Clientele Effects And Condo Conversions

2008

V36

3: pp. 611–634

REAL ESTATE ECONOMICS

Clientele Effects and Condo Conversions John D. Benjamin,∗ Peter Chinloy,∗∗ William G. Hardin III∗∗∗ and Zhonghua Wu∗∗∗∗ During an asset boom a property can develop a new usage. Appreciation investors emerge to change a property’s occupant mix or letter grade. Rental investors not intending to change the mix or grade are outbid. Sellers receive a capitalized premium from the new type of bidder. For apartments in MiamiFort Lauderdale during 2004–2006, there is an asset pricing premium from condo converters. The price of real estate depends on clienteles in addition to characteristics.

In real estate there are different clienteles bidding for a property. Rental investors retain the existing occupancy and letter grade. Appreciation investors intend to change the existing usage and occupancy of the property, such as changing the letter grade from Class C to Class B. In equilibrium, properties are in grade or type categories. The asset price difference between classes is equal to the cost of investment to make the grade difference. Appreciation investors emerge as successful bidders when a change in systematic risk raises their expected returns. Condo converters appeared in the real estate market after tax law changes in the 1980s, and during the abnormally low interest rate period after 2003. If converters outbid rental investors, there is an asset pricing characteristic in apartments that depends on the buyer’s type and there is a premium in the asset price by clientele. The asset pricing premium in multifamily buildings is what converters are willing to pay over and above those interested only in renting the apartments to the same tenant mix. A separate asset pricing premium or discount is how the housing market is affected by an increase in condo conversions. There are different implicit values attached to the same property by owner types. The differences between owners lie in their return objectives and intentions ∗

Kogod School of Business, American University, Washington, DC 20016 or [email protected]. ∗∗ Kogod School of Business, American University, Washington, DC 20016 or [email protected]. ∗∗∗ Department of Finance and Real Estate, Florida International University, Miami, FL 33199 or [email protected]. ∗∗∗∗ Department of Finance and Real Estate, Florida International University, Miami, FL 33199 or [email protected].  C 2008 American Real Estate and Urban Economics Association

612 Benjamin et al.

regarding capital expenditures and operation. Market shocks cause one group to dominate as actual buyers with predictable effects on prices and returns. These clientele effects lead to apartment buildings being competed for by rental investors and appreciation-oriented condo converters. Rental investors’ expected returns come from collecting lease payments. The property’s clientele is existing tenants by grade or classification. Rental investors plan limited capital expenditures and do not expect to change the quality of the property. Rental investors anticipate limited capital expenditure apart from ongoing maintenance. Their return comes from evaluating the cap rate, or ratio of net rent to price at the current occupancy. Appreciation investors have expected returns from capital gains in upgrading a property. The appeal is to higher quality occupants than the current tenant base. Appreciation investors plan extensive capital expenditures to upgrade the quality of the property. Condo converters acquiring apartments are appreciation investors. Appreciation investors anticipate extensive capital expenditure. Their intention is to change and upgrade the type of occupant and property, including its usage. These investors’ returns come from the expected capital gain. The expected capital gain is the difference in price derived from the change in net revenues above what occupants will pay for a property’s current usage. In the apartment market, rental investors compete with condo converters for property. An exogenous shock, such as monetary or fiscal policy measures favoring homeownership, causes condo converters to raise their bids for existing property. As their bids increase, there is a capitalized asset price benefit. Eventually, the asset price level rises to a level that reduces expected returns. Over time, the asset pricing premium disappears. Price increases and returns over time rise and then fall. The volume of sales is positively correlated with this price pattern, with the high-expected return clientele’s proportion of purchases rising and then falling. Prices and volume associated with the new clientele of converters are positively correlated. Pricing, returns and volume depend on the characteristics of market participants. Hedonic physical characteristics alone are insufficient determinants of the transaction price. Condo conversions soared in South Florida during 2004–2006. There were 62,000 apartment units converted to condos during this time period according to the real estate advisory service REIS and the brokerage firm Marcus & Millichap. More than one in every five apartments in South Florida was converted to condos in a 3-year window. From REIS, 20.7% of the stock of apartments in Miami-Fort Lauderdale and 22.9% in West Palm Beach were converted. Condo conversions were geographically concentrated around the

Clientele Effects and Condo Conversions 613

United States. In Washington, DC, the apartment-to-condo conversion rate in that 3-year period was 21.7%, while it was 19.5% in Orlando and 13.5% in Tampa. Las Vegas, San Diego and Jacksonville saw nearly 10% of their apartment stock similarly converted during the same 2004–2006 window. The application is focused on the Miami-Fort Lauderdale apartment market, the center of this conversion activity. The sample is from the third quarter of 2004 to the last quarter of 2006. During this 10-quarter period, rental and appreciation investors were bidding for apartment buildings in the Miami-Fort Lauderdale market. The premium that converters as appreciation investors paid over their rental counterparts is estimated on a quarterly basis. Over the entire 10-quarter window, converters paid a premium of 7% over rental investors. The premium is higher in the larger submarkets of Miami Beach and downtown Fort Lauderdale. In the second and third quarters of 2005, converters paid 17% more than rental investors for apartments. By early 2006 the premium declines to zero and later in the year there are few sales to converters. The competition between converters and rental investors leads to a sharp rise in apartment prices per unit. Adjusted for quality and neighborhood, apartment prices overall in Miami-Fort Lauderdale rose by 39% in 2005 regardless of buyer. In 2006, hedonic apartment prices rose by a considerably more modest 3.5%. The presence of condo converters led all buyers to pay more, including those who intended to operate the properties for rental. Real estate has a clientele effect in pricing and returns that is separate from conventional physical hedonics. Prices and returns depend on the identity of the owner. The next section describes real estate markets where parties with different characteristics bid for the same property. The third section includes the model where rental and appreciation investors compete. The data and empirical results on Miami-Fort Lauderdale apartments are presented in the fourth section. Concluding remarks are included in the last section. Background Instead of property hedonics alone, characteristics of the buyers or sellers are part of the asset pricing puzzle. One participant characteristic that has been shown to affect price is the bargaining strength of individual buyers and sellers. Colwell and Munneke (2006) find that bargaining strength impacts commercial real estate transactions.1 The application is to the office market in Chicago. 1 If D buy is a dummy variable representing a type of buyer and D sell is another for a seller, then the included term in a hedonic regression with the price or its logarithm as dependent variable is α[D buy − λD sell ]. If α > 0 a buyer type has the advantage in bargaining, and λ accounts for a seller performing differently on buying versus selling.

614 Benjamin et al.

Another characteristic of buyers and sellers is locality. Lambson, McQueen and Slade (2004) find that for apartments in Phoenix nonlocals pay more than local buyers because of higher search costs and an anchoring bias. The anchoring bias occurs particularly when the prices in a nonlocal buyer’s resident market are higher than in the target market. A buyer who lives in a market where apartments are more expensive than in Phoenix uses the home reference point and pays more. Malpezzi and Shilling (2000) find that institutional investors have tilts in the types of properties that they prefer. Hardin and Wolverton (1999) show that real estate investment trusts have at times paid acquisition premiums over other types of buyers. Harding, Rosenthal and Sirmans (2003a, b) show that buyer and seller characteristics influence real estate prices in the residential market. The physical hedonic characteristics that determine price are expanded by adding variables for the buyer and seller. One pair of dummy variables enters with opposite signs for buyer-seller types indicating differential bargaining strength. Another pair has the same signs representing property type effects not measured by physical characteristics. In the housing market, some sellers receive less than others because they are different bargainers, not because their houses are different. Rental investors are focused on yields, the property cap rate and long-term cash flows that determine per unit price. Appreciation investors, such as condo converters emphasize capital gains conditional on making improvements. Characteristics that assign investors to rental or appreciation status are tax rates, holding period, agency and whether investors are locally focused. Rental investors seek income and operating scale. They are not seeking to make extensive capital expenditure decisions particularly if distantly located, resulting in agency costs of monitoring. This conflict is acute in the apartment market because the two participant types have divergent property strategies. Appreciation investors as converters are active. They profit by buying and transforming apartments into condo units for sale. Converters provide skills in entitlement, renovation and marketing. The holding period is relatively short to match the conversion and sell-out phase. The investment is taxed as ordinary income instead of preferential capital gains. Other drivers for conversion include the relative costs of owning versus renting, demand for housing ownership, rent regulations such as rent control and low rates of rental housing starts. The different clienteles are motivated by shifts in asset price expectations and tax policy. Previous research on condo conversion has not emphasized investor clienteles and has focused on real estate booms driven by tax policy as opposed to pricing expectations. In Steele (1993), conversions come in relatively new apartment

Clientele Effects and Condo Conversions 615

buildings. High taxes on rental income and low taxes on capital gains induce property owners to sell to converters.2 Existing apartment residents, who are older, with children and higher incomes, have higher probabilities of purchasing their units as conversions. Because converters shrink the rental apartment supply, remaining multifamily property owners’ benefit. As the conversion cycle continues or as individual condos are acquired by small-scale investors, the increased rental inventory competes with larger scale apartment buildings. While prior studies have documented price and availability effects of real estate conversions primarily in the aggregate or with regard to rents, the present study evaluates investor clienteles and provides evidence of their impact on transaction prices and returns. The role of expectations between clienteles drives the differentials. Returns and Expectations In the real estate market, there are i = 1, . . . , I different investor clienteles, including rental and appreciation investors. In an initial market equilibrium, clienteles have optimized with respect to systematic risks such as tax policy and interest rates. This sorting leads to properties being in different types and grades, such as Class A, B and C. Shifting a property from one grade to another is the objective of appreciation investors. An unanticipated policy event, such as lower interest rates and available financing that favor homeownership leads to added bidding by appreciation investors. A model of an asset pricing premium for apartments that capitalizes the entry of new appreciation investors can be developed. The seller of the existing apartment earns a premium when the converter arrives. Analogously, land prices and values are affected by when a different developer usage emerges. The observation window is over time t = 1, . . . , T . Tenants pay a market rent Rt regardless of the landlord. In the short run, the inventory of the property type is fixed at S¯ t . The demand for rental space is Qt (Rt , Yt ), where Yt indicates income and other tenant characteristics. In the rent-quantity space market (Rt , Qt ), there is an equilibrium vacancy rate of v ∗t . 2

Previous condo conversions were motivated by tax considerations as opposed to clientele competition resulting from a fall in interest rates. That makes the condo conversions up to 2000 have a different systematic risk factor than those after 2000. In Whinihan (1984) increases in rental housing taxes for landlords increase the incentives to convert multifamily residential properties to condominiums. Rental housing landlords have high marginal tax brackets, so that further increases in marginal taxes and reduced capital gains taxes are followed by increases in condominium conversions. Diskin and Tashchian (1984) find these results from a logit analysis of condo conversions in the Atlanta metro area. Another factor driving conversion is avoidance of rent restrictions (Hansmann 1991). Tenant subsidies are discussed in Benjamin, Chinloy and Sirmans (2000).

616 Benjamin et al.

Rental investors form expectations about rents. Actual rent growth is rt = Rt+1 −Rt . Because the rent in the next period R t+1 is not observable, investors Rt form expectations about those rents as Et (R t+1 ). Expected rental growth is rte = Et (Rt+1 )−Rt . An owner subscript on Et (R t+1 ) is suppressed because only rental Rt investors have these expectations. Appreciation investors have no expectations about rent growth. The equilibrium vacancy rate satisfies the expectationsaugmented structure rt − rte = γ0 + γ1 vt ,

(1)

where γ 0 ≥ 0 and γ 1 ≤ 0 are parameters. The structure is as in Rosen and Smith (1983). The rent-vacancy trade-off is negatively sloped. If expectations are correct and equal to actual rent increases, then the equilibrium vacancy rate is vt∗ = − γγ01 ≥ 0. The rental market is in equilibrium when   Qt (Rt , Yt ) = 1 − vt∗ S¯ t .

(2)

Expectations on property use translate into different asset pricing beliefs. The t , but as with rents asset price of the property is Pt . Capital gains are pt ≡ Pt+1P−P t the one-period-ahead price P t+1 is not observable. Investors form expectations of the subsequent price when they are bidding in the current or spot market. Sellers see these bids enhanced if appreciation investors are in the market for purchase. The asset market is for apartment buildings. A separate asset market, which is not part of the model, is for completed condos. Investors are competing to buy existing apartment buildings. The price in the current market that investor clientele i is willing to pay is Pit . As appreciation and rental investors compete, sellers in the current apartment market may benefit if conditions favor converters. If converters are willing to pay a higher price, a premium emerges by type of buyer. That premium is observed in current spot price data. The structure develops the pricing premium by investor clientele in the current market, based on the expectations of the future. Investors have heterogeneous price expectations about the value of the apartment building as Eit (P it+1 ). Appreciation investors expect that the one-periodahead price will be higher than rental investors, as value-added investments are planned. A shock that raises appreciation investors’ price expectations causes them to increase their bidding in the spot apartment market. The expected )−Pit . capital gain differing across investors is pite ≡ Eit (Pit+1 Pit For apartments, rental investors and condo converters find themselves competing against each other. Expected capital gains are pit − pite = βi0 − βi1 Pit

βi0 ≥ 0 βi1 ≥ 0.

(3)

Clientele Effects and Condo Conversions 617

The observed price in the market is Pt = maxi Pit . The higher the price a current bidder is willing to pay in Pit , the greater the capital gains expectation. If a given investor group’s expectation of the capital gains is accurate, then pit − pite is zero. The equilibrium bid price for the property is Pit∗ =

βi0 ≥ 0. βi1

(4)

The price of an asset in equilibrium is the ratio of two statistics that effectively correspond to net operating income (NOI) β i0 divided by a yield, or cap rate β i1 . The NOI-cap rate ratio is embedded in the equilibrium structure. When rental investors are the successful bidders there is no added premium by buyer type. Then β i0 is the NOI measured as the intercept in the asset pricespace market and β i1 is the cap rate as the slope. If condo converters are the successful bidders, they view β i0 as the cash flow return and β i1 as the capital gain they require per dollar spent. Expressing the price expectations (3) in terms of the price level Pit =

 1  βi0 − pit + pite . βi1

(5)

Equation (5) summarizes how price expectations affect the actual level of prices. Actual capital gains pit are tied to the physical characteristics Xt . Expected capital gains pite differ across investor groups and are time-dependent. These variables can be represented by dummy variables Zi with coefficients δ t the clientele asset pricing premium. If the premium is zero, there is no difference in the price of an apartment from the location, type or nature of the buyer and only physical characteristics affect prices. Figures 1–4 describe the market for rental space and the market for the property as an asset in the short term, along with the equilibrium adjustments and expectations for rents and prices. Figure 1 shows the property market in long-run equilibrium. The market is in long-run equilibrium for space when expectations about rent increases are correct and fulfilled. On the right side is the space market. The demand function is downward sloping as Qt (Rt , Yt ) ¯ Expected rent increases are the actual rent growth, so and the inventory is S. the equilibrium vacancy is v ∗ . The space market clears at effective gross rent R(1 − v). Net rent is the NOI tied to gross rent. On the left side is the market for the asset. In equilibrium, investors have the same and correct expectations about capital gains. If the successful bidders are rental investors, then the NOI together with the correct capital gains expectation produces an asset-market equilibrium. The clearing price for the asset is the dark line for P .

618 Benjamin et al.

Figure 1  Equilibrium rent and asset price: expectations fulfilled.

Note: On the right graph is a quantity measure on the horizontal axis, such as square feet. The vertical axis is the price of services or rent per square foot. The demand is downward sloping and supply is perfectly inelastic in the short run. Frictional and other factors lead the gross rent to an equilibrium vacancy rate; the rent does not equate supply and demand. The gross rent corresponds to effective gross income; a lower level of net rent or net operating income is determined by the operator’s efficiency. On the left graph, the net rent divided by the cap rate or the slope of the line is the equilibrium asset price. An equilibrium vacancy, rent and asset price are determined together in the spot market with expectations fulfilled.

Equation (4) shows the conventional asset price as the ratio of NOI to the cap rate in an equilibrium structure with rent and asset price expectations fulfilled. In Figure 2, that situation for rent is where r = r e and rent increases are anticipated, determining the equilibrium vacancy v ∗ . In Figure 3, when p = pe capital gains are anticipated and the winning bidders are rental investors. The intercept on the vertical axis is the NOI and the slope is the cap rate. The intercept with the horizontal axis is the clearing price. That equilibrium is disturbed if expectations about asset prices and capital gains are not fulfilled. Then the asset price is not equal to the NOI divided by a conventional cap rate. Figure 4 indicates where rental and appreciation investors have different expectations about capital gains. On the vertical axis is the difference between actual

Clientele Effects and Condo Conversions 619

Figure 2  Adjustment, rent increase expectations.

Note: On the horizontal axis is the vacancy rate. The rental increase shock, as the difference between the actual and expected growth rate is on the vertical axis. When the rent growth rate is expected, vacancy is at its equilibrium level along the vertical axis. That equilibrium vacancy rate that eliminates rent growth shocks is where the function intersects the horizontal axis.

Figure 3  Expected capital gains and equilibrium price.

Note: On the horizontal axis is the asset price level. The vertical axis is the appreciation rate shock, the difference between actual and expected price growth. When price growth is anticipated, the function intersects with the horizontal axis, leading to the equilibrium price level.

620 Benjamin et al.

Figure 4  Disequilibrium in asset market rental and appreciation investors.

Note: The horizontal axis is the price level and the vertical axis the unanticipated appreciation. Rental and appreciation investors have different functions representing their expected appreciation for the property. The boldface function is for appreciation investors. They outbid rental investors at relatively high prices. In the figure, the appreciation investors bid a price level above the equilibrium that would satisfy or make sense in the rental market.

and expected rates of property price appreciation. The light-shaded negatively sloped line is the trade-off of capital gains and price level p 1 for rental investors. If rental investors correctly forecast property appreciation, then p = pe , and this group bids P ∗1 for the asset. This P ∗1 is the clearing price when the property is sought for rental income, as in Figure 1. With appreciation investors’ expectations having changed favorably after a demand shock, this is no longer the case. Appreciation investors have a trade-off (p − pe , P ) between capital gains pricing error and price level of p 2 , the dark-shaded line in Figure 4. Converters bidding a price such as P 2 in Figure 4 are successful, because that price is above the equilibrium bid P ∗1 from rental investors. There is a positive price effect or pricing premium by ownership type with appreciation investors as converters paying more. This process continues dynamically until eventually converters bid a price P 2 past where the dark-shaded line intersects the horizontal axis. At this stage, the converter has unanticipated capital losses. The price premium by ownership has an inverse U shape and is not permanent. The price premium is for the asset, here apartment buildings.

Clientele Effects and Condo Conversions 621

This is a model of apartment buildings for rental or purchased for conversion. It is not a model of condo sales and their impact on the housing market. In those markets there is potentially a separate premium. The converter is bidding for the building as an input in its production process. This bidding leads to an asset pricing premium for apartments. The price path for an asset does not have hedonic characteristics alone. Instead, owner characteristics can affect price. These owner characteristics vary over time and may be linked with changing capital market flows. Implementation of pricing structures requires knowledge of the characteristics of the buyers and sellers of the real estate asset as well as the characteristics of the property. In the application, properties are being sought by two investor groups, rental and appreciation investors. Appreciation investors convert properties to condos. Each group has an information set I j,t,t−1 , j = 1, 2 at date t. Bidding continues on individual properties during the time window with the groups competing against each other. The first test of the structure is whether one investor group has superior private information about the property. If the converters win the bidding process the property goes off the rental market as opposed to its current use. The decision at date t to convert a given property is D ∗t (I t,t−1 ). Parameters are α 1 and disturbance υ 1 with zero mean. Variables Wt describe the clientele in the condo conversion bidding group. If the converter wins the bidding then ⎧ ∗ ⎪ ⎨Dt (It,t−1 ) = Wt α1 + υ1t D1,t−1 = 1 Wt−1 α1 + υ1,t−1 > 0 (6) ⎪ ⎩ D1,t−1 = 0 Wt−1 α1 + υ1,t−1 ≤ 0. The indicator variable is D 1,t−1 = 1 if the property is won by a converter given information I t,t−1 . The indicator is D 1,t−1 = 0 if the rental investor wins. As the decision to accept an offer is contingent on willingness to sell, this self-selection must be addressed. The expected value of the disturbance is E(υ2t | υ1t ≥ −W1t α1 ) = λ

f (−W1t α1 ) ≡ λm1t . 1 − F (−W1t α1 )

(7)

Adjusting for the conditional expectation of allowing the asset to be sold, the sale price is determined by Pt = Xt α2 + λm1t + Zi δt + υ2t .

(8)

Equation (8) is the empirical implementation of the expectations structure (5) with self-selection. It tests for whether converters have superior private information and whether there is a premium by clientele. If converters have superior

622 Benjamin et al.

private information, then λ = 0. If there are differences between clienteles in price expectations, then δ t = 0. Observing the sequence of coefficients on clienteles, the premiums can vary over time. In an inverse U shape δ t are rising and then falling during a time window. Data and Empirical Results The sample consists of 1,236 apartment sales in the Miami-Fort Lauderdale metropolitan statistical area (MSA) from the third quarter of 2004 to the last quarter of 2006.3 The ten-quarter data are from the commercial real estate information company CoStar and include transactions on multifamily property sales during the window.4 CoStar tracks and confirms property sales through contacts with transaction participants and the use of required public filings and records. There are two groups of apartment purchasers. A majority of the buyers are investors primarily seeking rental income and long-term investment returns. A second group of buyers are investors acting as condo converters. Condo converters are identified by CoStar during their transaction confirmation process. These buyers terminate the income investment by transforming the rental units to condos and selling the units individually. The hypothesis is that capital gains expectations differ between the parties, particularly when short-term asset prices are rising.5 An initial test is whether converters have superior private information, despite their identities being known. This self-selection involves running a probit model where the dependent variable is a dummy variable for property conversion. Property characteristics and ownership characteristics are included in the initial probit model. The price equations investigate both self-selection and clientele effects. The testable construct is whether converters pay different prices than rental investors. Characteristics such as age, size and location are included as control variables. Various configurations of the converter variable are presented to evaluate the robustness of the findings. The converter premium δ t need not be static over time as both the coefficient and the self-selection factors are 3

According to National Real Estate Investor, the south Florida markets of Miami/Dade, Broward/Palm and Tampa represented 30% of the $2.6 billion of apartments acquired for condominium conversion in 2003. The data used in the study come from what is likely a census of transactions. Observations with missing data are excluded from the initial census, which creates the data set with 1,236 observations for the analysis.

4

CoStar started differentiating sales as conversions in these markets beginning in mid 2004. Designation is made through the transaction confirmation process.

5

The market for condo resales in South Florida turned downward in 2006. In December 2006, the South Florida Multiple Listing Service showed in Dade County there were 7,500 condo listings at a price greater than $400,000 and there were 155 sales.

Clientele Effects and Condo Conversions 623

allowed to vary across several of the models presented. Whether there is a price-volume correlation when converters are active involves examining the time series of condo conversion premiums. This is done by using interaction variables identifying property conversion over the 10 quarters. Table 1 shows basic variables including building condition and submarket location. The temporal distribution of conversion sales is in Table 2. The results of

Table 1  Miami-Fort Lauderdale MSA apartment sales.

Acres Age Property sales price ($1,000) Price/unit ($1,000) Floors Units Square feet Condo conversion Buyer nonlocal Seller nonlocal Condition excellent Condition good Condition average Condition fair Quarter 2004:3 Quarter 2004:4 Quarter 2005:1 Quarter 2005:2 Quarter 2005:3 Quarter 2005:4 Quarter 2006:1 Quarter 2006:2 Quarter 2006:3 Quarter 2006:4 Hialeah Gardens Outlying Broward West Miami Southwest Broward South Dade Sawgrass Park Pompano Beach Plantation Northeast Dade Northwest Broward Miami Lakes

Mean

Standard Deviation

Minimum

Maximum

1.86 43.60 6,136.80 107.26 2.35 52.49 728.04 0.12 0.08 0.11 0.01 0.02 0.92 0.03 0.14 0.12 0.13 0.15 0.11 0.07 0.08 0.01 0.06 0.03 0.01 0.01 0.01 0.01 0.02 0.00 0.05 0.02 0.07 0.02 0.01

4.90 17.40 15,827.86 54.19 2.48 112.88 291.89 0.32 0.28 0.31 0.09 0.16 0.25 0.18 0.34 0.33 0.33 0.35 0.31 0.25 0.27 0.29 0.23 0.17 0.09 0.09 0.07 0.09 0.13 0.04 0.22 0.13 0.26 0.15 0.08

0.02 0.00 400.00 12.70 1.00 5.00 63.62 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

48.30 90.00 160,000.00 438.46 39.00 1,300.00 3,916.40 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

624 Benjamin et al.

Table 1  continued

Miami Beach Miami Airport Miami Medley/Hialeah Kendall Hollywood Hallandale Fort Lauderdale Downtown Miami Downtown Fort Lauderdale Cypress Creek Coral Way Coral Gables Commercial Boulevard Coconut Grove Brickell Biscayne Corridor Aventura

Mean

Standard Deviation

Minimum

Maximum

0.16 0.03 0.12 0.03 0.01 0.09 0.02 0.14 0.00 0.06 0.01 0.01 0.02 0.01 0.01 0.01 0.03 0.01

0.37 0.17 0.33 0.20 0.11 0.29 0.15 0.35 0.05 0.24 0.08 0.11 0.13 0.08 0.11 0.11 0.16 0.06

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

Note: Sales data are for a ten-quarter window beginning July 1, 2004 and ending December 31, 2006. Data are from the Miami-Fort Lauderdale real estate market. The data are provided by CoStar.

Table 2  Distribution of conversion sales by quarter.

Quarter

Total Sales

Conversion Sales

% Conversion Sales

Ratio of Conversion Sales to Peak Conversion Sales

Quarter 2004:3 Quarter 2004:4 Quarter 2005:1 Quarter 2005:2 Quarter 2005:3 Quarter 2005:4 Quarter 2006:1 Quarter 2006:2 Quarter 2006:3 Quarter 2006:4

172 155 160 188 142 86 104 119 71 39

12 22 21 22 28 13 8 11 7 8

6.98 14.19 13.13 11.70 19.72 15.12 7.69 9.24 9.86 20.51

42.86 78.57 75.00 78.57 100.00 46.43 28.57 39.29 25.00 28.57

Note: Sales data are for a ten-quarter window beginning July 1, 2004 and ending December 31, 2006.

Clientele Effects and Condo Conversions 625

Table 3  Probability of converter purchase with probit estimates. Model Variable

Coefficient

Standard Error

Intercept Acres Age Building quality (E or G) Buyer nonlocal Seller nonlocal Floors (number) Units (100s) Unit size (1,000SF) Log likelihood McFadden R2 N

−1.512∗∗∗ 0.027∗ −0.007∗ 0.364 −0.635∗∗∗ 0.081 0.045∗ 0.163∗∗ 0.428∗∗ 155.565 0.169 1,236

0.239 0.016 0.004 0.230 0.220 0.165 0.024 0.001 0.001

Note: The dependent variable is a dummy variable reflecting status as a conversion property. The variables of interest are the hedonic property variables and purchaser characteristic variables. Sales data are for a ten-quarter window beginning July 1, 2004 and ending December 31, 2006. ∗∗∗ Significant at the 1% level. ∗∗ Significant at the 5% level. ∗ Significant at the 10% level.

the probit model delineating the factors that impact the selection of a particular apartment complex for conversion are in Table 3. The results for the regression models on per unit price and the impact of conversion on price are in Tables 4 and 5. From Table 2, the peak in condo acquisitions is the third quarter of 2005 when 28 complexes were acquired for conversion. The quarter has the highest percentage of conversions; 19.7% of sales were acquired for this purpose. The data show apartment sales decreased substantially in the latter half of 2006 as converters leave the market. The property and ownership characteristics associated with condo conversion properties are assessed using a probit model with results presented in Table 3. The probability of a property being bought as a condo conversion depends both on the physical and owner characteristics. Newer buildings with larger units have a higher probability of being bought by a converter, and facing competition in bidding. Having 1,000 square feet in a unit has a 0.43 impact on the probability of conversion. The age coefficient is negative, at −0.007 per year, indicating that older buildings are less attractive to converters. Nonlocal buyers are rental and not appreciation investors. The coefficient on nonlocal buyer is negative and significant at the 1% level. This is a higher level of significance than any of the physical characteristics. The condo converter is

T Statistic 17.80 2.75 −1.92 2.71 −1.83 1.38 3.53 0.91 7.36 2.29 2.15 1.44 5.24 8.48 8.36 8.97 8.89 8.29 8.65 7.31

Coef.

8.859∗∗∗ 0.016∗∗∗ −0.003∗ 0.226∗∗∗ −0.250∗ 0.051 0.029∗∗∗ 0.030 0.737∗∗∗ 0.072∗∗ 0.549∗∗ 0.053 0.189∗∗∗ 0.295∗∗∗ 0.313∗∗∗ 0.390∗∗∗ 0.364∗∗∗ 0.327∗∗∗ 0.403∗∗∗ 0.425∗∗∗

Variable

Intercept Acres Age (years) Building quality (E or G) Buyer nonlocal Seller nonlocal Floors (number) Units (100s) Unit size (1,000SF) Conversion = 1 Inverse mills ratio λ Quarter 2004:4 Quarter 2005:1 Quarter 2005:2 Quarter 2005:3 Quarter 2005:4 Quarter 2006:1 Quarter 2006:2 Quarter 2006:3 Quarter 2006:4

Model 1

17.2 3.13 −2.07 2.77 −2.00 1.43 3.89 0.77 7.43 2.36 1.89 5.42 8.23 7.90 9.19 9.18 8.34 8.66 7.42

0.610∗∗∗ 0.072∗ 0.205∗∗∗ 0.299∗∗∗ 0.316∗∗∗ 0.422∗∗∗ 0.390∗∗∗ 0.341∗∗∗ 0.422∗∗∗ 0.473∗∗∗

T Statistic

8.727∗∗∗ 0.018∗∗∗ −0.003∗∗ 0.235∗∗∗ −0.274∗∗ 0.054 0.033∗∗∗ 0.026 0.756∗∗∗

Coef.

Model 2 9.031∗∗∗ 0.017∗∗∗ −0.002 0.205∗∗ −0.240 0.057 0.027∗∗∗ 0.011 0.725∗∗∗ 0.093∗∗∗ 0.828∗

Coef.

Model 3

16.90 2.83 −1.39 2.29 −1.61 1.43 3.05 0.30 6.77 2.72 1.94

T Statistic 8.771∗∗∗ 0.016∗∗∗ −0.003∗∗ 0.212∗∗ −0.283∗∗ 0.060 0.001∗∗∗ 0.039 0.698∗∗ 0.085∗∗ 0.603∗∗ 0.062 0.193∗∗ 0.297∗∗∗ 0.311∗∗∗ 0.409∗∗∗ 0.364∗∗∗ 0.330∗∗∗ 0.400∗∗∗ 0.444∗∗∗

Coef.

Model 4

Table 4  Apartment pricing and conversions, Miami-Fort Lauderdale MSA 2004–2006 Using Data from All Submarkets.

17.71 2.88 −2.15 2.48 −2.09 1.61 3.92 1.22 7.15 2.41 2.38 1.52 4.88 7.79 7.62 8.99 7.61 7.61 8.52 7.42

T Statistics

626 Benjamin et al.

1,236 28.318 0.509

Conversion ∗ 2004:3 Conversion ∗ 2004:4 Conversion ∗ 2005:1 Conversion ∗ 2005:2 Conversion ∗ 2005:3 Conversion ∗ 2005:4 Conversion ∗ 2006:1 Conversion ∗ 2006:2 Conversion ∗ 2006:3 Conversion ∗ 2006:4 % Conversion N F value Adj. R2

T Statistic ∗∗∗

1,236 24.078 0.511

0.306 0.032 0.065 0.160∗∗ 0.138∗ −0.063 −0.084 0.053 0.004 −0.087

Coef.

Model 2

2.98 0.41 0.81 2.12 1.95 −0.62 −0.69 0.51 0.03 −0.67

T Statistic

0.990∗∗∗ 1,236 27.658 0.427

Coef.

Model 3

3.94

T Statistic

1,236 26.131 0.488

Coef.

Model 4 T Statistics

Note: The dependent variable is the log of the per unit sales price. The variables of interest are conversion, the quarterly conversion percentage of sales, and the converter interactions over the 10 quarters. Models 1, 2 and 3 are OLS models and Model 4 is a GLS model estimated with the same variables as presented in the baseline Model 1. ∗∗∗ Significant at 1%. ∗∗ Significant at 5%. ∗ Significant at 10%.

Coef.

Model 1

Variable

Table 4  continued

Clientele Effects and Condo Conversions 627

T Statistic 17.77 2.84 −0.85 3.48 −2.02 1.55 3.88 1.21 7.51 2.04 2.46 1.21 4.69 7.43 6.93 8.93 7.70 6.03 7.19 6.46

Coef.

9.689∗∗∗ 0.018∗∗∗ −0.001 0.315∗∗∗ −0.297∗∗ 0.061 0.034∗∗∗ 0.043 0.836∗∗∗ 0.068∗∗ 0.676∗∗ 0.049 0.186∗∗∗ 0.290∗∗∗ 0.290∗∗∗ 0.415∗∗∗ 0.365∗∗∗ 0.272∗∗∗ 0.390∗∗∗ 0.414∗∗∗

Variable

Intercept Acres Age Building quality (E or G) Buyer nonlocal Seller nonlocal Floors (number) Units (100s) Unit size (1,000SF) Conversion = 1 Inverse Mills ratio λ Quarter 2004:4 Quarter 2005:1 Quarter 2005:2 Quarter 2005:3 Quarter 2005:4 Quarter 2006:1 Quarter 2006:2 Quarter 2006:3 Quarter 2006:4

Model 1

17.35 3.21 −1.03 3.56 −2.24 1.66 4.22 1.14 7.66 2.71 1.45 4.79 6.91 6.26 9.16 7.90 5.92 6.90 6.31

0.751∗∗∗ 0.062 0.198∗∗∗ 0.283∗∗∗ 0.282∗∗∗ 0.451∗∗∗ 0.390∗∗∗ 0.279∗∗∗ 0.394∗∗∗ 0.450∗∗∗

T Statistic

9.563∗∗∗ 0.021∗∗∗ −0.002 0.324∗∗∗ −0.330∗∗ 0.066 0.038∗∗∗ 0.041 0.860∗∗∗

Coef.

Model 2 9.858∗∗∗ 0.021∗∗∗ −0.001 0.298∗∗∗ −0.295∗ 0.064 0.033∗∗∗ 0.022 0.803∗∗∗ 0.093∗∗∗ 0.662∗∗

Coef.

Model 3

16.85 3.04 −0.48 3.08 −1.88 1.53 3.56 0.59 6.76 2.59 2.26

T Statistic 10.661∗∗∗ 0.005 0.002 0.144∗∗ 0.009 0.030 0.026∗∗∗ −0.000 0.590∗∗∗ 0.084∗∗ 0.128 0.063 0.202∗∗∗ 0.310∗∗∗ 0.293∗∗∗ 0.452∗∗∗ 0.377∗∗∗ 0.270∗∗∗ 0.369∗∗∗ 0.434∗∗∗

Coef.

Model 4

26.76 1.11 1.49 2.12 0.10 0.82 3.03 −0.70 7.20 2.19 0.71 1.30 4.61 7.19 6.33 9.06 6.46 5.42 6.79 5.99

T Statistics

Table 5  Apartment pricing and conversions, Miami-Fort Lauderdale MSA 2004–2006 using data from submarkets with more than three condo conversions.

628 Benjamin et al.

942 22.233 0.441

Conversion ∗ 2004:3 Conversion ∗ 2004:4 Conversion ∗ 2005:1 Conversion ∗ 2005:2 Conversion ∗ 2005:3 Conversion ∗ 2005:4 Conversion ∗ 2006:1 Conversion ∗ 2006:2 Conversion ∗ 2006:3 Conversion ∗ 2006:4 % Conversion N F value Adj. R2

T Statistic

Model 2

942 39.849 0.723

0.241∗∗ 0.034 0.044 0.181∗∗ 0.145∗ −0.115 −0.110 0.067 0.096 −0.037

Coef. 2.04 0.38 0.51 2.30 1.94 −1.13 −0.86 0.61 0.63 −0.27

T Statistic

Model 3

1.109∗∗∗ 942 23.734 0.655

Coef.

3.94

T Statistic

Model 4

942 19.584 0.408

Coef.

T Statistics

Note: The dependent variable is the log of the per unit sales price. The variables of interest are conversion, the quarterly conversion percentage of sales, and the converter interactions over the 10 quarters. Models 1, 2 and 3 are OLS models and Model 4 is a GLS model estimated with the same variables as presented in the baseline Model 1. The model is restricted to submarkets having more than three conversion projects during the 10-quarter time period. The submarket variable coefficients are not reported. ∗∗∗ Significant at 1%. ∗∗ Significant at 5%. ∗ Significant at 10%.

Coef.

Model 1

Variable

Table 5  continued

Clientele Effects and Condo Conversions 629

630 Benjamin et al.

a local operator. Conversion takes a local presence to upgrade the property and market the property. The nonlocal buyer has a disadvantage in permitting and entitlement, managing the conversion process and supervising the marketing staff in the sales office. Condo conversion is time intensive and often requires some recourse borrowing. National apartment investors are at a disadvantage to local investors who are better able to handle the property acquisition, condo transformation and marketing risks. The self-selection inverse Mills ratio is computed to measure the impact of private information on pricing. That private information is included in the estimation of the apartment pricing. The probability of a condo conversion purchase depends on the characteristics of the owner, determining the inverse Mills ratio. If the private information is significant in the pricing equation, then the characteristics of the owner by residence is a determinant of asset valuation. Table 4 presents the results from four models of unit price to implement Equation (8). The dependent variable is the logarithm of the per-unit sales price. Model 1 includes a variable for whether the property was purchased for conversion along with other control variables. A positive coefficient yields a condo conversion asset pricing premium. Model 2 tests for whether the conversion premium is stable over time by interacting conversion with time dummies. Model 3 provides an assessment of the impact of the quarterly percentage of conversions on the transaction price. Model 4 estimates the base equation from Model 1 using generalized least squares (GLS). Each model includes the inverse Mills ratio of Heckman (1978) to control for private information. Table 4 includes all 29 Miami-Fort Lauderdale MSA submarkets.6 In Table 4, for all models price is positively related to size, as measured in acres, building quality, number of floors and average unit size. Having another floor raises the price of the building by 2.7% to 3.3%. Another 100 square feet raises the price by 7.3% to 7.6%. Building quality, rated as excellent or good, such as a Class A property, yields a premium of 20.5% to 23.5%. Apartment buildings depreciate. The price for a unit falls by 0.2% to 0.3% for each year of age. Local buyers pay higher prices, given their dominance of conversion.7 The characteristics of the owner affect price determination in addition to those of the property. 6

The submarket variable coefficients are not reported. There are between-market differences in effect. The activity of interest in this study is focused on property conversions so the submarket variables are control variables. A series of tests for spatial autocorrelation was carried out, including delineating the submarkets. The results are robust to these specifications.

7

Local investors are more likely to be influenced by the premiums associated with conversion. These investors see the actual property trades in the local market and change return expectations, especially ones that night be associated with the conversion option.

Clientele Effects and Condo Conversions 631

The coefficients by quarter constitute the growth rates of a hedonic price index for apartments in the Miami-Fort Lauderdale MSA over 2004–2006. In the last two quarters of 2004, prices were flat. Apartment unit prices then increased dramatically during 2005. In calendar year 2005, the price of a quality-adjusted apartment building rose by approximately 39%. During 2006 prices increased further, but by a more modest 3.5%. In Model 1, converters pay an average premium of 7.2% over the 10-quarter sample window. This premium controls for the characteristics of the property and the neighborhood effects. As with the residence of the owner, where locals pay more than nonlocals, the characteristics of the buyer affect the price. Those intending to convert the property pay more, as appreciation rather than rental investors. Model 2 in Table 4 creates cross variables between conversion and time to test the capital gains expectations. The tracking of the condo conversion premium can be determined by quarter. Except for the first quarter of the sample the price sequence has an inverse U shape. In the second quarter of 2005, the coefficient is 0.160 and in the third quarter it is 0.138. Taking the antilog of these coefficients and subtracting one yields 0.173 and 0.148. Converters paid 17.3% more than rental investors in the second quarter and 14.8% more in the third quarter. These variable coefficients are statistically significant at the 5% and 10% levels, respectively. By late 2005, converters were no longer paying a premium. They appear to be paying discounts for much of 2006, but the results are not statistically significant. During two of the middle quarters of the sample window, there is a significant difference in prices, which corresponds with the inverted U-shaped pricing pattern. Model 3 tests for the price-volume correlation. The proportion of properties bought by converters is labeled as “% conversion per quarter.” There is a higher price when the volume of sales accounted for by converters rises. The coefficient on this variable is statistically significant at the 1% level. This impact is in addition to the actual premium associated with converters who pay approximately 9.3% more for a comparable apartment building. As the volume and price coefficients are both positive, when converters increase their market presence prices rise. Kramer and Donninger (1987) evaluate ordinary and GLS when spatial autocorrelation may be present. In the limiting case, the relative efficiency of OLS is one or similar to GLS when there is a constant in the regression. Although there is a constant, both GLS and OLS regressions are reported. GLS estimates are reported as Model 4 for all markets, which is comparable with Model 1. Other

632 Benjamin et al.

tests were carried out for submarkets and the results were similar, indicating that spatial autocorrelation does not affect the conclusions.8 Table 5 carries out the estimation for another subsample. Only submarkets with more than three conversion transactions are used. The results are robust in confirming the initial results from Table 4. The coefficients on the conversion variables are statistically significant at the 5% and 1% levels across the models.9 Converters pay between 6.8% and 9.3% more for an apartment unit, controlled for quality and time. Out-of-town buyers pay less than locals, who dominate the conversion market. The conversion premium remains concentrated in mid 2005. In the second quarter, appreciation investors as converters paid 18.1% more, and 14.5% more in the third quarter of 2005.10 The entry of condo converters to the Miami-Fort Lauderdale apartment market in 2004 causes a bubble in prices, which eventually dissipates. The dissipation occurs in less than a year. Investor clienteles, specifically condo converters, are trying to arbitrage the price paid for cash flow and a retail selling price per unit that is not dependent on rental income. The converter expects that for some short-term period the value is out of equilibrium. First movers, who may be converting property acquired years in advance, make excess returns based on inelastic supply. As more supply comes on line in 2005 with investors acquiring property for conversion and rising, expected

8

The analysis was also run on the five largest submarkets. From Table 1 Miami Beach is the largest apartment submarket accounting for 16% of the sample transactions, while Fort Lauderdale has 14%. The other three large submarkets are Miami, Hollywood and Northeast Dade. These five submarkets account for 52% of the transactions. The tests and specification are repeated to determine whether the larger submarkets are typical. Instead of using a full set of 29, the transaction data are segmented into the five largest submarkets. The results are consistent with those reported. To conduct an additional robustness check, the 29 submarkets are grouped into five large submarkets based on the proportion sold to converters. Again, the regression results do not change qualitatively.

9

In addition to the model results presented, analysis of only submarkets with at least 10% of sales being conversion sales is completed with results similar to those presented. The findings appear to be robust across model specification.

10 There is a significant effect in the third quarter of 2004. That quarter has the lowest proportion of sales to converters in any of the sample of 7% from Table 2. The result is consistent, in that converters pay more than rental investors, and in no quarter do they pay less. The second and third quarters of 2005 more accurately describe the peak of the conversion boom, as shown in Table 2. Those two quarters resulted in the largest sales volumes of any in the sample, and the highest numbers of sales to converters. The positive converter premium in 2004:3 remains supportive of the underlying model, but could be influenced by outliers, as there were only 12 sales to appreciation investors out of 172 transactions during the quarter.

Clientele Effects and Condo Conversions 633

capital gains, profits are reduced and then eliminated. Apartment prices rose by 39% in Miami-Fort Lauderdale in 2005 alone because of the competition from appreciation-oriented converters. That higher price was paid even by rental investors. Appreciation investors paid an additional premium. Concluding Remarks The clienteles that purchase apartment buildings can be sorted by their characteristics. Active buyers such as developers and converters are more likely to be local. These buyers have a comparative advantage in the entitlement process for permitting, tenant negotiations, renovation and marketing. Passive buyers that are income-oriented are more likely to be nonlocal and are not generally contributors to the inverse U shape in prices and volumes. The data and empirical analysis indicate that converters pay acquisition premiums relative to other investor clienteles or groups. Converters lead the overall markets to higher per unit sale prices. The results have broader implications for real property acquisition strategies. Clienteles are shown to impact price and returns. Nonlocal buyers for these apartments actually pay less, as they are less likely to be converters. Converters pay more regardless of their location. The price paid for a real estate property depends on the identity of the buyer and the movement of investor clienteles into and out of markets. Price growth depends on investor clienteles, impacting returns. Condo conversions are based on expectations of developers. These can be influenced by systematic risks such as homeownership tax benefits for purchasers, housing appreciation and negative or low real rates of interest for homeowners. Those shocks have real implications for competition on properties and transaction prices. Nonlocals are significantly less likely than locals to be the buyers of properties for conversion. Nonlocals pay less for properties than locals. There may be abnormally large returns in condo conversion requiring local knowledge of the housing market. Alternatively, nonlocal buyers have a national perspective and are less likely to be caught up in a short-term speculative frenzy in one market. On the sell side, local and nonlocal owners tend to receive the same prices. The clientele effect comes in short bursts when there is a different, but resource-intensive usage for the property that favors locals. We are grateful to Richard Langhorne, Crocker Liu, Maury Seldin and three referees for their comments and suggestions. Support from the Homer Hoyt Institute is acknowledged as is support from the Jerome Bain Real Estate Institute at Florida International University.

634 Benjamin et al.

References Benjamin, J.D., P. Chinloy and G.S. Sirmans. 2000. Housing Vouchers, Tenant Quality and Apartment Values. The Journal of Real Estate Finance and Economics 20: 37–48. Colwell, P.F. and H.J. Munneke. 2006. Bargaining Strength and Property Class in the Office Markets. The Journal of Real Estate Finance and Economics 33: 197–213. Diskin, B.A. and A. Tashchian. 1984. Application of Logit Analysis to the Determination of Tenant Absorption in Condominium Conversion. AREUEA Journal 12: 191–205. Hardin, W.G., III and M.L. Wolverton. 1999. Equity REIT Property Acquisitions: Do Apartment REITs Pay a Premium? Journal of Real Estate Research 17: 113–126. Hansmann, H. 1991. Condominium and Cooperative Housing: Transactional Efficiency, Tax Subsidies, and Tenure Choice. Journal of Legal Studies 20: 25–71. Harding, J.P., S.S. Rosenthal and C.F. Sirmans. 2003a. Estimating Bargaining Power in the Market for Existing Homes. Review of Economics and Statistics 85: 178–188. ——. 2003b. Estimating Bargaining Effects in Hedonic Models: Evidence From the Housing Market. Real Estate Economics 31: 601–622. Heckman, J.J. 1978. Sample Selection Bias as a Specification Error. Econometrica 47: 153–162. Kramer, W. and C. Donninger. 1987. Spatial Autocorrelation Among Errors and the Relative Efficiency of OLS in the Linear Regression Model. Journal of the American Statistical Association 82: 577–579. Lambson, V.E., G.R. McQueen and B.A. Slade. 2004. Do Out-of-State Buyers Pay More for Real Estate? An Examination of Anchoring-Induced Bias and Search Costs. Real Estate Economics 32: 85–126. Malpezzi, S. and J.D. Shilling. 2000. Institutional Investors Tilt Their Real Estate Holdings Toward Quality Too. The Journal of Real Estate Finance and Economics 21: 113–140. Rosen, K.T. and L.B. Smith. 1983. The Price-Adjustment Process for Rental Housing and the Natural Vacancy Rate. American Economic Review 73: 779–786. Steele, M. 1993. Conversions, Condominiums and Capital Gains: The Transformation of the Ontario Rental Housing Market. Urban Studies 30: 103–126. Whinihan, M.J. 1984. Condominium Conversion and the Tax Reform Acts of 1969 and 1976. AREUEA Journal 12: 461–472.