220 Week 8 F /For WIZARD KIM

Rev Ind Organ (2015) 46:145–167 DOI 10.1007/s11151-014-9441-0

Explaining Variation in Title Charges: A Study of Five Metropolitan Residential Real Estate Markets

Robert M. Feinberg · Daniel Kuehn · Signe-Mary McKernan · Doug Wissoker · Sisi Zhang

Published online: 20 September 2014 © Springer Science+Business Media New York 2014

Abstract Title charges are costs that are related to establishing and insuring title in a real estate transaction. These charges vary considerably and include title insurance premiums and other fees involved in closing a mortgage loan. This study analyzes the variation in title charges using data from almost 3,000 FHA-insured, 30-year fixed-rate home purchase loans in five metropolitan areas. Even after controlling for cost-related and other characteristics, more than one-half of the variation in title charges remains unexplained in four of the five markets. Substantial differences in charges between settlement agents suggest that consumers would benefit by shopping for settlement services.

Keywords Title insurance · Real estate settlement · Housing

R. M. Feinberg (B) · D. Kuehn Department of Economics, American University, 4400 Massachusetts Avenue, Washington DC, NW 20016, USA e-mail: [email protected]

D. Kuehn · S.-M. McKernan · D. Wissoker The Urban Institute, 2100 M Street, Washington DC, NW 20037, USA e-mail: [email protected]

S.-M. McKernan e-mail: [email protected]

D. Wissoker e-mail: [email protected]

S. Zhang Shanghai University of Finance and Economics, 777 Guoding Road, Shanghai 200433, China e-mail: [email protected]

123

146 R. M. Feinberg et al.

1 Introduction

Title charges are costs that are related to establishing and insuring title in a real estate transaction. Title charges include the costs of performing a title search, preparing a title insurance binder, attorney or other settlement agent fees, the cost of the insurance that covers the lender and (often) the buyer against title defects, and other fees related to processing the title insurance policy. Establishing that a seller has legitimate ownership of his or her property is essential to the smooth operation of residential real estate transactions, given the large share of household wealth tied up in property (GAO 2007). Title insurance also merits attention because it represents a substantial proportion of residential real estate closing costs that are paid by consumers and adds significantly to the cost of purchasing a home.

Title charges vary considerably across the U.S.1: Woodward (2008) finds that nationally, title charges averaged $1,200 per loan for FHA-insured loans in 2001. The GAO’s (2007) study of the industry identifies a wide range of premiums—the amount that is paid for the insurance policy and the major component of title charges—from an average of $700 in Des Moines, Iowa, to an average of $2,190 in New York City. Title charges are also an important component of the wider array of closing costs. A widely cited, regularly conducted survey of closing costs by bankrate.com consis- tently concludes that the substantial variation in title insurance charges across real estate markets accounts for much of the observed variation in closing costs (Lewis 2014).

Title charges not only vary widely across markets, but also vary considerably within a metropolitan area. This study uses data from almost 3,000 FHA-insured, 30-year fixed-rate home purchase loans from May and June, 2001, in counties covering five major metropolitan real estate markets to examine the determinants of variation in title charges faced by consumers within each area.2 To our knowledge this is the first paper that examines variation in title charges within metropolitan areas, con- trolling for characteristics of homebuyers, houses, neighborhoods, and settlement agencies. The model is reduced-form and does not attempt to estimate structural parameters.

The five metropolitan areas studied (Cook County, Illinois; Philadelphia County, Pennsylvania; Broward County, Florida; Maricopa County, Arizona; and Sacramento County, California) were selected due to variation in the regulatory regimes in each market and especially the perception of Cook County and Philadelphia as being on opposite ends of the regulatory spectrum. Qualitative interviews with market partic- ipants (settlement agents, underwriters, and academic and legal experts) inform the

1 The large variation in title charges could result from the relatively non-competitive nature of the title insurance market. Researchers argue that title charges are held substantially above cost by local industry practice, state regulation, and the infrequency with which consumers shop around (Woodward and Hall 2010). Critics allege that in competing for the business of real estate agents and underwriters, settlement agents have less incentive to satisfy what they perceive to be a captive customer (Dumm et al. 2007; GAO 2007; Owen 1977; White 1984). Even more sympathetic analyses highlight such “reverse competition” as a genuine problem that is inherent in the market and requires a solution (Roussel and Rosenberg 1981). 2 HUD selected the May 21-June 30 closing date period based on an examination of interest rate trends. The goal was to select a period during which interest rates were stable.

123

Explaining Variation in Title Charges 147

empirical model and interpretation of the regression analyses. Multivariate regression analysis provides insights into the role of market-specific institutions in shaping the title charges.

Our analysis of title charges across the five metropolitan areas suggests that the amounts paid by consumers vary substantially, both between and within areas. The price of a housing unit, amounts paid to real estate agents, and settlement agent fixed effects are found to be associated with title charges, though the association differs by market and component of title charges. In most of the markets studied, title charges for comparable properties vary substantially across settlement agents. Minority home- buyers pay higher total title charges than do white homebuyers in Broward and Cook counties. After controlling for various characteristics of homebuyers, housing units, neighborhoods, and the identities of the top settlement agents, between 84 % (Cook County) and 46 % (Philadelphia) of the variation in total title charges remains unex- plained.

Section 2 provides a brief description of the existing research on title charges. Section 3 describes market characteristics in each of the five metropolitan real estate markets. Section 4 discusses data and methods. Section 5 presents regression results, and Sect. 6 concludes.

2 Background

Although some studies examine the factors that are associated with the variation in total real estate closing costs (Birnbaum 2005; GAO 2007; Woodward 2008), little is known about what factors explain the wide variation in title charges, which is an important component of closing costs. McKernan et al. (2009) find that although the largest component of closing costs is the lender fee, a substantial cost—with a median value of $1,284—was attributable to title insurance and related settlement services. Woodward (2008) finds that title charges rise with loan amount.

While the risk associated with title defects rises with the loan amount, payments for title claims account for such a small share of the premium that the effect of loan size on premiums should be minimal. This pattern can be interpreted as either price discrimination against buyers of higher-valued properties or subsidization of insurance for buyers of lower-valued properties. In addition, the modest discounts that are given to customers who are refinancing or buying recently sold properties (which should have substantially lower title search costs) can also be viewed as indicators of price discrimination or market power (Eaton and Eaton 2007).

The title insurance industry is primarily regulated at the state level, with regu- lations that are often focused on premium rates.3 Of the five markets studied in this paper, only Cook County has no regulation of premiums. Twenty-one states, including Arizona and California (studied here), have “file-and-use” provisions, where agents can set their own rates but must register them with a state regulator that often has some authority to review and reject filed rates. Sixteen states, including Pennsylvania

3 For another aspect of state regulatory policy and a good discussion of the title insurance industry, see Jaffee (2006).

123

148 R. M. Feinberg et al.

(studied here), are “prior–approval” states, which are different from file-and-use states in that rates must be approved by the state before settlement agents can use them. How- ever Pennsylvania also allows “rating bureaus”, which are widely thought to act as cartels of title companies that set rates collectively. Four states “promulgate” or set rates for settlement agents, including Florida, which is where the fifth market covered in this study is located.

The data for these analyses are from a sample that was compiled by the Department of Housing and Urban Development (HUD) in 2001. Because closing cost information is not compiled digitally, they remain the best available for studying patterns of charges. Interviews with agents indicate that practices and regulations change only very slowly, although new regulations from HUD in 2008 that simplified the lenders’ “good faith” estimates of closing costs are expected to have improved homebuyers’ ability to shop for title insurance.

3 Market Regulation

To understand better the title insurance industry in specific real estate markets, we conducted qualitative interviews with settlement agents, underwriters, and academic and legal experts. These interviews covered local title insurance industry practices and inform the empirical model and interpretation of the regression analysis. This section describes what we learned from the interviews about market characteristics.

Philadelphia and Broward County are the most constrained of the five markets studied. Philadelphia cannot be strictly categorized as a “regulated” market. State regulation is minimal and consists of approval of premiums agreed to by a rating bureau, which sets common rates on a complete package of settlement services, from title insurance underwriting through the actual closing of the purchase. Premiums are determined by a formula and increase as the price of the home increases. The market is therefore tightly constrained by rating bureau decisions, but not by a public regulator. Some smaller underwriters do not participate in the rating bureau and may charge somewhat less than the rating bureau rates.

Broward County has the tightest title insurance regulation imposed by government of all of the markets in this study. Similar to the rating bureau in Philadelphia, premiums in Broward County are a function of home price. Florida does not regulate title search or abstracting fees,4 which are added separately.

Maricopa County and Sacramento County are moderately regulated. The title insur- ance agents in those counties are only required to file rates with the state prior to using them. Title insurance fees and search fees must be filed in both of these metropolitan areas, and abstracting fees must also be filed in Maricopa County.

Cook County is the least regulated of the five markets. The Cook County title insurance market is widely recognized as one of the least regulated markets in the country, which led one interviewee to refer to it as “the Wild West.” A unique aspect of the title market in Cook County is that attorney fees (which are a small or

4 An abstract is a collection of legal documents that are associated with a title and that are compiled by the title agent.

123

Explaining Variation in Title Charges 149

non-existent settlement expense elsewhere) represent a significant share (approxi- mately one-third) of total non-lender closing costs. High attorney usage in Cook County is driven by the finding in The Chicago Bar Association v. Quinlan and Tyson, Inc., a (1966) Illinois Supreme Court ruling that many tasks associated with real estate settlements constitute the practice of law, and a settlement agent performing these tasks would be guilty of the “unauthorized practice of law.”

4 Data and Methods

The primary data source for this study is the Metropolitan HUD-1 Settlement Cost Database. It contains 2,839 HUD-1 settlements for FHA-insured, fixed-rate 30-year mortgage loans that were sampled from the population of all loans that were closed between May 21 and June 30, 2001, in the five metropolitan areas that were listed in the Introduction. These loans are for the purchase of an owner-occupied home and do not include refinancing loans. In addition, 370 HUD-1 additional settlements from the same time and locations were added from the National HUD-1 Settle- ment Cost Database, a parallel, nationally representative database of 7,600 HUD-1 settlements.

The 2,839 settlements from the metropolitan database and the 370 additional set- tlements from the national database were matched by HUD to the FHA data that correspond to the loan. These data provide basic demographic information on the bor- rower, as well as the price, location, and other characteristics of the house itself. The FHA data include the census block group of the house, which allows for a match of the tract-level data from the 2000 Census with the settlement data.

Most of the title insurance premium is retained by the settlement agent to cover search costs, but a portion of it is paid to the title insurance underwriter to insure against problems with the title. This portion, known as the “retention rate”, is usually negotiated between the underwriter and the agent (Florida is the exception among our five states, since it imposes a mandated retention rate). In some markets, components such as “abstract or title search fees” or “title examination fees” are itemized separately, while in others they are included in the premium. In most cases, both lenders and buyers receive title insurance, with coverage for the second parties usually available at a substantial discount.

The HUD-1 form provides additional lines for “endorsement fees,” which pro- vide additional title insurance coverage and are usually added at the request of the lender. Standard endorsements include “restrictions, easements, and minerals” endorsements, which protect land; “environmental protection lien” endorsements, which insure against losses that are associated with state environmental claims on the property; and “planned unit development” endorsements, which insure that the house’s development was organized in accordance with regulations.

The settlement records are supplemented with Census data by block group, obtained from the Summary File 3 (SF-3) of the 2000 Census. The SF-3 data provide neighbor- hood characteristics for variables that are not available for individual home purchases, such as the amount of turnover in the housing stock and the average age, education, and income levels of residents of the census block group. To investigate variation in

123

150 R. M. Feinberg et al.

Table 1 Definition of title charge categories

Title charge Definition HUD-1 Line

Total title charges Total title charges are all of the charges that were paid by consumers to the settlement agent or to attorneys for the title and closing work that was associated with a house purchase. This is the total of all charges that are listed in the 1100 series of the HUD-1 form

1101, 1102, 1103, 1104, 1105, 1106, 1107, 1108, 1109, 1110, 1111, 1112, 1113

Premium plus endorsement charges

Premiums are typically listed in line 1108, and are paid directly to the settlement agent. When information in line 1108 is missing, lines 1109 and 1110 (indicating the price that was paid for buyer and lender coverage) are used instead. Standard endorsements are occasionally included in line 1108 with the premium, but also appear in the extra lines that are provided on the HUD-1 form

1108, 1109, 1110, 1111, 1112, 1113

The HUD-1 line numbers refer to those of the HUD-1 that was in use through 2009

title charges across settlement agents, we used the settlement agent’s name and address from the HUD-1 forms to create unique settlement agent identifiers.

Outliers in the title charge and house price data are removed from the sample. The top and bottom 1 % of values in total title charges in each metropolitan area are excluded from the samples for the regressions that are presented.5 After trimming and removing cases with missing or outlier values in the covariates discussed below, the number of observations used in the five metropolitan area regressions is 2,874. Regressions using both levels and logs were estimated, although only the logged version is shown here.

4.1 Dependent Variables in the Regression Analysis

Multivariate regression models are constructed to describe the relationship between title charges that homebuyers pay and the characteristics of the homebuyers, their houses, and their local communities. The analyses in this study combine buyer and seller charges for title services, with the assumption that who pays for a specific component of the costs of a home sale is fungible. Payment arrangements determined at closing may be determined by factors such as the availability of cash on hand to pay closing costs, but it is assumed that all costs are ultimately passed on to the buyer.

This study examines both broad and narrow measures of title and settlement costs: total title charges, as well as premium plus endorsement charges and a residual category of “other” title charges. Table 1 defines these categories, and notes the line numbers from the existing HUD-1 form that were used to construct the variables.

5 Regressions that trimmed the top and bottom 0.5 and 2 % of values were also estimated to ensure that the findings are robust to different trimming strategies.

123

Explaining Variation in Title Charges 151

4.2 Explanatory Variables in the Regression Analysis

As noted above, we conducted qualitative interviews with market participants to assess the likely determinants of title charges to include as explanatory variables in the regres- sions.6 The interviewees noted several determinants of the costs faced by settlement agents and title insurance underwriters that may be passed on to consumers. They identified differences in costs as primarily resulting from variation in the difficulty of the title search required and differentials in the extent to which properties are likely to face title problems.

The interviewees pointed out that any characteristic of a property that might increase search costs has the potential to increase fees, particularly in a market where settlement agents have substantial discretion in charging fees. They suggested that search costs may be higher for older houses, because those houses have longer title histories. Some agents indicate that they add predetermined fees for houses that are older than 30 or 40 years for this reason. Properties that combine multiple prior properties raise search costs as well, since a separate search must be conducted for each title chain.

The relationship between housing stock turnover and title costs is more ambiguous. Houses with greater turnover have more complicated title chains to search, with a greater likelihood of prior mistakes or liens. However, more recent purchases ensure that title searches have been done more recently, which reduces risk and search costs. Properties that are located in areas with high foreclosure rates or borrowers who pose a high risk of foreclosure might pay higher title insurance rates, either because underwriters demand a higher portion of the premium to cover the added risk or because lenders require more endorsements.

Title insurance charges can also be higher for reasons that have no apparent relation to insurance or search costs. For example, charges can be higher if a fee is assessed for the fairly typical simultaneous issuance of buyer and lender policies. This can be as low as $25, but in some markets such as Florida there is no statutory cap.

The interviewees suggested that premiums will tend to increase with the house price even if the relationship is not built into a set formula. One theory was that higher premiums are charged on more expensive houses for the simple purpose of cross-subsidizing lower-income consumers. In explaining closing fees specifically (sometimes called escrow fees), another expert suggested that the higher liability that is associated with greater amounts of funds being handled for high-priced houses justified higher charges. Another settlement agent disagreed, stating that title search and closing fees generally are fixed costs and do not depend on house value. Interviewees did not mention the potential for higher losses when higher loan amounts are at risk as a reason for higher premiums assessed against more expensive houses, perhaps because of the low share of premiums that are actually used to pay claims.

Based on these interviews, we defined three classes of explanatory variables and their association with title charges. The first class includes characteristics of the house that are identified by interviewees and the literature as influencing the costs

6 We interviewed two industry specialists who worked in non-profit research and advocacy organizations, one representative from a title insurance trade association, one settlement agent in each of four of our study markets (Phoenix, Chicago, Sacramento, and Broward County), and three settlement agents in Philadelphia.

123

152 R. M. Feinberg et al.

of settlement agents and underwriters, and which therefore could affect fees charged to consumers. The second class of variables includes characteristics of the homebuyer, house, and neighborhood that are not directly related to the costs that are incurred by the title firm, and therefore should not be associated with higher or lower charges. The third class of variables is a set of fixed effects for the ten largest agents that operated in each metropolitan area. These three classes are described below, and the mean values for each variable can be found in Table 2.

4.2.1 Cost-Related Factors

Philadelphia and Broward County settlement agencies determine title insurance pre- miums through the use of an explicit formula that is dependent on house price. Settle- ment agencies in all markets might reasonably consider house prices relevant for costs because insurance claims will increase with house price, although all interviewees sug- gested this was not an important factor. Title agencies may charge higher premiums for more expensive properties for other reasons. The optimal time spent examining a title is most likely greater as the risk assumed by the title insurer increases. More expen- sive properties represent a greater risk for insurers, but one would expect a nonlinear relationship between home price and search effort due to the diminishing productivity of effort spent in search activities. A logarithmic function of home price and other continuous explanatory variables is used to allow for a nonlinear relationship, and so relationships can be interpreted as elasticities.7

The qualitative analysis of the five metropolitan markets concluded that title insur- ance claims are more likely to be paid when properties go into foreclosure. To account for this, the models include an indicator of whether a homebuyer’s loan-to-value ratio exceeds the FHA recommendation of 97.5 % as a proxy for foreclosure risk. Older properties, which are likely to have longer title chains, are also identified as increasing search costs for agents. The models therefore include indicators for houses that are less than 5 years old and greater than 30 years old (with properties between 6 and 30 years old as the reference group).

Housing stock turnover, which has an ambiguous expected relationship to costs, is controlled for with the inclusion of a measure of the percentage of households that moved into the census block group in the last 5 years. Finally, since larger residential lots may increase search costs due to the potential incorporation of multiple title chains, the models control for the square footage of the lot.

4.2.2 Non-Cost-Related Factors

We include several non-cost-related factors in the regression models to determine whether certain homebuyers pay higher charges. Interpretation of the association of these variables with title charges is delicate; while strong empirical relationships may be a sign of overt discrimination, they may also suggest substantial market segmenta- tion with certain settlement agents’ serving a restricted set of neighborhoods.

7 We also estimate regressions in levels and use quadratic versions of house price to capture any non- linearities. These models are not shown here, but the results are largely comparable to the preferred log specification.

123

Explaining Variation in Title Charges 153

Table 2 Explanatory variable means

Cook Co. Philadelphia Maricopa Co. Sacramento Broward Co. IL PA AZ CA FL

Home price

Price/10,000 $13.87 $7.18 $10.99 $14.21 $11.31

Loan-to-value ratio >97.5 % 15 % 80 % 46 % 25 % 54 %

Cost related

Unit ≤5 years old 4 % 3 % 12 % 4 % 13 % Unit >30 years old 68 % 80 % 21 % 28 % 19 %

% Households in block group that moved in last 5 years

41 % 37 % 55 % 52 % 54 %

Property lot size/100 53.82 24.34 67.64 62.29 54.15

Non-cost related

Buyer is minority 62 % 48 % 45 % 35 % 60 %

Buyer race unknown 4 % 18 % 4 % 0 % 7.6 %

Buyer lives in urban area 33 % 83 % 25 % 3 % 35 %

Buyer’s monthly effective income/1,000

$4.71 $3.04 $3.67 $4.50 $4.17

Percent of households in block group 25+ with no high school degree

24 % 26 % 21 % 20 % 21 %

Percent of households in block group 25+ with college degree

17 % 13 % 18 % 18 % 18 %

Avg. block group HH income/1,000

$56.42 $43.97 $52.49 $51.79 $50.76

Median block group age 33.7 35.1 31.4 32.4 36.6

Settlement

Largest agency 21 % 11 % 12 % 18 % 4 %

Second largest agency 14 % 5 % 13 % 17 % 3 %

Third largest agency 10 % 5 % 10 % 15 % 2 %

Fourth largest agency 8 % 6 % 9 % 10 % 2 %

Fifth largest agency 9 % 3 % 7 % 8 % 2 %

Sixth largest agency 6 % 3 % 6 % 8 % 2 %

Seventh largest agency 3 % 3 % 5 % 6 % 2 %

Eighth largest agency 2 % 2 % 5 % 5 % 2 %

Ninth largest agency 2 % 3 % 4 % 5 % 2 %

Tenth largest agency 1 % 3 % 4 % 3 % 2 %

Loan to value ratio 96.20 97.5 96.89 96.56 95.9

Amount paid to real estate agents $4,860.13 $4,007.96 $4,850.24 $6,693.73 $4,200.74

YSP $2,794.46 $1,559.30 $2,189.12 $2,785.06 $2,645.31

Cash payment $3,209.31 $1,168.96 $3,019.66 $4,039.21 $3,582.21

Data source: Metropolitan HUD-1 Settlement Cost Database, 2001 Non-monotonically decreasing settlement agent shares result from settlement agencies being ranked before data trimming on the dependent variable

123

154 R. M. Feinberg et al.

Minorities have a long history of housing market discrimination (Turner et al. 2002) and may also pay higher title costs. A dummy variable for whether the buyer is a minor- ity (black, Hispanic, other race) is included in the regression to determine whether minorities pay higher title charges than similarly situated whites. An indicator is also included for census block groups that are urban (as opposed to rural and suburban). While urbanity should not influence costs, more densely populated neighborhoods may host more title agencies and thus have lower rates due to competition.

Higher income levels for a homebuyer or a neighborhood should not increase title costs, after controlling for price, because the title search process remains the same between these buyers. However, if title agencies price discriminate, or if high-cost real estate agents partner with high-cost settlement agents, there may be a relationship between income and title charges. Title agencies may also operate in specific neigh- borhoods, on the basis of neighborhood income levels. The monthly income of the homebuyer and average household income in the census block group are included as control variables to determine whether these charges vary with income.

Less-educated homebuyers may be more susceptible to settlement agents who attempt to take advantage of them by raising costs, and they may lack the infor- mation or confidence to investigate and negotiate their closing costs, including their title charges. The models include two education variables: (1) the percentage of cen- sus block group residents who are 25-years-old or older with less than a high school diploma and (2) the percentage of block group residents with a college degree. The median age of residents in a census block group is included in the models.8 Although age should have no impact on search costs, savvy title agents may exploit the relative lack of confidence or experience of younger homebuyers.

4.2.3 Settlement Agent Factors

In a competitive market, with no cost to consumers of obtaining information, individual settlement agents should not charge substantially different fees for title insurance services after controlling for all of the factors that are expected to influence title costs. In a market where certain agents charge consistently higher or consistently lower fees than other agents, consumers could benefit from comparison shopping. To test whether settlement agents charge different fees, the models include fixed effects for each of the ten largest title insurance agents in each market.9 The coefficients on the fixed effects provide an estimate of the difference between each agent’s charges and the charges of smaller agents or those for whom the agent is unknown.

The results indicate whether there is a significant range of title charges between the largest agents after controlling for other relevant factors. Care is required in interpret- ing these results as a reflection of the nature of competition between title companies. In

8 Alternative median neighborhood age specifications are discussed in Sect. 5, but are not ultimately included in reported output. 9 Models with five settlement agent fixed effects are also estimated to determine the robustness of the results to the number of fixed effects specified. The results are consistent between the two specifications. Models with ten settlement agent fixed effects are presented here to describe the range of title charges within a larger share of the market.

123

Explaining Variation in Title Charges 155

both a highly competitive market and a tightly regulated (or cartelized) market, these agent fixed effects are expected to be negligible. However, in a less-regulated market, especially where consumer information or shopping around is limited, substantial vari- ation in charges would be expected across agents. Research on home buyers by Lacko and Pappalardo (2010) also suggests that many do not understand standard mortgage disclosures, which suggests further difficulties in shopping for title insurance.

In addition to the settlement agent fixed effects, the models include the amount of money that was paid to the real estate agent, the yield spread premium (YSP),10 and the amount of cash that was paid at settlement. Previous research by Woodward (2008) suggests that these factors capture the capacity of settlement agents to exploit unwary or otherwise vulnerable customers. Real estate agents who are able to extract large rents from home buyers may pass these customers on to settlement agents who are equally willing to extract these rents (in some cases, the former may enjoy kickbacks from the settlement agents for the service). Any payments to settlement agents must of course be paid in cash, so the amount of cash that a home buyer has on hand to pay plays a limiting role in the amount of rent available for extraction in the settlement process.

4.3 Comparison of Conditions and Charges Across the Five Markets

Summary statistics on the explanatory variables suggest that the five metropolitan areas differ in their demographic and economic conditions and in the characteristics of the housing stock (Table 2). Homebuyers with FHA mortgages in Philadelphia are much more likely to have high loan-to-value ratios and much lower house prices than do homebuyers in any other metropolitan area. This suggests that in Philadelphia, title insurance underwriters may face greater foreclosure risk or that higher leverage is more feasible for buyers because the price-to-income ratio is low.

Homes are generally older in Cook County and Philadelphia, where more than two-thirds of the homes are over 30 years old. On the contrary, less than a third of the properties are more than 30 years old in the other three counties. The lot size is much smaller in Philadelphia than the other four counties (2,434 square feet versus over 5,000 square feet). Homebuyers’ characteristics also differ. The majority of homebuyers are minorities in Cook and Broward County, and homebuyers are on average older in Broward County.

None of the five markets is dominated by a single settlement agency, but a few agencies account for a nontrivial share of the market in Cook County, Maricopa County, and Sacramento. In Sacramento, four settlement agencies account for 60 % of the market for title insurance on FHA mortgages. Finally, as noted earlier, Cook County is the only market where a substantial share of homebuyers use an attorney (not shown in Table 2 because very low variation in attorney usage resulted in the variable’s exclusion from the model). Minimums and maximums for continuous variables across the five markets are presented in Table 3.

10 The YSP is the compensation that is paid by the lender to the mortgage broker (but generally passed onto the borrower in lower cash origination charges) for selling a borrower a loan at a higher interest rate than what that lender could otherwise have qualified for.

123

156 R. M. Feinberg et al.

T ab

le 3

C on

ti nu

ou s

va ri

ab le

m in

im um

s/ m

ax im

um s

C oo

k C

o. P

hi la

de lp

hi a

M ar

ic op

a C

o. S

ac ra

m en

to B

ro w

ar d

C o.

IL PA

A Z

C A

F L

H o m

e p ri

ce

P ri

ce /1

0, 00

0 $4

.2 0/

$2 5.

50 $1

.9 0/

$1 6.

50 $1

4. 50

/$ 16

.5 0

$5 .5

0/ $2

2. 40

$4 .5

0/ $1

8. 90

C o st

re la

te d

P er

ce nt

of ho

us eh

ol ds

in bl

oc k

gr ou

p th

at m

ov ed

in la

st 5

ye ar

s 8

% /8

9 %

4 %

/8 2

% 16

% /1

00 %

17 %

/9 2

% 16

% /9

7 %

P ro

pe rt

y lo

t si

ze /1

00 6/

39 0

5/ 23

6 8/

20 1

7/ 21

6 7/

15 6

N o n -c

o st

re la

te d

B uy

er ’s

m on

th ly

ef fe

ct iv

e in

co m

e/ 1,

00 0

$1 .5

6/ $2

0. 98

$1 .0

6/ $9

.6 2

$1 .1

4/ $1

4. 67

$1 .3

0/ $2

0. 35

$1 .2

4/ $1

7. 58

P er

ce nt

of ho

us eh

ol ds

in bl

oc k

gr ou

p 25

+ w

it h

no hi

gh sc

ho ol

de gr

ee 2

% /7

2 %

1 %

/7 4

% 2

% /8

2 %

2 %

/6 2

% 3

% /6

1 %

P er

ce nt

of ho

us eh

ol ds

in bl

oc k

gr ou

p 25

+ w

it h

a co

ll eg

e de

gr ee

1 %

/6 5

% 1

% /7

8 %

1 %

/6 8

% 1

% /5

1 %

2 %

/5 6

%

A vg

.b lo

ck gr

ou p

H H

.i nc

om e/

1, 00

0 $1

2. 01

/$ 11

6. 09

$1 2.

55 /$

97 .8

8 $2

2. 61

/$ 10

9. 29

$2 0.

00 /$

10 7.

73 $2

5. 39

/$ 10

6. 37

M ed

ia n

bl oc

k gr

ou p

ag e

19 .3

/6 7.

3 22

.5 /7

6. 3

20 .7

/7 3.

2 18

.2 /5

2. 9

24 /7

5. 5

S et

tl em

en t

L oa

n to

va lu

e ra

ti o

74 .0

7/ 98

.7 5

76 .5

6/ 98

.7 5

73 .9

5/ 98

.7 4

37 .9

4/ 10

3. 2

39 .2

1/ 97

.7 5

A m

ou nt

pa id

to re

al es

ta te

ag en

ts $0

/$ 16

,1 67

$0 /$

9, 30

0 $0

/$ 10

,1 50

$0 /$

12 ,5

40 $0

/$ 11

,2 50

Y S

P $1

8/ $8

,9 94

$1 24

/$ 5,

98 8

$9 8/

$8 ,0

30 $2

88 /$

7, 71

6 $7

9/ $6

,5 46

C as

h pa

ym en

t $3

35 /

$1 5,

77 0

$2 5/

$7 ,0

06 $1

75 /

$1 2,

17 0

$5 0/

$1 2,

98 7

$2 7/

$1 1,

27 9

D a ta

so u rc

e: M

et ro

po li

ta n

H U

D -1

S et

tl em

en t

C os

t D

at ab

as e,

20 01

123

Explaining Variation in Title Charges 157

Table 4 Summary statistics in title charges

Metropolitan area Total charges Premium plus endorsement charges

Other charges

Mean

Cook County, IL 1,827 866 961

Philadelphia, PA 1,007 781 226

Maricopa County, AZ 1,558 1,044 514

Sacramento County, CA 1,772 1,102 670

Broward County, FL 1,882 884 998

Minimum

Cook County, IL 991 335 190

Philadelphia, PA 613 150 0

Maricopa County, AZ 897 177 140

Sacramento County, CA 990 303 134

Broward County, FL 1,030 362 200

Maximum

Cook County, IL 4,494 3,151 3,725

Philadelphia, PA 4,245 1,303 3,267

Maricopa County, AZ 2,238 1,445 1,439

Sacramento County, CA 2,843 1,539 1,952

Broward County, FL 2,924 1,459 1,830

Coefficient of variation

Cook County, IL 0.254 0.167 0.459

Philadelphia, PA 0.334 0.202 1.255

Maricopa County, AZ 0.144 0.151 0.225

Sacramento County, CA 0.145 0.15 0.244

Broward County, FL 0.19 0.211 0.27

Data source: Metropolitan HUD-1 settlement cost database. Top and bottom 1 % outliers in total title charges are trimmed Title charges are in 2001 dollars. The “coefficient of variation” is the standard deviation normalized by (divided by) the mean

Title charges vary widely across the five metropolitan areas. Mean total title charges in Philadelphia are roughly half the mean charges in Cook and Broward Counties. While total title charges are highest, on average, in Cook and Broward Counties, premiums and endorsements are the highest in Maricopa County and Sacramento (Table 4).

The coefficients of variation for total title charges range from below 0.15 to over 0.33 and suggest that even after controlling for regulatory or institutional environment by focusing on specific metropolitan areas, substantial variation in title charges exists. There is greater variation within Philadelphia (0.34) and Cook County (0.24) than there is within any other metropolitan area considered in this study. We believe that the wider differential for Cook County is due to the high use of attorneys (attorney costs seem to vary widely) and relative lack of regulation. The finding for Philadelphia is somewhat surprising given its reputation for being tightly controlled by the rating bureau.

123

158 R. M. Feinberg et al.

While Eaton (2009) suggests that premiums that are set by the rating bureau in Penn- sylvania are unnecessarily high, it has not been widely suggested that Pennsylvania experiences the wide variation in title charges that have been identified in metropolitan areas such as Washington, D.C. (Woodward 2008). The greater variation in Philadel- phia is likely due to greater variation in home prices than in other markets, because it also exhibits high variability in the premiums (in contrast to Cook County which shows lower premium variability despite high variability in total title charges).11

4.4 Empirical Approach and Specification

We apply ordinary least squares to estimate each dependent variable as a function of house price (represented in Eq. 1 as PRICEi) and all of the factors that are hypothesized to be associated with the costs of the settlement agency or the underwriter that could be passed on to consumers, as well as those noted above that are not expected to influence title costs. These include characteristics of individual homebuyers (Xi), as well as characteristics of a census block group where multiple homebuyers in the sample may reside (Zc). We also include the ten settlement agent fixed effects and all other settlement factors described above besides the agent fixed effects (Ri). All continuous variables are logged.12 The model specification is:

lnYi = α + β′1ln (PRICEi) + β′2Xi + β′3Zc + β′4Ri + ui (1)

This model is estimated separately (in Tables 5, 6, and 7, respectively) for logged versions of total title charges, title premium plus endorsements, and other title charges in each of the five metropolitan areas.13 Premiums plus endorsements are modeled separately to focus more exclusively on the charges of title agents themselves, apart from the broader suite of costs that are associated with closing.

The relationships between the dependent variables and the explanatory variables are estimated in separate regressions for each metropolitan area because variation in institutions across metropolitan areas is expected to produce different patterns of association. Chow tests that were performed on pooled regressions for each dependent variable all confirmed that unpooled regressions offered the appropriate specification.

11 The coefficient of variation of home prices is 0.40 in Philadelphia, which is much higher than in the other four markets: Cook County, 0.30; Sacramento County, 0.24; Broward County, 0.21; Maricopa County, 0.19 (not shown in table). 12 One variable—the amount that is paid to the real estate agents that were involved in the transaction—is equal to zero for between 15 and 30 % of cases, depending on the market. Instead of dropping this significant portion of the sample, we take the log of the reported payment to real estate agents plus one dollar. Results are robust to whether one dollar is added before logging the payment to real estate agents or other amounts such as 10 or 25 dollars are added. These are likely cases where no real estate agent is used. Elder et al. (2000) report that in a 1996 survey of home buyers, 16 % of transactions do not use a real estate agent. 13 An F-test revealed that the estimated regression models for the two components of total title charges (premiums plus endorsements and other charges) are significantly different in all five counties.

123

Explaining Variation in Title Charges 159

Table 5 Explaining variation in logged total title charges

Philadelphia Broward County

Maricopa County

Sacramento Cook County

Logged house price/$10,000 0.307*** 0.329*** 0.342*** 0.276*** 0.231***

(0.0435) (0.0489) (0.0413) (0.0385) (0.0444)

Loan to value ratio exceeds 97.5%

−0.0330 0.00704 0.0205* 0.00690 0.0183

(0.0241) (0.0162) (0.0112) (0.0157) (0.0265)

Unit <5 years old −0.0987* 0.00839 −0.0466*** −0.0293 −0.0664 (0.0546) (0.0257) (0.0156) (0.0278) (0.0425)

Unit >30 years old −0.0320 0.00187 −0.0496*** 0.00691 −0.0293 (0.0231) (0.0200) (0.0131) (0.0117) (0.0190)

Log % of HH in block group moved in last 5 years

0.0327 0.0418 −0.0561*** −0.0386 −0.0138

(0.0240) (0.0363) (0.0212) (0.0251) (0.0279)

Logged property lot size/100 0.0623*** 0.00978 0.00134 0.0285** −0.0243 (0.0193) (0.0147) (0.0118) (0.0118) (0.0156)

Buyer is minority 0.0219 0.0300* −0.0115 0.0109 0.0374* (0.0208) (0.0177) (0.0126) (0.0114) (0.0200)

Buyer race unknown −0.00253 −0.0210 −0.0373 −0.0510 0.0801* (0.0265) (0.0309) (0.0253) (0.0828) (0.0442)

Buyer lives in urban area −0.0819*** 0.0143 −0.0109 −0.0152 0.00104 (0.0295) (0.0163) (0.0115) (0.0281) (0.0206)

Buyer’s log monthly effective income/$1,000

0.00340 0.00267 0.0147 0.00828 0.0546*

(0.0269) (0.0282) (0.0175) (0.0168) (0.0309)

Log %. of HH in block group 25+ with no HS degree

−0.0130 −0.0472 −0.0135 −0.0231* 0.00518

(0.0241) (0.0294) (0.0123) (0.0138) (0.0212)

Log % of HH in block group 25+ with college degree

0.00566 −0.0343 0.00849 0.0174 0.0187

(0.0164) (0.0267) (0.0118) (0.0125) (0.0204)

Log block group avg. HH income/$1,000

0.0439 0.00386 −0.0115 −0.0257 −0.108**

(0.0490) (0.0501) (0.0296) (0.0318) (0.0462)

Log median block group age −0.0134 0.0162 0.0229 −0.0847* −0.153* (0.0700) (0.0503) (0.0438) (0.0484) (0.0794)

Largest agency 0.0166 0.0680* −0.00505 0.0766*** −0.0408* (0.0285) (0.0374) (0.0173) (0.0235) (0.0246)

Second largest agency −0.0746* −0.161*** 0.0198 0.0352 0.0349 (0.0381) (0.0446) (0.0173) (0.0237) (0.0281)

Third largest agency 0.0104 −0.155** −0.0499*** 0.00866 0.00314 (0.0377) (0.0602) (0.0187) (0.0241) (0.0314)

123

160 R. M. Feinberg et al.

Table 5 continued

Philadelphia Broward County

Maricopa County

Sacramento Cook County

Fourth largest agency 0.0508 0.0893* 0.0445** −0.0361 −0.0157 (0.0365) (0.0489) (0.0192) (0.0256) (0.0330)

Fifth largest agency −0.0955* 0.0851* −0.00879 −0.0862*** −0.0319 (0.0508) (0.0516) (0.0211) (0.0266) (0.0328)

Sixth largest agency −0.0284 −0.0426 0.00150 −0.000722 0.0644* (0.0488) (0.0488) (0.0225) (0.0277) (0.0380)

Seventh largest agency −0.0514 −0.0835 −0.0526** 0.0102 −0.0941* (0.0529) (0.0514) (0.0245) (0.0289) (0.0486)

Eighth largest agency 0.126** 0.0188 −0.0983*** −0.115*** 0.0706 (0.0580) (0.0529) (0.0249) (0.0295) (0.0643)

Ninth largest agency 0.0491 0.0305 −0.0495* 0.0185 −0.00799 (0.0523) (0.0558) (0.0272) (0.0309) (0.0701)

Tenth largest agency −0.0895 −0.136** −0.0233 −0.0399 0.0349 (0.0573) (0.0554) (0.0259) (0.0361) (0.0810)

Log loan to value ratio 0.000613 0.000471 0.000470 −0.000208 0.00938*** (0.00510) (0.00154) (0.00230) (0.00121) (0.00303)

Log amount paid to real estate agents

−0.0159*** 0.00593*** 0.00940*** 0.00746*** 0.00625**

(0.00392) (0.00204) (0.00180) (0.00199) (0.00273)

Log YSP 0.00405 −0.00291 0.0163 −0.0111 −0.0286 (0.0190) (0.0170) (0.0118) (0.0105) (0.0201)

Log cash payment −0.00835 0.0202** 0.00662 −0.00236 0.0104 (0.00762) (0.00923) (0.00770) (0.00816) (0.00968)

Constant 6.227*** 6.264*** 6.160*** 7.035*** 7.090***

(0.600) (0.375) (0.313) (0.253) (0.459)

F-test for joint significance of ten settlement agent coefficients

2.07** 3.91*** 4.49*** 10.99*** 1.80*

F-test for joint significance of non−cost related variable coefficients

1.68* 1.01 1.68 1.31 3.10***

Observations 493 527 594 539 722

R-squared 0.543 0.285 0.426 0.450 0.157

Data source: Metropolitan HUD-1 database for FHA mortgages, 2001 Standard errors in parentheses *** p < 0.01; ** p < 0.05; * p < 0.1

5 Regression Results

5.1 Home Price

The price a consumer pays for his or her house has a strong positive association with total title charges in all five markets. This relationship is likely driven by the

123

Explaining Variation in Title Charges 161

Table 6 Explaining variation in logged premiums and endorsement charges

Philadelphia Broward County

Maricopa County

Sacramento Cook County

Log house price/$10,000 0.463*** 0.693*** 0.447*** 0.342*** 0.211***

(0.0274) (0.0408) (0.0492) (0.0442) (0.0278)

Loan to value ratio exceeds 97.5%

0.0217 0.00474 0.0238* 0.0202 0.00185

(0.0152) (0.0135) (0.0133) (0.0181) (0.0166)

Unit <5 years old 0.0360 0.0425** −0.0578*** −0.0574* −0.0860*** (0.0345) (0.0214) (0.0186) (0.0320) (0.0266)

Unit >30 years old −0.00187 0.00708 −0.0362** −0.00664 0.00690 (0.0146) (0.0167) (0.0156) (0.0134) (0.0119)

Log % of HH in block group moved in last 5 years

−0.0215 −0.0113 −0.0370 −0.0257 −0.00338

(0.0152) (0.0303) (0.0252) (0.0288) (0.0175)

Log property lot size/100 0.0247** −0.0220* 0.0208 0.0235* −0.0220** (0.0122) (0.0123) (0.0141) (0.0135) (0.00980)

Buyer is minority 0.00546 0.00198 −0.0156 0.00600 −0.000646 (0.0131) (0.0148) (0.0150) (0.0131) (0.0125)

Buyer race unknown −0.0318* −0.0526** −0.0431 −0.0587 −0.0479* (0.0167) (0.0258) (0.0302) (0.0950) (0.0277)

Buyer lives in urban area 0.0141 0.0254* −0.00758 −0.0333 −0.0247* (0.0186) (0.0136) (0.0137) (0.0322) (0.0129)

Buyer’s log monthly effective income/$1,000

0.0288* 0.0222 0.00297 −0.00263 −0.00704

(0.0170) (0.0235) (0.0209) (0.0193) (0.0194)

Log %. of HH in block group 25+ with no HS degree

0.0251 −0.0225 0.00727 −0.0299* 0.0197

(0.0152) (0.0245) (0.0147) (0.0158) (0.0133)

Log % of HH in block group 25+ with college degree

−0.0136 −0.0474** 0.0142 0.00617 0.00558

(0.0104) (0.0223) (0.0140) (0.0143) (0.0128)

Log block group Avg. HH income/$10,000

−0.0470 0.0253 4.07e−05 0.0166 −0.0295

(0.0309) (0.0419) (0.0353) (0.0365) (0.0290)

Log median block group age 0.00385 0.0496 0.0710 −0.0580 0.000113 (0.0442) (0.0420) (0.0522) (0.0556) (0.0498)

Largest agency 0.0336* 0.107*** −0.0209 0.0609** 0.0175 (0.0180) (0.0312) (0.0206) (0.0270) (0.0154)

Second largest agency 0.00767 −0.0505 −0.0114 0.0447 −0.0122 (0.0241) (0.0372) (0.0206) (0.0272) (0.0176)

Third largest agency 0.0254 −0.112** −0.0361 0.0411 −0.0293 (0.0238) (0.0503) (0.0222) (0.0277) (0.0197)

123

162 R. M. Feinberg et al.

Table 6 continued

Philadelphia Broward County

Maricopa County

Sacramento Cook County

Fourth largest agency 0.0313 0.0768* 0.0263 0.0664** 0.0628***

(0.0230) (0.0408) (0.0229) (0.0294) (0.0207)

Fifth largest agency 0.0237 0.110** −0.0493** −0.0322 −0.0207 (0.0320) (0.0430) (0.0251) (0.0305) (0.0205)

Sixth largest agency −0.0372 0.00280 0.00998 0.0429 0.0829*** (0.0308) (0.0407) (0.0268) (0.0318) (0.0238)

Seventh largest agency 0.0255 −0.120*** −0.0873*** 0.0262 0.00266 (0.0334) (0.0430) (0.0292) (0.0332) (0.0305)

Eighth largest agency 0.0384 0.0472 −0.0759** −0.0574* −0.0115 (0.0366) (0.0442) (0.0297) (0.0339) (0.0403)

Ninth largest agency 0.0124 0.117** −0.0253 0.0115 0.128*** (0.0330) (0.0466) (0.0324) (0.0354) (0.0439)

Tenth largest agency −0.0114 −0.131*** −0.0718** 0.0159 0.217*** (0.0362) (0.0462) (0.0309) (0.0414) (0.0508)

Log loan to value ratio −0.00168 −0.000175 0.000725 0.000701 −0.000447 (0.00322) (0.00128) (0.00274) (0.00139) (0.00190)

Log amount paid to real estate agents

0.000375 0.00271 0.00925*** 0.00620*** 0.00245

(0.00247) (0.00170) (0.00215) (0.00229) (0.00171)

Log YSP 0.0247** 0.0110 0.0155 −0.000898 0.00190 (0.0120) (0.0142) (0.0141) (0.0120) (0.0126)

Log cash payment −0.0113** 0.0209*** 0.00729 −0.00558 −0.0116* (0.00481) (0.00771) (0.00917) (0.00936) (0.00606)

Constant 5.828*** 4.473*** 5.278*** 5.981*** 6.558***

(0.379) (0.313) (0.373) (0.290) (0.288)

F-test for joint significance of ten settlement agent coefficients

0.85 5.28*** 2.72*** 3.50*** 5.73***

F-test for joint significance of non-cost related variable coefficients

3.50*** 1.72* 1.31 1.36 1.38

Observations 493 527 594 539 722

R-squared 0.734 0.599 0.389 0.422 0.255

Data source: Metropolitan HUD-1 database for FHA mortgages, 2001 Standard errors in parentheses *** p < 0.01; ** p < 0.05; * p < 0.1

strong association between house price and premiums; premiums are explicitly set as a function of house price in some markets. A 10 % increase in house price is associ- ated with an increase in total title charges of 3.42 % in Maricopa County, 3.29 % in Broward County, 3.07 % in Philadelphia, and with lower degrees of association for Sacramento and Cook County (Table 5). House price also has a positive association with increased premiums plus endorsements in all five markets (Table 6). These elas-

123

Explaining Variation in Title Charges 163

Table 7 Explaining variation in logged other charges

Philadelphia Broward County

Maricopa County

Sacramento Cook County

Log house price/$10,000 −0.265 0.00773 0.0862 0.199** 0.288*** (0.164) (0.0872) (0.0696) (0.0773) (0.0842)

Loan to value ratio exceeds 97.5%

−0.136 0.0117 0.0150 −0.0210 0.0322

(0.0908) (0.0289) (0.0189) (0.0316) (0.0502)

Unit <5 years old −0.352* −0.0236 −0.0232 0.0131 −0.0565 (0.206) (0.0458) (0.0263) (0.0559) (0.0806)

Unit >30 years old −0.0123 −0.0107 −0.0921*** 0.0353 −0.0732** (0.0869) (0.0357) (0.0221) (0.0234) (0.0360)

Log % of HH in block group moved in last 5 years

0.0685 0.0856 −0.0899** −0.0701 −0.0137

(0.0905) (0.0648) (0.0357) (0.0504) (0.0528)

Log property lot size/100 0.233*** 0.0352 −0.0117 0.0113 −0.0297 (0.0728) (0.0263) (0.0199) (0.0236) (0.0296)

Buyer is minority 0.0102 0.0410 −0.0100 0.0277 0.0745** (0.0784) (0.0317) (0.0212) (0.0228) (0.0379)

Buyer race unknown −0.00159 0.0160 −0.0266 −0.0295 0.156* (0.100) (0.0552) (0.0427) (0.166) (0.0837)

Buyer lives in urban area −0.454*** 0.00685 −0.0140 0.0260 0.0141 (0.111) (0.0290) (0.0194) (0.0563) (0.0390)

Buyer’s log monthly effective income/$1,000

−0.00163 −0.0147 0.0427 0.0340 0.0929

(0.101) (0.0503) (0.0296) (0.0337) (0.0586)

Log %. of HH in block group 25+ with no HS degree

−0.0795 −0.0700 −0.0505** −0.0127 0.000431

(0.0909) (0.0524) (0.0208) (0.0277) (0.0401)

Log % of HH in block group 25+ with college degree

−0.0452 −0.0311 −0.00464 0.0243 0.0392

(0.0619) (0.0476) (0.0198) (0.0250) (0.0387)

Log block group Avg. HH income/$10,000

0.118 −0.0325 −0.0178 −0.0809 −0.196**

(0.184) (0.0895) (0.0499) (0.0638) (0.0877)

Log median block group age −0.136 −0.0309 −0.0385 −0.0999 −0.235 (0.264) (0.0898) (0.0738) (0.0972) (0.151)

Largest agency 0.0960 0.0530 0.0234 0.151*** −0.0948** (0.107) (0.0668) (0.0291) (0.0472) (0.0466)

Second largest agency −0.362** −0.304*** 0.0618** 0.0686 0.0744 (0.144) (0.0795) (0.0291) (0.0476) (0.0533)

Third largest agency −0.199 −0.214** −0.0828*** 0.00735 0.00516 (0.142) (0.107) (0.0315) (0.0484) (0.0596)

123

164 R. M. Feinberg et al.

Table 7 continued

Philadelphia Broward County

Maricopa County

Sacramento Cook County

Fourth largest agency 0.0824 0.118 0.0667** −0.159*** −0.125** (0.137) (0.0873) (0.0324) (0.0515) (0.0626)

Fifth largest agency −0.500*** 0.0966 0.0607* −0.137** −0.0790 (0.191) (0.0920) (0.0355) (0.0534) (0.0622)

Sixth largest agency 0.110 −0.0677 −0.0176 −0.0297 −0.00129 (0.184) (0.0871) (0.0379) (0.0555) (0.0721)

Seventh largest agency −0.289 −0.0416 0.00478 0.0349 −0.219** (0.199) (0.0918) (0.0413) (0.0580) (0.0922)

Eighth largest agency 0.380* 0.0114 −0.162*** −0.180*** 0.124 (0.218) (0.0944) (0.0421) (0.0592) (0.122)

Ninth largest agency −0.0277 −0.0181 −0.104** 0.0826 −0.115 (0.197) (0.0996) (0.0458) (0.0619) (0.133)

Tenth largest agency −0.537** −0.125 0.0650 −0.0728 −0.156 (0.216) (0.0988) (0.0437) (0.0724) (0.154)

Log loan to value ratio −0.00555 0.000503 −0.00156 −0.00232 0.0201*** (0.0192) (0.00274) (0.00388) (0.00244) (0.00575)

Log amount paid to real estate agents

−0.0422*** 0.0114*** 0.0102*** 0.0106*** 0.0107**

(0.0148) (0.00364) (0.00304) (0.00400) (0.00518)

Log YSP 0.00135 0.00282 0.0265 −0.0128 −0.0820** (0.0716) (0.0304) (0.0200) (0.0210) (0.0381)

Log cash payment −7.22e−05 0.0230 0.00516 −0.00318 0.0282 (0.0287) (0.0165) (0.0130) (0.0164) (0.0183)

Constant 6.212*** 6.510*** 5.905*** 6.746*** 6.114***

(2.260) (0.669) (0.528) (0.507) (0.871)

F-test for joint significance of ten settlement agent coefficients

2.96*** 2.46*** 5.30 *** 11.24*** 2.21**

F-test for joint significance of non-cost related variable coefficients

2.71*** 0.50 1.75* 0.72 2.63***

Observations 493 527 594 539 722

R-squared 0.220 0.119 0.230 0.256 0.116

Data source: Metropolitan HUD-1 database for FHA mortgages, 2001 Standard errors in parentheses *** p < 0.01; ** p < 0.05; * p < 0.1

ticities are generally higher than those for total title charges. Statistically significant relationships between home price and other title charges are only (and in these cases, weakly) positive in Sacramento and Cook counties (Table 7). Consumers in most mar- kets, from those that are stringently regulated to those with light regulation, can expect to pay higher title charges for more expensive houses.

123

Explaining Variation in Title Charges 165

5.2 Cost-Related Factors

Characteristics of the household and neighborhood that are expected to be related to title costs had their clearest association with title charges in Maricopa County. Newer housing units are associated with reduced total title charges in both Philadelphia and Maricopa County and with reduced premiums plus endorsements charges in Maricopa, Sacramento, and Cook counties (Tables 5, 6).

Other cost-related factors are often significant in Maricopa County, but are not as significant across markets. For example, more recent housing stock turnover and older units are only significant in Maricopa County, where they reduced total title charges. Loan-to-value ratios that exceeded the FHA’s 97.5 % threshold are associated with higher title charges in Maricopa.

5.3 Non-cost-Related Factors

Non-cost-related factors, such as demographic or educational characteristics of the household or neighborhood, exhibit a relatively weak relationship with title charges, although some factors are more consistently significant than others. Being a minority homebuyer is associated with title charges that are 3 % higher in Broward and 3.74 % higher in Cook County than for white homebuyers.

Living in an urban area in Philadelphia is associated with 8.19 % lower title charges than for non-urban areas of the market, although premiums and endorsements are unrelated to urban residence in Philadelphia. An increase in the median age of a block group is associated with reduced total charges in Sacramento and Cook County, although it is unrelated to premiums plus endorsements.14 Most other non-cost-related factors (notably, education level) are unrelated to total charges.

5.4 Settlement Agent Factors

Settlement agent fixed effects have a substantial range of values and are jointly sig- nificant in all of the markets in the total and the other title charges regressions, and all of the markets except for Philadelphia for premiums plus endorsements. This is a sensible result, since the strong regulation of premiums in Philadelphia would reduce variability in premiums across firms (but not necessarily non-premium charges).

More of the settlement agent fixed effects are significant in the total title charges regression for Broward County than in any other market. Twelve of the 19 settlement agent fixed effects that are significant across all five markets offer lower total title charges than does the reference population of smaller agents, which suggests the

14 To test whether age exhibited a non-linear association with costs (for example, with neighborhoods of relatively young or relatively old residents both at a disadvantage relative to more average block group median age levels) we substituted dummy variables for younger, middle, and older median age block groups (roughly corresponding to the interquartile range and the lowest and highest quartile). No notable non-linearities were identified. The categorical version (not shown) is consistent with the continuous version in showing that neighborhoods with younger residents pay higher title charges. Median block group age was also interacted with average block group household earnings, but this was statistically insignificant.

123

166 R. M. Feinberg et al.

potential either for scale economies or market segmentation. The cartel-type rating bureau system in place in Philadelphia likely explains the somewhat limited variation found among agent fixed effects for that market in Table 6. Significant settlement agent fixed effects are unlikely to reflect consumer perceptions of agent quality, since consumers are known to be poorly informed about the title industry.

The amount of money paid to real estate agents has a positive and statistically signif- icant relationship with total title charges in Broward, Maricopa, Cook, and Sacramento counties (although it has an unexpected negative association in Philadelphia), and a positive but weak association with premiums plus endorsements in Maricopa and Sacramento counties. Contrary to prior research by Woodward (2008), YSP does not seem to be closely related to title charges. Cash payments are associated with premi- ums plus endorsements in Cook County and Philadelphia, although in two of these cases it has an unexpected negative association.

6 Summary and Implications

This study examines title charges for a sample of FHA-insured residential closings within five metropolitan areas, defined at the county level. Previous research has shown substantial variation between states in title charges and the associated regulations and institutions. We take this into account by examining the variation in charges within mar- kets rather than across markets. The empirical models in this study explain variation in three measures of title charges (total title charges, premiums plus endorsements, and other charges) within five different metropolitan areas with the use of three classes of control variables: cost-related factors, non-cost-related factors, and settlement factors.

Cost-related characteristics of the home or neighborhood, especially house sales price but also age and turnover of housing units, are generally associated with title charges. Factors not related to the costs incurred by settlement agents are generally insignificant across all five metropolitan areas, although race seems to play a modest role in explaining total title charges (though not title insurance premiums) in Broward and Cook Counties.

In all markets, there is significant variation in charges across settlement agents, with many of the ten largest settlement agents charging fees that are significantly higher or lower than those of smaller settlement agents or with unknown agents. In Philadelphia, substantial differences in premiums plus endorsements across agents are not observed, which is likely due to the presence of the all-inclusive premium that is set by the rating bureau there.

Significant fixed effects suggest that homebuyers may benefit by shopping for title insurance with more than one settlement agent. However, the unconditional variation in charges across agents is likely to exist in part because of the relationships and referrals between real estate agents and settlement agents. The positive association between the amount of money paid to real estate agents and title charges across several markets supports this hypothesis, which may make it challenging for consumers to compare the rates of different agents. Consumers may also be constrained in shopping by the location of the house in question. Some of the variation between settlement agents identified in the regression models may be due to settlement agent specialization in

123

Explaining Variation in Title Charges 167

specific neighborhoods or developments. While not a perfect solution, another option would be for one of the parties that are involved in settlements—perhaps the lender— to bundle title charges with other loan costs, so that the consumer may be better able to compare the prices of the complete packages of services.

Acknowledgments This article is based on a report funded by the U.S. Department of Housing and Urban Development Order Number C-CHI-01027CHI-T0001. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders. We thank William Reid, William Reeder, Kurt Usowski, and all of HUD’s Office of Policy Development and Research staff for their helpful suggestions. The authors also thank the settlement agents, industry representatives, and title insurance experts who contributed to our understanding of the industry, as well as Larry White and two anonymous reviewers for their comments on an earlier draft.

References

Birnbaum, B. (2005). An analysis of competition in the California title insurance and Escrow industry. Report to the California Insurance Commissioner. Retrieved from: http://www.insurance.ca.gov/0400-news/ 0200-studies-reports/upload/CATitleCompetitionReport0512Public.pdf

The Chicago Bar Association v. Quinlan and Tyson, Inc. 34 Ill. 2d 117, 1966. Dumm, R. E., Macpherson, D. A., & Sirmans, G. S. (2007). The title insurance industry: Examining a

decade of growth. Journal of Insurance Regulation, 25(4), 24–51. Eaton, J. W. (2009, May 28). Title insurance reform: A long overlooked target of opportunity—The governor

and the legislature should embrace it. Testimony to the hearing on title insurance rate regulation under the auspices of the Pennsylvania Insurance Commission. Harrisburg, PA.

Eaton, J., & Eaton, D. (2007). The American Title Insurance Industry: How a Cartel Fleeces the American Consumer. New York: NYU Press.

Elder, Harold, Zumpano, Leonard, & Baryla, Edward. (2000). Buyer-brokers: Do they make a difference? Their influence on selling price and search duration. Real Estate Economics, 28(2), 337–362.

Jaffee, D. (2006). Monoline restrictions, with applications to mortgage insurance and title insurance. Review of Industrial Organization, 28(2), 83–108.

Lacko, James, & Pappalardo, Janis. (2010). The failures and promise of mandated consumer mortgage dis- closures: Evidence from qualitative interviews and a controlled experiment with mortgage borrowers. The American Economic Review, 100(2), 516–521.

Lewis, H. (2014). Closing Costs Study. Bankrate.com. Retrieved from: http://www.bankrate.com/finance/ mortgages/closing-costs/closing-costs-by-state.aspx

McKernan, S., Wissoker, D., Ratcliffe, C., Van Order, B., & Scott, M. (2009). Descriptive analysis of FHA settlement charges. Report prepared for the U.S. Department of Housing and Urban Development. Washington, DC: The Urban Institute.

Owen, B. M. (1977). Kickbacks, specialization, price fixing, and efficiency in residential real estate markets. Stanford Law Review, 29(5), 931–967.

Roussel, H. L., & Rosenberg, M. K. (1981). The high price of ‘Reform’: Title insurance rates and the benefits of rating bureaus. Journal of Risk and Insurance, 48(4), 638–648.

Turner, M. A., Ross, S. L., Galster, G. C., & Yinger, J. (2002). Discrimination in metropolitan housing markets: National results from phase I HDS2000. Report prepared for the U.S. Department of Housing and Urban Development. Washington, DC: The Urban Institute.

U.S. Government Accountability Office. (2007). Title insurance: Actions needed to improve oversight of the title industry and better protect consumers. Washington, DC: GAO.

White, L. J. (1984). The title insurance industry, reverse competition, and controlled business: A different view. Journal of Risk and Insurance, 51(2), 308–319.

Woodward, S. E. (2008). A study of closing costs for FHA mortgages. Report prepared for the U.S. Depart- ment of Housing and Urban Development, Washington, DC: The Urban Institute.

Woodward, S. E., & Hall, R. (2010). Consumer confusion in the mortgage market: Evidence of less than a perfectly transparent and competitive market. American Economic Review: Papers and Proceedings, 100, 511–515.

123

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.