UNIVERSITEIT VAN AMSTERDAM
The effects of the Loan-to-Value ratio on
consumer switching incentives
Empirical evidence from the Dutch mortgage market
Jeroen Steinfort, 6064035 [10-12-2015]
Study: Msc. Business Economics – Finance track Type of document: Master‟s Thesis
1 ABSTRACT
This thesis analyzes consumer switching incentives in the Dutch mortgage market and identifies the current Loan-to-Value ratio as a switching cost deterring consumers from switching providers through adverse selection and loss of bargaining power. A probit model is used to estimate the relation between the current Loan-to-Value ratio and consumer switching; with “Fee paid for mortgage switching service Yes/No” as a proxy for switching. A two-stage probit model using the 2007 sub-prime mortgage crisis is used to address possible endogeneity issues with the instrumental variable defined as “Mortgage closed in 2006-2008 Yes/No”. The results show that the two variables affecting consumer mortgage switching are net savings and the current Loan-to-Value, with the former having a significant positive effect and the latter a significantly negative effect on consumer switching incentives. These findings provide support for further reduction of maximum LTV ratios in order to increase switching between providers and possibly facilitate new market entry.
2 Table of Contents
1. INTRODUCTION... 3
2. LITERATURE ... 5
2.1 Switching costs ... 5 2.2 Switching behavior ... 72.3 The Loan-to-Value ratio ... 9
2.4 Empirical evidence ... 10
3. METHODOLOGY ... 12
4. DATA & DESCRIPTIVE STATISTICS ... 18
5. RESULTS ... 21
6. CONCLUSION ... 26
7. REFERENCE LIST ... 29
3 1. Introduction
Residential mortgage rates in surrounding countries have been lower as compared to rates in the Netherlands1. While the possible determinants of the pricing discrepancy have been extensively researched and various explanations have been proposed (funding capacity restrictions, price leadership bans and barriers to entry, among others)2, there hasn‟t been any research on consumer switching costs, behavior and their possible impact on competition.
This thesis analyzes the determinants of consumer switching incentives in the context of residential mortgages and identifies the borrower Loan-to-Value ratio as a driver of borrower switching incentives. The following research question is formulated: what are the effects of the current borrower Loan-to-Value ratio on consumer switching incentives?
Answering the research question would not only provide insight into the general determinants of consumer mortgage switching, but could provide an answer as to why new competitors aren‟t entering the Dutch mortgage market; high LTVs could deter borrowers from switching, making the Dutch residential mortgage market less appealing from a new entrants‟ perspective.
As switching costs lower a consumer´s net switching benefit, they increase the consumer‟s reservation price3; when the reservation price is sufficiently high, incumbent firms can effectively monopolize their existing customers as sufficiently high switching costs reduce wealth beyond benefits (Klemperer, 1987). This in turn incentivizes firms to raise prices and switching costs in order to capture additional surplus.
Moreover, actual consumer switching behavior is driven by consumers‟ expectations about switching costs and benefits which are subject to strategic complementarities4 (for example when home owners in the same neighborhood agree to wait with the sale of their house in order to maintain the value of their real estate) and spillover effects5 (for example when a house price drops due to the sale of another); causing switching behavior to resemble a prisoner‟s dilemma (Cooper and John, 1988).
1
Dijkstra, Randag and Schinkel (2014) and Mulder (2014)
2 Dijkstra, Randag and Schinkel (2014), Mulder (2014), ACM (2013), CPB (2013) and NMa (2011) 3 The price at which switching to a different product or service becomes feasible
4
Strategic complementarities refer to interactions between agents on the level of strategy 5
4 In the context of Dutch residential mortgage markets, this thesis identifies the borrower Loan-to-Value ratio as a driver for consumer mortgage switching because it captures a borrower‟s bargaining power; low LTV homeowners are less risky borrowers and therefore are able to shop for more attractive terms. Also, as break-fees depend on the difference between the new and old interest rate, LTV‟s are a proxy of break-fees since high LTV borrowers are likely to have a high interest rate (Ben-Shahar, 2009). Furthermore, LTV‟s function as a signaling device to lenders as risky borrowers self-select into higher LTV contracts (Milde and Riley, 1988). Finally, the fact that LTVs are used by policy makers as a proxy for default risk (Bank of England, 2014 and Bank for International Settlements, 2010) adds to the importance of analyzing LTVs in a wider context.
To estimate the possible effects of the LTV on consumer switching, a probit multiple regression model was used including current LTV, net savings, property value, loan size and fixed interest rate period. A probit instrumental variable model, using ‘2006, 2007 or 2008 mortgage
yes/no’ as an instrument, was then estimated to infer causality and address possible endogeneity
between the LTV ratio and dependent variable. A cross-sectional dataset was used consisting of 599 Dutch respondents and collected through an online survey by www.ikbenfrits.nl.
After conducting tests for non-linearity, adjusting for heteroscedasticity and serial correlation, a negative relation is found between the borrower current Loan-to-Value and consumer switching incentives, implying that the LTV is a switching cost keeping consumers from switching.
Switching costs and their determinants will be discussed in Section 2. Section 3 will elaborate on the consumer switching modeling framework where a regression model is used in order to investigate the relation between consumer switching and LTVs. Section 4 discusses the cross-sectional dataset and contains descriptive statistics; Section 5 consists of regression results. The final section concludes that Loan-to-Value ratios negatively affect consumer switching incentives and discusses the implications of this finding with regard to policy measures.
5 2. Literature
2.1 Switching costs
Klemperer (1987) and Sharpe (1997) define switching costs as: “all costs that cause ex ante6 homogeneous products to be viewed as heterogeneous products ex post”. Switching costs deter existing customers from changing supplier by lowering the net benefits of switching to competitors, increasing the customer reservation price7 (Klemperer, 1995). When reservation prices become sufficiently high, consumers may effectively be monopolized by them, allowing the incumbent to sustain a long run price premium over competitors; decreasing incentives to enter new markets, reducing competition and incentivizing higher price setting.
Assuming a one-period closed oligopoly and constant market size, switching costs incentivize firms in mature markets to maximize profits from existing customers instead of competing for additional market share, since the marginal costs of price competition are higher than the marginal benefits of increasing market share (Klemperer, 1987 and 1995)8. This causes markets to shift away from oligopolistic competition and allows incumbents to set monopolistic prices (Klemperer, 1987 and 1995).
In a many-period model firms have to choose each period between lowering prices to attract new customers hence building market share and increasing future profits or increasing prices to exploit existing customers. These two effects balance in an economy where there is no discounting and each firm commits to a price that cannot be changed; if firms committed to constant prices with an infinite horizon consumers would never want to switch and current consumers would have the same value as future ones, rendering the exploitation of existing consumers meaningless, resulting in the same prices set by firms in an economy with switching costs as an economy without them (Klemperer, 1987).
6
Before and after the purchase of a product or service
7 With reservation price defined as the maximum price a customer is willing to endure, before switching to a different type of product or service (Steedman, 1987)
8
Assuming symmetric firms, linear demand, constant marginal costs, homogeneous products, quantity competition and firms cannot differ between new and old customers
6 However, a more realistic model implies higher prices for an economy with switching costs for several reasons (Klemperer, 1987):
First, discounting reduces the desire to attract new customers compared to the benefits of exploiting existing ones. Second, if one company raises prices today, its competitors gain market share today and will raise prices tomorrow. Therefore, each firm has an incentive to set a high price today, to make its competitor “fatter” and less aggressive tomorrow. Third, a new consumers‟ demand is less elastic in an economy with switching costs than without, because consumers recognize that a lower price today will mean a higher price tomorrow.
When considering an open, oligopolistic economy with high growth in market size, switching costs may actually facilitate market entry as higher prices, due to switching costs, attract new entrants that are able to undercut incumbents. Also, high market growth provides a basis for increased price competition as lower prices can be off-set by increased market share. However, as markets converge to low growth in infinity, market conditions will again resemble those described in an open oligopolistic market with low growth (Schmalensee, 1983).
Switching costs can be divided into four categories: transaction costs, learning costs, contractual costs and network effects (Klemperer, 1987; Farrell & Klemperer, 2007).
Transaction costs are the costs associated with participating in a market. They exist between brands of products or services and are standardized across specific markets. For example, transaction costs could be administration fees but also the amount of effort it requires to search and switch telecom providers. Given that increased switching costs raise reservation prices and allow for additional rent extraction, firms are incentivized to stress the amount of effort by requiring difficult to obtain documents, constructing complex procedures or providing very little information.
Learning costs refer to non-transferable knowledge of different brands of the same product. An example of learning costs are the different skills required to use a computer running Windows as its operating system, versus a computer running on a Mac operating system. While there are no monetary fees or expenses, a customer that already invested time or money into learning the required skills has an incentive to stick to a particular brand.
7 Contractual switching costs arise entirely at the firms‟ discretion, and are not associated with any social costs9. An example is producer specific savings points such as stamps at a coffee shop or air miles. What differentiates contractual switching costs from other switching costs is that contractual switching costs are producer specific and can differ per producer, unlike transaction costs which are sector specific; contractual switching costs are not incurred in the form of monetary losses but in the form of foregone perks10.
Networking effects lock customers into product choices even before actual decision making is required. They hinder customers from changing suppliers in response to differences in efficiency or pricing, and increase vendor market power. An example of networking effects are the incompatibility between Apple products and its competitors which cause users to lock into all Apple products early on; even though they might not be of the highest quality or offer the best prices11
In the context of mortgage markets, examples of switching costs are: advisory and notary fees (transaction costs), switching time and effort (transaction costs), the level of transparency about lenders and pricing (transaction costs), understanding different processes, terms and conditions between banks (learning costs), break-up fees or bundle discounts (contractual costs) and compatibility12 between different financial products (networking effects).
2.2 Switching behavior
Farrell and Klemperer (2007) show that the consumers‟ switching decision and behavior are not only determined by an assessment of financial costs, benefits and required effort, but also by the consumers‟ expectation about switching costs; actual switching costs are likely to be overestimated increasing the consumers‟ reservation price and resulting in less switching activity (Farrell and Klemperer, 2007).
9 Social costs are costs that reduce consumer wealth and allocate this to suppliers
10 For example, Air-miles, complementary toys when ordering a kids meal or discount stamps 11
Network effects only correspond to already owned products or already used services and does not include any form of learning (for example: Apple products are easily compatible with other Apple products whereas they are less with other products)
12
For example integration of online banking services at one provider allows for monitoring of all bank accounts, loans and credit cards
8 The difference between switching costs and their expectation occurs due to both imperfect information: consumers may not have complete information and tend to view financial products as difficult to understand causing them to overestimate switching costs (Price, Webster and Zhu, 2013). (Klemperer, 1995; Mankiw, Reis and Wolfers, 2003). Furthermore, the expectations and behavior of a single consumer can depend on that of others, resulting in less aggregate switching activity than would be efficient.
Also, expectations and behavior of a single consumer can depend on those of others in the aggregate economy as illustrated by Diamond (1982): in the model, the key consideration is that, since production is costly and inventory holding costs oblige producers to trade, individuals will only produce when there are sufficient others that will also produce in expectation. Therefore, what individuals believe others will do is crucial in determining the aggregate economic activity.
Applying the Diamond model (1982) to mortgage markets, consider an economy with any number of consumers and low concentration of mortgage providers. Furthermore, consumers are unable to coordinate their decision making while maximizing their individual payoff, which depends on the benefits of switching minus the costs of switching. In turn, the benefits of switching depend on aggregate switching activity as more switching leads to price competition resulting in lower interest rates. This causes endogeneity within the model: consumers switch only when sufficient others will.
9 2.3 The Loan-to-Value ratio
This thesis identifies the borrower Loan-to-Value ratio as a switching cost that affects borrower switching behavior for the following reasons: a lower LTV corresponds to increased borrower bargaining power as increased capital gains decrease default risk (Campbell and Dietrich, 1983), allowing for lower interest rates and higher expected savings, functions as a signaling device to lenders as risky borrowers self-select into higher LTV contracts (Milde and Riley, 1988) and higher LTVs could correspond to increased screening efforts by the lender increasing the amount of effort required to switch by the borrower. The LTV ratio used in this thesis is defined as:
(1)
Campbell and Dietrich (1983) analyze the determinants of residential mortgage defaults and find that the amount of homeowner equity has a negative relation to homeowner defaults, as additional homeowner equity incentivizes borrowers to repay in fear of losing capital gains of the property value. From a lender‟s perspective, the amount of homeowner equity reduces the loss given default as equity provides additional residual value, lenders reclaim more value in case of a foreclosure or fire sale resulting in lower expected losses. In case of a low LTV, borrowers could gain increased bargaining power as more lenders would be willing to lend; resulting in lower interest rates (i.e. increased benefits of switching) or more beneficial terms & conditions (i.e. lower contractual switching costs).
From a lender‟s point of view, the Loan-to-Value can be seen as a signaling device through which borrowers signal their risk preferences: lenders issue large and small loan sizes to separate equilibriums and distinguish between borrower risk preferences: risk-averse borrowers self-select into smaller loans, resulting in lower LTVs; while risky borrowers would do the opposite. The rationale behind this is that safe types know their project will succeed with higher certainty, and therefore are able to repay a larger loan whereas risky borrowers are more uncertain; thus they will prefer smaller loan sizes in case they can no longer service their debt (Milde and Riley, 1988).
10 2.4 Empirical evidence
2.4.1 Switching costs
Shy (2002) uses 1997 cross-sectional data (consisting of 15.75 million bank accounts across 4 major banks) on the Finnish demand for bank deposits and Israeli cellular phone demand (including 2.15 million subscribers across 2 telephone companies), to show that, in markets with switching costs, with switching costs defined as the long-term price premium over marginal costs, competitors set prices so as to discourage consumers from switching; allowing for higher prices compared to markets without switching costs.
However, Dube, Hitsch and Rossi (2008) find that prices in markets with switching costs are 18% lower compared to markets without switching costs. They analyzed panel data from 2,100 Midwestern U.S. households between 1993 and 1995 on refrigerated orange juice and 16 oz tub margarine categories, defining switching costs as the difference between the transaction price and the price under Bertrand oligopoly. They argue that the benefits of acquiring new customers outweigh the benefits of exploiting existing ones; prices are set lower so as to compensate consumers for incurred losses.
A possible reason as to why the findings of Dube, Hitsch and Rossi (2008) conflict with the evidence found by Shy (2002), could be due to the nature of the products: compared to bank deposits and telephone subscriptions switching between orange juice and margarine is easier due to the simplicity of the transaction. Also, retail products are often one-off purchases, lack the complexity and corresponding switching costs of recurring, long-term commitments such as bank deposits and telephone subscriptions.
11 2.4.2 Switching behavior
Giulietti, Price and Waterson (2005), show that, based on 692 consumer surveys between December 1998 and January 1999 in the U.K., searching costs reduce the probability to switch within the energy sector, with searching costs measured by the time that‟s required to switch. Also, search costs have a larger negative impact on switching for consumers with little switching experience in other markets, as indicated by prior switching between car and house insurance.
Price, Webster and Zhu (2013), use 2005 survey data of 2,027 respondents from the U.K. They estimate the effects of switching costs on consumer switching behavior in several markets: electricity, mobile phones, fixed telephone lines, calls, broadband internet, car insurance, current bank accounts and mortgages. They divide switching cost and benefit determinants into two categories: expected monthly gain in GBP per month and expected switching time in hours.
Price, Webster and Zhu (2013) find expected switching time was the most important determinant of consumer switching, reducing the probability of switching by 12.6% for every additional hour while the expected monthly gain in GBP per month increased the probability of switching by 1.83% for every additional GBP saved per month.
12 2.4.3 Loan-to-Value ratios
Using 1983, 1989, 1992, 1995 and 1998 U.S. survey data on 19,756 households, Edelberg (2004) finds that borrowers with a higher default rate self-select into higher interest rate contracts and pledge less collateral while low risk borrowers do the opposite; this could lead to increased switching costs for high LTV borrowers for example, lenders could impose more comprehensive screening requirements or more restrictive terms & conditions.
Furthermore, depending on their LTV, borrowers have access to more differentiated pricing (Rabobank, 2015; ABN AMRO, 2015, ING 2015): the spread between a 65% and >85% LTV is 0.85% annual interest for ABN AMRO; reflecting lower default risk (Edelberg, 2004) and allowing low LTV borrowers to shop around for better rates.
Figure 1 shows that Dutch Loan-to-Value ratios have been increasing over the 2010 – 2014 period while mortgage switching has been decreasing. While one cannot establish a relation between the two variables given this information, it does provide a basis for further investigation and a possibility for the presence of both signaling and bargaining power effects.
Figure 1. Aggregate Loan-to-Value ratios for Dutch homeowners with a mortgage and the percentage of Dutch mortgage switching.
Source: AFM, 2014; CBS, 2015. 66% 69% 70% 79% 80% 34% 34% 25% 25% 23% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 2010 2011 2012 2013 2014
Current Loan-to-Value ratio
13 3. Methodology
3.1 Modelling framework
The proposed model for the probit estimation of switching probability P [ ] is as follows: [ |
Where = 1 if individual i paid the €50 service fee and therefore entered the switching process with ikbenfrits.nl. While the current definition does not capture any effects during the switching process that might deter consumers from switching, as the respondents are not observed after they have actually switched, it does give insight in the drivers that cause consumers‟ willingness to switch as their intention to switch is observed. This causes the model to reflect the consumer‟s perception and expectation about benefits & costs of mortgage switching; while this does not lower the internal validity of the proposed framework, it does induce bias within the estimation parameters or dataset, it should be considered when assessing external validity, meaning that the results cannot be extrapolated to reflect actual switching, as there may be factors during the negotiation process that cause respondents who expressed their intention to switch to be unable or unwilling: such as lower benefits than expected, ineligibility after screening or lower than expected appraisal values.
Combining the current Loan-to-Value ratio, and controls, net savings13, current property value, current loan size, remaining fixed interest rate period, interest differential14, top three mortgage provider15, and a quadratic LTV term yields the following probit multiple regression model:
13 Net savings are defined as interest rate savings minus break fees over the entire fixed interest rate period 14 Current interest rate – Expected interest rate; the expected interest rate is the lowest rate available given the current borrower LTV and based on a 10-year fixed annuity mortgage
15
Top three mortgage providers consist of: Rabobank, ING, ABN AMRO, Florius (ABN AMRO) and MoneYou (ABN AMRO)
14 As the variables net savings, current property value and current loan size are positively skewed (appendix 1 & 4); a logarithmic transformation has been used to help fit the data to the model, as log transformations make a positively skewed distribution more normal. Also, the log transformation changes the interpretation of coefficients from marginal effects to percentage changes.
To address the model´s possible endogeneity issues, as the current LTV ratio might be affected by omitted factors, this thesis proposes using a two-stage probit regression model with the variable ‘2006, 2007 or 2008 mortgage yes/no’ as an instrument to estimate the current LTV ratio; with the subprime mortgage crisis used as an exogenous event unaffected by any factors that might influence the LTV ratio. These specific years have been chosen because the shock to homeowner LTVs has been largest around this period16; borrowers who purchased their homes just before the start of the crisis in 2007 experienced the greatest drop in LTV. The resulting two-stage probit model is:
̂
̂
With = 1 if individual i closed his or her mortgage during the 2006 – 2008 period and 0 if not. To validate the use of the instrument, its strength will be assessed by conducting Sargan‟s J-test, while its exogeneity will be implied by conducting a Hausman test for exogeneity.
16
15 3.2 Variable descriptions
The coefficient on the Loan-to-Value ratio is expected to be significant and positive as a lower LTV ratio is expected to correspond to greater borrower bargaining power leading to lower interest rates and therefore greater benefits of switching while also signaling being a risk-averse borrower type.
Price, Webster and Zhu (2013), Giulietti, Price and Waterson, (2005) and Edelberg (2004) show that the amount of possible savings has a significant positive impact on consumer switching and therefore should be included in the analysis. As the net savings variable consists of savings minus break-fees, the variable is expected to have significant and positive impact on consumer switching.
While the dataset does not include borrower specific controls such as income, educational level or age, these factors may be still reflected in the property value and loan size (Kocenda and Vojtek, 2009). As the amount of possible financing depends directly on income (Rabobank, 2015) and income both correlates with age and education and education depends on age (Griliches and Mason, 1972). Therefore, absolute loan size and property value should be included in the analysis. The effects of both variables are expected to be ambiguous.
Also, the remaining fixed interest period should be included in the model as a longer fixed rate period allows for greater possible savings (Paiella & Pozzolo, 2007). While both rates reflect borrower specific and market risk, fixed rates are set at one point in time, causing expectations about both borrower specific and future market risk to be discounted in the fixed rate, compared to floating rates where changing market circumstances are accounted for on a continued basis and only borrower specific risk is fixed (Paiella & Pozzolo, 2007). Over time, fixed rates can become mispriced due to a mismatch in expectations about both market and borrower risk, while floating rates can only misprice borrower risk. Therefore, as the fixed interest period increases, switching becomes increasingly attractive as this would allow the borrower to gain benefits from pricing differences in current risk as compared to expected risk at the time of fixing. The remaining fixed interest period is expected to have a significant and positive effect on consumer switching.
The variable ´Top three provider´ should be included to reflect possible differences among mortgage providers and the effects they might have on respondents; for example differences in a bank‟s terms & conditions or capital requirements leading to differences in rates. The size of the top
16 three mortgage providers, networking effects through other financial products and their combined market concentration could provide power over consumer and allow them to extract more value or keep them from switching.
The interest rate differential might be an important factor because it partly determines the expected switching benefits. Also, because of its close relation to the LTV ratio and net savings, it might capture some of the bias possibly affecting the LTV thus enhancing precision.
The quadratic term is included to test for non-linearity of the LTV: it could be the case that an increase in the LTV only lowers the switching probability to a certain level and that a sufficiently high LTV actually incentivizes switching as the borrower needs to restructure his or her debt to reduce the debt burden.
17 3.3 Motivation for the model
The model proposed in this thesis differs from earlier models in a number of ways:
First, it allows switching costs and their effects on consumer switching to be analyzed individually as opposed to micro-economic models of switching costs such as the ones used by Klemperer (1987 and 1995), Farrell and Klemperer (2007) and Sharpe (1997). Also, the model analyzes switching probability and costs on an individual level providing insight in homeowner behavior.
Second, the introduction of the Loan-to-Value ratio as a possible switching cost provides additional insight into switching incentives within the Dutch mortgage market as LTV ratios have been steadily increasing over the 2010 – 2014 period while homeowner switching rates have been decreasing over the same period as mentioned in paragraph 2.4.3 (figure 1); possibly indicating a relation between the two.
Finally, the use of an instrumental variable probit model to address any possible endogeneity sets the methodology apart from similar models on switching behavior such as those used by Giulietti, Price and Waterson, (2005) who use single stage maximum likelihood models without including educational level, age or prior switching experience which may be relevant.
18 4. Data & descriptive statistics
Data has been collected by www.ikbenfrits.nl through their free online mortgage savings check consisting of 599 Dutch individuals over the August 2015 – October 2015 period. The respondents arrived at the www.ikbenfrits.nl webpage themselves, thus were not selected randomly, possibly resulting in sample selection bias as respondents were already curious about mortgage savings before filling out the questionnaires; this likely resulted in the current sample consisting of a high proportion of net savers and switchers compared to the population. Also, information regarding current housing prices was provided by the respondents themselves. Therefore, housing values may be subject to personal biases.
Furthermore, as www.ikbenfrits.nl has been published in the Dutch newspaper “Het Financieele Dagblad” and the financial online blog “Follow the Money”, its readers might be over-represented in the overall respondent base.
Also, consumer switching is not fully observed in the available dataset, as the switching procedure can take several weeks or months to finish. Therefore, only the intention to switch is observed with the payment and effort exerted by ikbenfrits.nl clients as a proxy for actual consumer switching. The entire ikbenfirts.nl process involves a free check based on general parameters in order to determine whether consumers can save on their existing mortgage at the current provider. If so, the borrower can choose to pay a service fee of €50. After paying the service fee the borrower then submits all documents required by ikbenfrits.nl to make a more detailed assessment of savings opportunities and enter switching negotiations with the particular bank. Also, the borrower needs to pay additional fees associated with mortgage switching such as appraisal fees and break fees.
The resulting database is unique both in the scope of respondents and the obtained variables as they contain information regarding Dutch homeowner‟s switching behavior, homeowner savings on switching and switching costs; while also providing insight in market concentration amongst mortgage providers as respondents provided their specific mortgage provider.
Income related data has been dropped from the analysis due to poor data quality (450 out of 599 observations were missing data). Furthermore, observations were dropped if they met any of the
19 following non-logical conditions: the current home value was less than €10.000, the WOZ value of a home was less than €10.000, the current Loan-to-Value ratio was less than 0, the Loan-to-Value ratio at closing was less than 0.
The resulting dataset meets the aforementioned criteria and consists of 555 observations; descriptive statistics for all variables are presented in table 2.
Table 2. Descriptive statistics
Descriptive statistics
Variable Mean Std. Dev. Min Max
Paid yes/no (1 = Yes; 0 = No) 0.14 0.35 0.00 1.00
Current home value (€) 416734.50 331299.50 100000.00 6000000.00
Current WOZ value (€) 358232.60 191494.40 27100.00 1959000.00
Current loan size (€) 339007.50 175560.80 60453.71 1706954.00
Home value at closing (€) 392608.50 254246.20 65000.00 3200000.00 Loan size at closing (€) 362800.80 186735.10 76458.78 1699459.00
Current Loan-to-Value 0.81 0.28 0.01 1.52
Loan-to-Value at closing 0.92 0.27 0.17 1.85
Monthly savings (€) 363.29 338.37 -390.02 2820.00
Total break fee (€) 20188.45 17005.34 1200.00 125445.00
Remaining fixed interest period
(in months) 89.96 67.81 1.00 343.00
Interest differential (in %) -1.56 0.81 -6.2 1.95
Mortgage type Number of respondents Non-amortizing 416 Saving 103 Annuity 32 Life 19
Hybrid (combination of Leven and Saving) 11
Investment 10
Linear 6
Total 597
As can be seen in the descriptive statistics, the mean of current home value and WOZ value are relatively large values in the sample compared to the national average of €211,000 (CBS, 2015). Appendix 1 shows the distribution of the current home values across different classes; the sample
20 average is skewed by a number of observation within the >€500,000 value class. This is also shown when comparing the mean of €416,734.5 versus the sample median of €350,000. Also, when looking at the home value at closing, the same pattern arises from the dataset; implying that the sample is biased towards more wealthy respondents in terms of housing value.
Furthermore, when comparing the current LTV and LTV at closing, the data shows a decrease in LTV of approximately 11%; given the average closing year of 2007, the data implies an improvement in LTV of 11% over the past 8 years. The data shows that the increase in LTV is driven by both increasing average property values (6.15%) and decreasing average principals (5.2%).
While the decrease in average principal sizes is counterintuitive given the large amount of non-amortizing loans (table 2), this is explained by lower leverage levels for new mortgages, hence lower LTV‟s, after the 2009 sub-prime mortgage crisis (appendix 2).
People who have closed their mortgage in the last 15 years make up 90.62% of the sample (appendix 3). Furthermore, the number of respondents who closed their mortgage in the three years17 prior to the sub-prime mortgage crisis of 2008-2009 accounted for 35.34% of all respondents.
While there is a difference between the sample mean of 81% current LTV and the national average of 85% (NVB, 2014), the difference is not significant.
Average net savings18 over the entire remaining fixed interest period amounted to €7,223.95 with 61.04% of respondents being able to save on their mortgage. Among the respondents who could save on their mortgage (appendix 4), 57.81% could save €1,000 – 15,000 and 27.12% could save €15,000 – 50,000 over the remaining fixed interest period.
Appendix 5 shows the distribution of mortgage providers within the sample; where the largest 20% of Dutch mortgage providers provide approximately 75% of all mortgages. The largest three Dutch banks provide 60.43% of all mortgages in the sample (appendix 6); with the remaining 39.57% being provided by a set of 38 other financial services providers. The top-5 largest amongst the other financial services providers are Aegon (5.53%), Direktbank (5.19%), Nationale Nederlanden (3.18%),
17 The years referred to are 2006, 2007 and 2008 18
Net savings are the savings over the fixed interest rate period due to a lower rate, minus the break-fees on the existing mortgage
21 Delta Lloyd (2.85%) and WestlandUtrecht (2.01%); leaving a 20.81% distributed among the 33 smallest providers.
According to the sample homeowners could reduce their interest rate by 1.56% by switching mortgage providers. Given the average current interest rate of 4.74%, a decrease of 1.56% would be a 33% reduction in interest rate.
5. Results
In section 5.1 the results of the single stage probit regression model are presented and discussed with the two-stage probit model and related robustness checks to be discussed in 5.2.
5.1 Main findings
As shown in table 3, the current Loan-to-Value ratio is negative and significant at the 1% level for all estimation models; this is in line with current literature that implies a lower LTV, should correspond to increased borrower bargaining power, lowering interest rates and allowing for greater savings, risk-averseness of the borrower and possibly increased screening efforts by the lender; as compared to a higher LTV.
Comparing the different estimation models, the first model shows the smallest effect of current Loan-to-Value ratio‟s on consumer switching: a one percent increase in current LTV would lower the probability of switching by 2.0890%. When including the controls in the model, the current LTV coefficient changes to -1.7223% due to correlation between the controls and the LTV. In the fourth model, net savings, current property value and the current loan size are significant at the 5% level while the quadratic term is significant at the 1% level, with mortgage lender significant at the 10% level. Important to note is that the variable becomes non-significant when using robust standard errors but becomes significant again after using heteroscedasticity and autocorrelation adjusted standard errors (HAC); using HAC standard errors can reduce the standard errors when there is negative autocorrelation within the cluster. An explanation could be that if respondents switched
22 within a top three bank, this would cause other respondents to be less likely to switch. Another explanation could be negative autocorrelation induced by omitted variables.
In model (8), the insignificant variables were dropped from the equation in order to enhance precision, resulting in significant coefficients at the 1% level for current LTV, net savings and the constant while property value was significant at the 5%. Furthermore, the current loan size has been dropped due to collinearity issues. Interesting is the non-significance of the quadratic term in the final model as it was significant within the prior estimation models. This is possibly due to correlation between the quadratic term and the omitted variables. Also, the use of clustering around top three mortgage provider increased the standard errors, indicating some autocorrelation might exist. This could be the case when respondents with a top three mortgage provider were likely to have similar LTV‟s.
The coefficient of the current LTV is -1.4991, indicating that a 1% increase in LTV results in a 1.4991% decrease in switching probability. The coefficient of the logarithm of savings is 0.3070, meaning that a 1% increase in savings, results in a 0.3070% increase in switching probability. A change in the current home value of 1% corresponds to an increase in switching probability of 1.1181% and having a mortgage at one of the largest three Dutch banks decreases switching by 0.5421%.
23 Table 3. Homeowner Switching: Single Stage Probit Model
Models (1) & (2) contain naïve probit regressions; all controls have been added in models (3) – (5) with all non-significant variables having been dropped in (6) – (8). Also, the log(Current loan size) has been dropped due to collinearity issues. The coefficient on the current Loan-to-Value ratio is a marginal effect while the log transformations reflect percentage changes. With *** p < 0.01, ** p < 0.05, * p < 0.1. Standard errors in parentheses.
Dependent variable: Paid fee Yes/No
(1) (2) (3) (4) (5) (6) (7) (8) Current Loan-to-Value -2.0890*** -2.0890*** -1.7223*** -1.7223*** -1.7223*** -1.4991*** -1.4991*** -1.4991*** (0.2316) (0.2316) (0.2420) (0.6100) (0.6380) (0.2565) (0.2545) (0.4035) (Current Loan-to-Value)2 8.3171*** 8.3171*** 8.3171*** 2.898* 2.898* 2.898 (1.1263) (1.9342) (1.3998) (1.5693) (1.5922) (2.0392) Log(Net savings19) 0.1953*** 0.1953** 0.1953** 0.3070*** 0.3070*** 0.3070*** (0.0698) (0.0831) (0.0988) (0.1050) (0.1010) (0.0473)
Log(Current property value) -4.6644*** -4.6644** -4.6644** 1.1181*** 1.1181*** 1.1181**
(0.4730) (2.0362) (2.2787) (0.2631) (0.2764) (0.4471)
Log(Current loan size) 4.8572*** 4.8572** 4.8572**
(0.4351) (2.0421) (2.0523) Log(remaining fixed interest
period) -0.0957 -0.0957 -0.0957
(0.0874) (0.0165) (0.1720) Mortgage provided by largest 3
banks Yes/No -0.2073* -0.2073 -0.2073* -0.5421** -0.5421** -0.5421*** (0.1241) (0.1473) (0.1220) (0.2542) (0.2581) (0.08263) Interest differential 0.0878 0.0878 0.0878 (0.0779) (0.0981) (0.0675) Constant -0.3710** -0.3710** 9.2300*** 9.2300** 9.2300 -13.97*** -13.97*** -13.97*** (0.1873) (0.1862) (2.6471) (4.6619) (7.4752) (2.6653) (2.8864) (3.3931) Observations 597 597 597 596 596 596 596 596
Robust Std. Errors YES YES YES
Clustered Std. Errors20 YES YES
Pseudo R-squared 0.0305 0.0305 0.1448 0.1448 0.1448 0.122 0.122 0.122
Prob > Chi2 0.0001 0.0001 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000
19Net savings = total savings minus break fees (“oversluitkosten”) over the remaining fixed interest period 20
24 5.2 Robustness checks
Table 4 shows the two-stage probit model results, with the variable of interest significant at the 1% level and a coefficient of -0.1117; instrumented by „Closing date 2006-2008 yes/no‟. The significance of the coefficient implies the causal relation between the LTV and consumer switching. Also, the Sargan J-test shows that the instrument is indeed valid given the p-value of 0.5312 as the null hypothesis, that the instruments are uncorrelated to some set of residuals, is not rejected. Furthermore, appendix 8 implies the exogeneity of the instrument through the Hausman test; no relation can be found between the instrument and the used controls; supporting the premise that the sub-prime mortgage crisis was an exogenous shock and a valid instrument.
The coefficient on the current LTV is reduced from -0.4991 under model (8) in table 3, to -0.1117 in model (3) of the IV probit model. An explanation for this reduction could be endogeneity in model (8), for example, omitted variables that were reflected through the LTV coefficient which were removed by the use of the instrument.
As mentioned under chapter 3, the quadratic term in table 3 is included to test for non-linearity of the LTV: it could be that an increase in the LTV only lowers the switching probability to a certain level and that a sufficiently high LTV actually incentivizes switching as the borrower needs to restructure his or her debt to reduce the debt burden. While the quadratic term was significant in model (3) through (7) in table 3, it is not significant in the IV probit model; possibly due to some correlation with the current loan size as that was the only variable dropped compared to models (3) through (5) in table 3.
While the variable „top three mortgage provider‟ was not significant in the model (2), it was included in model (3) for clustering purposes.
25 Table 4. Homeowner Switching: Endogeneity Assessment & Two-stage Probit Estimation
Models (1) – (3) exclude the log(Current loan size) due to collinearity issues. To arrive at model (3), all insignificant control variables were dropped from model (2) while the variable “Mortgage provided by largest 3 banks Yes/No” was still included for clustering purposes. Furthermore, a Sargan J-test has been conducted to assess the strength of the instrument. The coefficient on the current Loan-to-Value ratio is a marginal effect while the log transformations reflect percentage changes. With *** p < 0.01, ** p < 0.05, * p < 0.1. Standard errors in parentheses.
Dependent variable: Paid fee Yes/No
(1) (2) (3)
Current Loan-to-Value (instrumented) -0.0089*** -0.3183 -1.1117***
(0.0023) (0.4180) (0.2521)
(Current Loan-to-Value)2 0.4192
(0.7514)
Log(Net savings) 0.3239*** 0.2427***
(0.0800) (0.0544)
Log(Current property value) 0.0500
(0.1546)
Log(remaining fixed interest period) -0.2007
(0.2014)
Mortgage provided by largest 3 banks Yes/No -0.2136 -0.1929
(0.1351) (0.1333)
Interest differential 0.0977
(0.0894)
Constant -0.3707** -2.3272 -2.2532***
(0.1873) (2.0440) (0.5015)
Sargan's J-test (p-value) 0.5312 0.5312 0.5312
Observations 596 596 596
Clustered Std. Errors21 YES YES
Pseudo R-squared 0.0301 0.0301 0.0906
Prob > Chi2 0.0001 0.0001 0.0001
21
26 6. Conclusion
This thesis analyzes the determinants of consumer switching in the Netherlands and addressed the research question: what are the effects of the current borrower Loan-to-Value ratio on consumer switching incentives?
Using probit and instrumental variable regression models based on cross-sectional, observational data collected by www.ikbenfrits.nl over the period August 2015 – October 2015, an increase in the current Loan-to-Value ratio of 1% leads to a reduction in consumer switching probability of 1.1117%. Also, an increase in net savings of 1% leads to an increase in the switching probability of 0.2427% while the other controls, current property value, loan size and the remaining fixed interest period, are implied to have no effect.
The result supports the hypothesis that the current Loan-to-Value ratio is a proxy for switching costs within the Dutch mortgage market; possibly through bargaining power and signaling of borrower types leading to increased screening requirements or other imposed switching costs such as increased monitoring. This is in line with Milde and Riley (1988) and Edelberg (2004) who show the importance of current Loan-to-Value ratios in assessing borrower default risk; the LTV provides lenders with information on borrower riskiness and repayment behavior causing them to avoid high LTV borrowers.
Also, lower LTV‟s could provide increased bargaining power to borrowers as suggested by Campbell and Dietrich (1983) and Wong, Fung, Fong and Sze (2004). As mortgage providers prefer less risky borrowers, low LTV borrowers could use their attractive risk profile to shop for better rates as opposed to riskier borrowers who only have limited access to providers and products.
Furthermore, expectations of high LTV borrowers could deter them from switching; when borrowers expect not to be able to save on their current mortgage due to their inability to refinance their mortgage they might self-select not to switch.
27 7. Limitations
The limitations of this thesis are its dataset which contains only partial information on consumer switching as it does not contain information on borrowers who have actually finished the entire switching process. Also, the data potentially suffers from selection bias as respondents were not randomly chosen; homeowners who did not actively search for mortgage savings possible did not find the ikbenfrits.nl website. In addition, ikbenfrits.nl was published in several papers and online financial blogs whose reader bases do not consist of a representative sample of all Dutch homeowners.
Also, a number of possibly important variables on switching costs such as age, educational level, prior switching experience and income are not included within the used dataset. Literature suggests that educational level is an important switching cost given the perceived complexity of financial products by consumers (Price, Webster and Zhu, 2013). While possible endogeneity issues between these omitted variables and the current LTV have been addressed by the use of an instrumental variable, they might provide valuable information on switching incentives.
8. Policy implications
The results presented in this thesis provide some support for policy makers to lower the maximum allowed current Loan-to-Value ratio in order to increase consumer switching incentives. Increased consumer switching might be desired considering the possible effects of consumer switching on competition and in light of concentration in the Dutch mortgage market (appendix 5 and 6). Apart from additional consumer switching among existing mortgage providers, lower LTV‟s could facilitate additional market entrants as argued by Dijkstra, Randag and Schinkel (2014) and Mulder (2014); as a lower LTV makes Dutch borrowers more attractive to foreign lenders.
The LTV ratio is already being used as a policy measure by the Dutch government to reduce default risk within the Dutch residential mortgage market: the maximum allowed LTV at closing will be reduced from 103% in 2015 to 100% by 2018 (Rijksoverheid, 2015). Apart from its use to reduce default risk, the change in LTV could also benefit switching: a reduction of 3% corresponds to an increase of 3.34% in switching probability. As an increase of 3.34% in switching probability might
28 not be sufficient to support any significant change in market competition, a further decrease in LTV to, for example, 90%22 could prove more effective as it would increase switching consumer probability among existing providers by 14.45% while facilitating new entry by competitors as LTV‟s are closer to average levels in surrounding countries23.
However, it is important to consider the possible negative effects of lowering the maximum allowed borrower LTV as new borrowers would not be able to borrower as much, resulting in lower housing prices thus causing existing homeowners to lose value. This, in turn, would actually increase current LTVs for existing homeowners; causing the opposite to occur and possibly reducing homeowner switching on the whole in the short term.
22 Arbitrarily chosen to show a larger decrease in LTV from the current level of 103% 23
Belgium, Germany, France and the UK have average LTV ratios of 82%, 75%, 85% and 70% respectively (NVB, 2014)
29 9. Reference list
ABN AMRO, 2015, https://www.abnamro.nl/nl/prive/hypotheken/actuele-hypotheekrente/index.html
ACM, 2013, Concurrentie op de hypotheekmarkt: Een update van de margeontwikkelingen sinds begin 2011, Monitor Financiële Sector
AFM, 2014, https://www.afm.nl/~/media/files/consumenten-monitor/2014-voorjaar-hypotheken1.ashx
Bank for International Settlements, 2010, Loan-to-value ratio as a macro prudential tool – Hong Kong SAR‟s experience and cross-country evidence, Working Paper
Bank of England, 2014, Financial Policy Committee statement on housing market powers of Direction from its policy meeting, Financial Policy Committee statement
Ben-Shahar, O., 2009, A Bargaining Power Theory of Default Rules, Columbia Law Review Vol. 109, No. 2, 396-430
Campbell, T.S., and Dietrich, J.K., 1983, The Determinants of Default on Insured Conventional Residential Mortgage Loans, The Journal of the American Finance Association, Volume 38, Issue 5, 1569–1581
CBS, 2015, Financieel Risico Hypotheekschuld Eigenwoningbezitters, StatLine
30 Cooper, R., and John, A., 1988, Coordinating Coordination Failures in Keynesian Models, The
Quarterly Journal of Economics, Vol. 103, No. 3, 44-463
CPB, 2013, De Nederlandse Woningmarkt - Hypotheekrente, Huizenprijzen en Consumptie, CPB
Notitie
Diamond, P.A., 1982, Aggregate Demand Management in Search Equilibrium, Journal of Political
Economy Vol. 90, No. 5, 881-894
Dijkstra M.A., Randag F. and Schinkel M.P., 2014, High Mortgage Rates in the Low Countries: What Happened in the Spring of 2009? Amsterdam Center for Law & Economics Working Paper, No. 2014-05
Dube, J.P., Hitsch, G.J., and Rossi, P.E., 2008, Do Switching Costs Make Markets Less Competitive?
Edelberg W., 2004, Testing for Adverse Selection and Moral Hazard in Consumer Loan Markets,
FEDS Working Paper, No. 2004-09
Farrell, J., 1986, Moral Hazard as an Entry Barrier, The RAND Journal of Economic, Vol. 17, No. 3, 440-449
Farrell, J., and Klemperer, P., 2007, Coordination and lock-in: Competition with switching costs and network effects, Handbook of Industrial Organization, Vol. 3, 1967–2072
Giulietti, M., Price, C.W., and Waterson, M., 2005, Consumer Choice and Competition Policy: A Study of U.K. Energy Markets, The Economic Journal, 115, 949–968
31 Griliches, Z., and Mason, W.M., 1972, Education, Income, and Ability, Journal of Political Economy, Vol. 80, No. 3, Part 2: Investment in Education: The Equity-Efficiency Quandary, S74-S103
ING, 2015, https://www.ing.nl/particulier/hypotheken/actuele-hypotheekrente/opbouw-vaste-hypotheekrente/index.html
Klemperer, P., 1987, Markets with Consumer Switching Costs, The Quarterly Journal of Economics, Vol. 102, No. 2, 375-394
Klemperer, P., 1995, Competition when Consumers have Switching Costs: An Overview with Applications to Industrial Organization, Macroeconomics, and International Trade, Review of
Economic Studies, Vol. 62-4, 515-539
Kocenda, E., and Vojtek, M., 2009, Default Predictors and Credit Scoring Models for Retail Banking,
CESifo Working Paper Series, No. 2862
Mankiw, G.N., Reis, R., Wolfers, J., 2003, Disagreement about Inflation Expectations, NBER
Working Paper, No. 9796
Milde, H., and Riley, J.G., 1988, Signaling in Credit Markets, The Quarterly Journal of Economics, Vol. 103, No. 1, 101-129
Mulder M., 2014, The Impact of Concentration and Regulation on Competition in the Dutch Mortgage Market, Journal of Competition Law & Economics, No. 10, 795-817
NMa, 2011, Sectorstudie Hypotheekmarkt: Een onderzoek naar de concurrentieomstandigheden op de Nederlandse hypotheekmarkt, Monitor Financiële Sector
32 NVB, 2014, www.nvb.nl/media/document/001406_the-dutch-mortgage-market.pdf
Paiella, M., and Pozzolo A.F., 2007, Choosing Between Fixed and Adjustable Rate Mortgages
Price, C.W., Webster, C., and Zhu, M., 2013, Searching and Switching: Empirical estimates of consumer behavior in regulated markets, CCP Working Paper 13-11
Rabobank, 2015, https://www.rabobank.nl/particulieren/hypotheek/hypotheekrente/
Rijksoverheid, 2015, https://www.rijksoverheid.nl/onderwerpen/koopwoning/inhoud/nieuwe-regels-hypotheek
Sharpe, S.A., 1997, The Effect of Consumer Switching Costs on Prices: A Theory and its Application to the Bank Deposit Market, Review of Industrial Organization, Vol. 12, Issue 1, 79-94
Schmalensee, R., 1983, Advertising and Entry Deterrence: An Exploratory Model, Journal of
Political Economy, Vol. 91, No. 4, 636-653
Steedman, I., 1987, Reservation price and reservation demand, The New Palgrave: A Dictionary of
Economics, Vol. 4, 158–59
Shy, O., 2002, A quick-and-easy method for estimating switching costs, International Journal of
Industrial Organization, Vol. 20, 71–87
Wong, J., Fung, L., Fong, T., and Sze, A., 2004, Residential mortgage default risk and the loan-to-value ratio, Hong Kong Monetary Authority Quarterly Bulletin
33 10. Appendix
Appendix 1. Distribution of property values
Appendix 2. Distribution of Loan-to-Value ratio‟s according to closing year 0 20 40 60 80 100 120 140 Nu m b er o f resp o n d en ts
Property values (in EUR) Current home values Home values at closing
0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 L o an -to -Valu e Closing year
34 Appendix 3. Distribution of closing years
Appendix 4. Distribution of Net savers*
*Respondents who could not save (i.e. savings equal to or lower than 0) were excluded from this chart
0 10 20 30 40 50 60 70 80 90 100 2 0 1 5 2 0 1 4 2 0 1 3 2 0 1 2 2 0 1 1 2 0 1 0 2 0 0 9 2 0 0 8 2 0 0 7 2 0 0 6 2 0 0 5 2 0 0 4 2 0 0 3 2 0 0 2 2 0 0 1 2 0 0 0 1 9 9 9 1 9 9 8 1 9 9 7 1 9 9 6 1 9 9 5 1 9 9 4 1 9 9 3 1 9 9 2 1 9 9 1 1 9 9 0 1 9 8 9 1 9 8 8 1 9 8 7 1 9 8 6 1 9 8 5 Nu m b er o f resp o n d en ts Closing year 0 10 20 30 40 50 60 70 80 90 Nu m b er o f resp o n d en ts Net savings
35 Appendix 5. Lorenz-curve of mortgage providers
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 2% 4% 7% 9% 11% 13% 16% 18% %02 22% 24% 27% 29% %31 33% 36% 38% 40% %24 44% 47% 49% 51% %35 56% 58% 60% 62% %46 67% 69% 71% 73% %67 78% 80% 82% 84% %87 89% 91% %93 96% 98% 1 0 0 %
36 Appendix 6. Distribution of mortgage providers
Distribution of mortgage providers
Mortgage provider Number of respondents Mortgage provider Number of respondents
Rabobank 127 BNP Paribas 12
ING 91 a.s.r. 10
ABN AMRO 69 Argenta 10
Florius 39 SNS Bank 9
Aegon 33 Van Lanschot Bankiers 7
Direktbank 31 BLG Wonen 7
Obvion 28 DBV 6
Nationale Nederlanden 19 MoneYou 5
Delta Lloyd 17 Hypotrust 5
WestlandUtrecht 12 GMAC (nu: CMIS) 5
Mortgage provider
Number of respondents Mortgage provider Number of respondents
REAAL Verzekeringen 5
Royal Residentie
Hypotheken 2
Bank of Scotland 5 Philips Pensioenfonds 2
Syntrus Achmea 4 Bouwfonds (nu: Florius) 2
WoonNexxt 4 Robeco 1
Quion 4 Avero Achmea 1
ZwitserLeven 3 EuropeLife 1
FBTO - Eigen Huis 3 Triodos Bank N.V. 1
Centraal Beheer 3 AXA 1
Woonfonds 3 UCB 1
Solide Koers
37 Mortgage provider Number of respondents MNF Bank 1 Generali 1 Erasmus 1 NIBC 1 Allianz 1 Total 597
Market share ABN AMRO, Rabobank & ING 60.43%
(including subsidiaries*)
38 Appendix 7. Cross-correlation table
Paid yes/no Current home value Current WOZ value Current loan size Home value at closing Loan size at closing Current Loan-to-Value Loan-to-Value at closing Loan-to-Value change over time Monthly savings Total break fee Remaining fixed interest period Total savings minus break fee Paid yes/no 1
Current home value 0.0949 1
Current WOZ value 0.1302 0.7033 1
Current loan size 0.1012 0.4828 0.6948 1
Home value at closing 0.1448 0.6564 0.9089 0.7386 1
Loan size at closing 0.1277 0.5176 0.7385 0.9289 0.7814 1
Current Loan-to-Value -0.1644 -0.234 -0.2333 0.3955 -0.1341 0.2655 1
Loan-to-Value at closing -0.1455 -0.1899 -0.2142 0.2934 -0.2397 0.2906 0.7692 1
Loan-to-Value change over time -0.0386 -0.0795 -0.0437 0.1739 0.142 -0.0176 0.4007 -0.2772 1
Monthly savings 0.201 0.4255 0.5881 0.5491 0.6052 0.5683 -0.0418 -0.019 -0.0356 1
Total break fee 0.0495 0.2718 0.381 0.4846 0.4003 0.4584 0.1315 0.06 0.1116 0.401 1
Remaining fixed interest period 0.038 -0.0588 -0.0447 0.0434 -0.0272 0.0372 0.1206 0.0508 0.1084 -0.2122 0.4334 1
39 Appendix 8.
Table 6. Homeowner Switching: Implied Exogeneity of the Instrument
Dependent variable: Closing date 2006-2008 Yes/No
(1)
Log(Net savings) -0.029
(0.0411)
Log(Current loan) (0.183)
(0.119)
Log(Current property value) -0.0714
(0.126)
Log(Remaining fixed interest period) -1.46
(1.412)
Interest differential -1.320
(1.412)
Top three provider -0.461
(0.812)
Observations 596
Robust Std. Errors YES
40 Appendix 9. Ikbenfrits.nl questions
