of variances, with applications to the outer continental shelf bidding data.
Technometrics, 23(4), pp. 351-361.
Crotty, J. (2009) Structural causes of the global financial crisis: a critical assessment of the
‘new financial architecture’. Cambridge Journal of Economics, 33(4), pp. 563-580.
Davis, K. (2011) The Australian financial system in the 2000s: dodging the bullet. In H.
Geard and J. Kearns (Eds.) The Australian Economy in the 2000s, Reserve Bank of
Australia: Sydney, pp.313-314.
DeJong, D.N., Nankervis, J., Savin, N.E. & Whiteman, C.H. (1992) The power problems of
unit root tests in time series with autoregressive errors. Journal of Econometrics, 53, pp.
32343.
Dixit, A.K. & Stiglitz, J.E. (1977) Monopolistic competition and optimum product diversity.
American Economic Review, 67, pp. 297-308.
15
Haucap, J. and Heimeshoff, U. (2014) Google, Facebook, Amazon, eBay: Is the Internet
driving competition or market monopolization? International Economics and Economic
Policy, 11(1-2), pp. 49-61.
Hillebrand, E. and Koray, F., (2008). Interest rate volatility and home mortgage loans.
Applied Economics, 40(18), pp.2381-2385.
Im, K. S., Pesaran, M. H. & Shin, Y. (2003) Testing for unit roots in heterogeneous panels.
Journal of Econometrics, 115(1), pp. 53-74.
Klemperer, P. (1992). Equilibrium product lines: competing head-to-head may be less
competitive. American Economic Review, 740-755.
Krugman, P. (1980) Scale economies, product differentiation, and the pattern of trade.
American Economic Review, 70, pp. 950-959.
Lim, G.C., Tsiaplias, S. & Chua, C.L. (2013) Bank and official interest rates: how do they
interact over time? Economic Record, 89(285), pp. 160-174.
Levin, A., Lin, C. F. & Chu, C.S.J. (2002) Unit root tests in panel data: asymptotic and finite-
sample properties. Journal of Econometrics, 108(1), pp. 1-24.
Lumsdaine, R.L. & Papell, D.H. (1997). Multiple trend breaks and the unit-root hypothesis.
Review of Economics and Statistics, 79(2), pp. 212-218.
Maddala, G.S. & Wu, S. (1999) A comparative study of unit root tests with panel data and a
new simple test. Oxford Bulletin of Economics and Statistics, 61, pp. 631-652.
Månsson, K., Shukur, G. and Sjölander, P. (2013). Asymmetric quantile analysis of the
Swedish mortgage price discovery process. Applied Economics, 45(21), pp.3088-3101.
Neal, P. (2011). Banking Competition: The Rhetoric and the Reality. South Australian Centre
for Economic Studies, University of Adelaide, Adelaide.
Pesaran, M. H. (2004) General diagnostic tests for cross section dependence in panels,
University of Cambridge, Faculty of Economics, Cambridge Working Papers in
Economics No. 0435.
Pesaran, M. H. (2007) A simple panel unit root test in the presence of cross-section
dependence. Journal of Applied Econometrics, 22, pp. 265-312.
Renaud, B. (2005) Mortgage financing in emerging markets: constraints and feasible
development paths. Working Paper. Renaud Advisors, McLean, VA.
Satterthwaite, F.E. (1946) An approximate distribution of estimates of variance components.
Biometrics Bulletin, 2(6), pp. 110-114.
16
Smales, L.A. (2012) RBA monetary policy communication: The response of Australian
interest rate futures to changes in RBA monetary policy. Pacific-Basin Finance Journal,
20(5), pp. 793-808.
Stewart, C., Robertson, B. & Heath, A. (2013) Trends in the funding and lending behaviour of
Australian banks. Reserve Bank of Australia, Research Discussion Paper RDP 2013-15.
Valadkhani, A. (2014) Analysing interest rate mark-ups in the Australian mortgage market.
Journal of International Financial Markets, Institutions & Money, 31, pp. 343361.
Valadkhani, A., Arjomandi, A. and O'Brien, M., (2013). Does the interest rate for business
loans respond asymmetrically to changes in the cash rate? Applied Economics Letters,
20(9), pp.869-874.
Yates, J. & Berry, M. (2011) Housing and mortgage markets in turbulent times: is Australia
different? Housing Studies, 26(7-8), pp.1133-1156.
Ward, D.R. (2009) Product differentiation and consumption efficiency in mortgage markets.
Journal of Business Research, 62(8), pp. 805-809.
Welch, B.L. (1951) On the comparison of several mean values: an alternative approach.
Biometrika, 38, pp. 330336.
Windsor, C., La Cava, G. and Hansen, J. (2015). Home price beliefs: Evidence from
Australia. Journal of Housing Economics, 29, 41-58.
17
Table 1 Summary statistics for individual lenders across over time (2001-2012).
Lender
A
it
R
(%)
Max
it
R
(%)
Number of mortgage
Products Nit
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev.
Adelaide Bank
7.26
0.76
7.89
0.84
5.3
3.6
AIMS Home Loans
7.06
0.87
7.66
0.96
6.5
2.9
AMP Bank
7.31
0.81
7.88
0.86
4.4
1.6
ANZ
7.10
0.68
7.77
0.89
6.3
4.5
Arab Bank Australia
7.43
0.82
7.85
0.83
3.3
1.3
Aussie Home Loans
7.05
0.63
7.87
0.88
10.8
7.8
Austral Mortgage
6.88
0.71
7.41
0.84
10.4
7.9
B & E Personal Banking
7.04
0.74
7.62
0.84
6.3
4.5
Bankmecu
6.88
0.70
7.58
0.83
8.8
5.0
BankSA
7.34
0.77
7.91
0.82
4.3
1.1
Bankstown City CU
6.94
0.59
7.30
0.77
2.8
1.9
Bankwest
7.22
0.76
7.70
0.84
4.0
1.5
Bendigo Bank
7.33
0.81
7.78
0.84
3.6
0.7
Big Sky Building Society
7.04
0.69
7.63
0.76
5.1
2.3
Big Sky Credit Union
7.64
1.11
8.08
1.26
4.8
2.4
BMC Mortgage
6.95
0.82
7.66
0.77
9.3
5.0
BOQ
7.38
0.82
7.91
0.84
3.3
0.5
Catalyst Mutual
7.29
0.75
7.58
0.72
3.9
2.2
Citibank
7.40
0.79
8.28
0.85
15.8
14.3
Coastline Credit Union
7.03
0.75
7.65
0.78
7.2
2.6
Collins Home Loans
6.95
0.79
7.78
0.97
9.4
3.9
Commonwealth Bank
7.26
0.86
7.81
0.85
5.9
1.4
Community CPS Australia
7.35
0.71
8.16
1.08
5.3
2.6
Community First CU
7.13
0.78
7.62
0.83
3.8
1.7
CUA
7.06
0.74
7.64
0.81
6.3
2.3
ECU Australia
7.20
0.78
7.55
0.88
2.6
1.4
Encompass Credit Union
7.17
0.75
7.62
0.77
2.9
0.9
Family First CU
7.04
0.68
7.55
0.67
3.4
0.8
FCCS Credit Union
7.05
0.68
7.67
0.82
12.5
10.2
Gateway Credit Union
6.81
0.77
7.34
0.86
3.5
1.1
Greater Building Society
7.14
0.75
7.70
0.82
4.3
1.4
Heritage Bank
7.05
0.82
7.60
0.94
5.1
1.5
Holiday Coast CU
7.26
0.79
8.04
1.38
18.8
20.1
Homeloans
6.94
0.76
7.87
1.71
10.3
5.6
HomeSide Lending
7.25
0.83
7.84
0.85
6.4
4.5
Horizon Credit Union
7.27
0.76
8.14
0.99
5.7
2.8
HSBC
7.26
0.81
7.74
0.85
5.2
1.1
Hume Building Society
7.06
0.76
7.44
0.68
2.8
1.2
Illawarra CU NSW
7.19
0.69
7.80
0.85
4.3
1.9
IMB
7.21
0.81
7.66
0.83
3.7
0.9
ING DIRECT
6.98
0.74
7.68
0.81
9.3
5.3
La Trobe Financial Servs
7.56
0.93
7.96
0.91
2.3
1.5
Maritime Mining & Power
6.89
0.66
7.30
0.74
4.3
3.1
Memberfirst Credit Union
7.10
0.57
7.44
0.60
4.8
1.8
NAB
7.03
0.65
7.80
0.85
6.0
3.2
Newcastle Permanent
7.12
0.79
7.62
0.82
5.7
1.4
People's Choice Credit Union
7.27
0.74
7.69
0.84
3.7
1.0
Police&NursesMut Banking
7.26
0.78
7.74
0.81
2.7
0.9
Qantas Staff CU
6.81
0.73
7.38
0.87
10.8
7.9
Railways Credit Union
6.82
0.77
7.38
0.75
6.2
3.1
RAMS Home Loans
7.17
0.67
8.01
0.71
11.8
8.8
RESI Mortgage Corp
6.88
0.75
7.63
0.75
7.2
5.6
SCU
7.10
0.64
7.53
0.77
3.8
1.3
Select Credit Union
6.82
0.68
7.25
0.75
2.9
1.2
Service One Members Bank
7.32
0.71
8.02
0.81
3.5
1.2
St George Bank
7.34
0.78
7.91
0.82
4.4
1.2
Suncorp Bank
7.35
0.76
7.94
0.71
4.7
1.8
Sutherland Credit Union
7.18
0.67
7.88
0.69
4.8
1.2
Teachers Credit Union
7.14
0.71
7.58
0.74
4.8
2.9
The Capricornian
7.26
0.69
7.91
0.60
4.7
2.2
The Mutual
7.18
0.79
7.74
0.81
3.7
1.1
Unicredit-WA
6.99
0.69
7.53
0.72
3.1
0.3
United Community
7.32
0.83
7.89
1.06
4.4
2.7
Victoria Teachers Mutual Bank
7.10
0.77
7.53
0.80
5.4
3.8
Westpac
7.34
0.80
7.85
0.84
5.3
1.8
Cash rate
5.00
1.00
Total loans, $billion
(owner occupier+investment)
814
306
18
Table 2 Summary statistics over time averaged across 65 lenders.
Year
A
it
R
(%)
Max
it
R
(%)
Number of mortgage
Products Nit
Mean
Std.
Dev.
Mean
Std.
Dev.
Mean
Std.
Dev.
2001
6.98
0.25
8.21
0.53
3.6
1.4
2002
6.33
0.24
6.74
0.40
3.4
1.4
2003
6.59
0.16
7.09
0.22
3.9
2.0
2004
6.99
0.15
7.13
0.18
4.1
3.0
2005
7.15
0.17
7.37
0.33
5.3
4.7
2006
7.42
0.21
8.03
0.32
7.1
7.1
2007
7.95
0.26
8.51
0.25
6.3
5.5
2008
8.74
0.27
9.57
0.53
6.6
6.1
2009
6.01
0.52
7.12
0.73
5.9
3.6
2010
7.06
0.41
7.78
0.47
8.4
6.7
2011
7.47
0.30
7.75
0.37
9.7
9.1
2012
7.07
0.32
7.33
0.32
5.5
4.3
19
Table 3 Test for equality of means and variances of the number of mortgage products
(2001-2003 vs. 2010-2011).
Variables
Count
Mean
Std.
Dev.
Average number of mortgage products during 2001-2003
65
3.63
1.34
Average number of mortgage products during 2010-2012
65
7.88
6.38
Testing for equality of means using:
Stat.
p-value
t-test
-5.25
0.00
Satterthwaite (1946) t-test
-5.25
0.00
Anova F-test
27.57
0.00
Welch (1951) F-test
27.57
0.00
Test for equality of variances using:
F-test
22.54
0.00
Bartlett
115.30
0.00
Levene (Conover et al. 1981) F test
F(1,128)=18.28
0.00
Brown and Forsythe (1974) F test
F(1,128)=13.70
0.00
20
Table 4 Panel unit root test results.
Null hypothesis
()
A
it
Ln R
()
Max
it
Ln R
Ln(NOit)
Common unit root process:
LLC t test(a)
-20.36***
-12.23***
-10.85***
Individual unit root process:
W test(b)
-10.32***
-8.72***
-4.80***
ADF-Fisher
2 test(c)
317.30***
280.40***
198.87***
PP-Fisher
2 test(c)
319.06***
283.11***
199.66***
Pesaran unit root test with CD(d)
-1.983*
-3.025***
-2.345***
Pesaran CD test(e)
145.6***
136.7***
49.4***
Note: (a) Levin, Lin & Chu (LLC, 2002); (b) Im, Pesaran and Shin (2003); (c) Maddala and Wu
(1999); (d) Pesaran (2007); (e) Pesaran (2004). * and *** indicate the corresponding null is
rejected at the 10% and 1% level of significance, respectively.
21
Table 5 Estimated dynamic panel models.
Model 1:         
  
    
Model 2: 
        
  
  
  
Model 3: 
         
  
  
  
Description
Model 1
Model 2
Model 3
Coeff.
t ratio
p-value
Coeff.
t ratio
p-value
Coeff.
t ratio
p-value
3.871
2.035
0.042
-4.610
-7.167
0.000
-1.191
-0.579
0.563
+
-0.463
-3.276
0.001
0.530
13.284
0.000
0.289
2.536
0.011
-0.879
-3.937
0.000
0.588
8.673
0.000
0.203
0.984
0.325
-6.705
-1.532
0.126
1.650
1.492
0.136
-2.488
-0.731
0.465
1
5.775
1.538
0.125
-1.215
-1.293
0.197
2.283
0.791
0.430
0.741
28.419
0.000
-0.015
-0.280
0.779
0.452
2.504
0.013
Hypothesis Tests:
+=
-
0.416
4.528
0.000
-0.058
-2.025
0.043
0.086
0.927
0.354
+=1
-0.470
-11.767
0.000
-0.711
-6.254
0.000
-=1
-0.412
-6.073
0.000
-0.797
-3.870
0.000
R2
0.612
0.733
0.707
Adjusted R2
0.610
0.732
0.705
DW
2.115
1.364
2.488
Durbin’s h
-0.192
1.071
-1.009
J-statistic
0.000
>0.99
0.000
>0.99
0.000
>0.99
22
Fig. 1. Comparing the average number of mortgage products in the first, and last, three years of the sample period.
0
10
20
30
40
50
60
2001-2003 2010-2012
... We should also note that, compared to other differentiation metrics, the developed instruments do not depend on endogenous market shares (e.g., (Tallman and Li 1996)) or external information, such as financial statements or industry classifications (e.g., (Varadarajan 1986)) or other domain-specific information (e.g., (Nguyen et al. 2018)). Besides, following the proposed approach, we avoid overfitting and inducing any violation of the independence of treatment without sacrificing efficiency. ...
Preprint
In this paper, we examine the effectiveness of various recommendation strategies in the mobile channel and their impact on consumers' utility and demand levels for individual products. We find significant differences in effectiveness among various recommendation strategies. Interestingly, recommendation strategies that directly embed social proofs for the recommended alternatives outperform other recommendations. Besides, recommendation strategies combining social proofs with higher levels of induced awareness due to the prescribed temporal diversity have an even stronger effect on the mobile channel. In addition, we examine the heterogeneity of the demand effect across items, users, and contextual settings, further verifying empirically the aforementioned information and persuasion mechanisms and generating rich insights. We also facilitate the estimation of causal effects in the presence of endogeneity using machine-learning methods. Specifically, we develop novel econometric instruments that capture product differentiation (isolation) based on deep-learning models of user-generated reviews. Our empirical findings extend the current knowledge regarding the heterogeneous impact of recommender systems, reconcile contradictory prior results in the related literature, and have significant business implications.
Article
Full-text available
ÖZET: Türkiye konut finansmanı piyasasının en önemli kaynağı olan konut kredileri son yıllarda büyük gelişim göstermiştir. Konut kredilerinin gelişimini etkileyen en önemli değişken ise konut kredisi maliyeti olan kredi faiz oranlarıdır. Diğer taraftan, faiz oranı düzeyinden bağımsız olarak konut kredisi hacmi ile faiz oranı volatilitesi arasında bir ilişkinin varlığı söz konusudur. Bu çalışmada, Türk bankacılık sektöründe kullandırılan konut kredisi hacmi ile faiz oranı volatilitesi arasındaki ilişkinin ekonometrik yöntemlerle analiz edilmesi hedeflenmiştir. Bu amaçla, “Üç Aşamalı En Küçük Kareler” (ÜAEEK) yöntemi ile tahmin edilen eşanlı denklem modeli esas alınmıştır. Bu çalışma ile Türk bankacılık sektöründeki konut kredilerinin piyasa faiz oranı volatilitesi ile ilişkisi incelenerek bu alanda literatürdeki eksikliğin giderilmesine katkı sağlanması hedeflenmektedir. Modelde yer alan ve genel olarak durağan olmayan konut kredisi, gelir, enflasyon ve faiz oranı volatilitesi zaman serilerinin ÜAEEK yöntemi tahmini için uzun dönemli eşbütünleşme yaklaşımından faydalanmıştır. Analiz sonucunda, piyasa volatilitesinin konut kredisi hacmi üzerinde önemli bir etkisi olduğu bulgusuna ulaşılmıştır. Ayrıca, bu durumun faiz oranlarının genel olarak negatif eğilimli olduğu dönemlerde tasarruf sahiplerinin daha yüksek getirili araçlara yönelmesinden kaynaklanmış olabileceği değerlendirilmiştir. ABSTRACT: Housing loans, which are the most important source of the housing finance market in Turkey, have developed in recent years. The most important variable affecting the development of housing loans is the housing loan cost, in other words, housing loans interest rates. On the other hand, there is a relationship between housing loan volume and interest rate volatility regardless of the interest rate level. In this study, we aimed to analyse the relationship between the volume of housing loans in the Turkish banking sector and interest rate volatility using econometric methods. For this purpose, the simultaneous equation model estimated by the “Three-Stage Least Squares” (3SLS) method was used. We aimed to contribute to the literature by analyzing the relationship between the housing loans of Turkish banking sector and market interest rate volatility. It has benefited from the long-term co-integration approach for the 3SLS method estimation of the non-stationary time series for housing loan, income, inflation and interest rate volatility in the model. As a result of the analysis, we found that market volatility has a significant impact on the mortgage loan volume. In addition, it has been evaluated that this situation may have resulted from the fact that savers tend to use higher-yielding instruments during periods when interest rates are generally negative.
Article
Full-text available
This paper discusses the general characteristics of online markets from a competition theory perspective and the implications for competition policy. Three important Internet markets are analyzed in more detail: search engines, online auction platforms, and social networks. Given the high level of market concentration and the development of competition over time, we use our theoretical insights to examine whether (a) leading Internet platforms have non-temporary market power and, based on this analysis, (b) whether any specific market regulation beyond general competition law rules is warranted in these three online markets.
Article
New facts are documented about self-assessed home valuations using household panel data and a near-census of home sale prices. Between 2002 and 2012, homeowners’ display a small positive bias of around 1 per cent in estimating the market value of their homes, although there is considerable dispersion in beliefs and prices. Household characteristics, including age, tenure, and income and local area characteristics, such as unemployment, are associated with differences between beliefs and prices. The extent of overvaluation is positively associated with household spending, leverage and risky-asset holdings. Over the housing cycle, homeowner valuations appear less volatile than sale prices and are backward-looking; homeowners also learn from past ‘errors’. These facts support recent literature on the importance of belief formation for household decision-making.
Article
Alternative formulations of Levene's test statistic for equality of variances are found to be robust under nonnormality. These statistics use more robust estimators of central location in place of the mean. They are compared with the unmodified Levene's statistic, a jackknife procedure, and a χ2 test suggested by Layard which are all found to be less robust under nonnormality.
Article
This article examines the dynamic relationship between the Reserve Bank of Australia's (RBA's) cash rate and the variable interest rate for lending to small businesses. The relationship is evaluated via an asymmetric GARCH model using monthly data spanning from August 1990 to October 2012. Our results show that a 1 percentage point increase in the cash rate results in an instantaneous 1.086 percentage point rise in the variable rate for small businesses, whereas an equivalent 1 percentage point cut only leads to a 0.862 percentage point fall with a delay of up to 2 months. This outcome has obvious implications for the RBA's monetary policy transmission mechanism and the effectiveness of the expansionary policy versus contractionary policy.
The aim of this paper is to enhance transparency and competition in Australia's mortgage market by examining the behaviour of the mortgage interest rate spread of 39 individual lenders. Using a time-varying probability regime-switching model and monthly data (2000M9–2012M3), we identify two very distinctive regimes: a low mark-up regime (R1) and a high mark-up regime (R2). Without setting any given date a priori, the results from both the regime-switching approach and a sequential search method indicate that the spread exhibits a significant upward shift around early 2008 for all lenders. The estimated switching model controls for the rising wholesale funding costs and changes in global consumer confidence over the sample period. The results show that building societies are generally less inclined to remain in R2 than most banks (particularly foreign subsidiary banks) and some credit unions, and are consequently more likely to offer competitive mortgages to borrowers.
Article
We document three facts concerning how the structure of housing finance affects the monetary transmission mechanism: first, the characteristics of residential mortgage markets differ markedly across industrialized countries; second, the impact of monetary policy shocks to residential investment and house prices is significantly stronger in those countries with larger flexibility/development of mortgage markets; third, the transmission to consumption is stronger only in those countries where mortgage equity release is common and mortgage contracts are predominantly of the variable-rate type. We then build a two-sector DSGE model with financial constraints to rationalize those facts.
Article
This study implements a procedure to evaluate time-varying bank-interest rate adjustments over a sample period which includes changes in industry structure, market and credit conditions and varying episodes of monetary policy. The model draws attention to the pivotal role of official rates and provides estimates of a bank equilibrium policy rate. The changing sensitivity of official rates to banking conditions is identified. Results are also provided for the variation in intermediation margins and pass-throughs as well as the interactions between lending and borrowing behaviour over the years, including behaviour before, during and after the global financial crisis. The empirical methodology is applied to the US and the Australian banking systems.