1 / 0

Commercial Real Estate Property Sales in Hot and Cold Markets: Evidence from 35 Years of NCREIF Sales Data (1978 – 201

Commercial Real Estate Property Sales in Hot and Cold Markets: Evidence from 35 Years of NCREIF Sales Data (1978 – 2012). Rebel A. Cole DePaul University AREUEA National Meetings Washington, DC May 30, 2013. Introduction.

heba
Télécharger la présentation

Commercial Real Estate Property Sales in Hot and Cold Markets: Evidence from 35 Years of NCREIF Sales Data (1978 – 201

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Commercial Real Estate Property Sales in Hot and Cold Markets: Evidence from 35 Years of NCREIF Sales Data(1978 – 2012)

    Rebel A. Cole DePaul University AREUEA National Meetings Washington, DC May 30, 2013
  2. Introduction What kinds of commercial real estate (“CRE”) properties do investors sell and when do they sell them? Are there differences in the kinds of properties they sell during hot and cold markets? Are there systematic biases in properties sold by investors?
  3. Introduction These questions also have important implications for: Investors - as transaction frequency is a measure of market liquidity and an indicator of market conditions, with higher frequencies being observed in hot markets and lower frequencies in cold markets. Researchers analyzing property sales, be it to measure the accuracy of appraisals , or to construct transaction-based return indices.
  4. Introduction If sold properties are not a random sample of all properties (and they are not), then it is important to account for this non-randomness when using sold properties for such tasks as: assessing appraisal accuracy, or estimating transaction values for non-sold properties.
  5. Introduction In this study, I provide new evidence on this issue by examining what factors explain the probability that a property was sold out of the NCREIF National Property Index (NPI) during the past 35 years, which spans four up cycles and three down cycles of the CRE market. (1978 – 2012) By examining sales over the past 35 years covered by NCREIF, I provide evidence regarding whether determinants of property sales differ in hot and cold markets. I also shed new light on whether these determinants have changed during the past decade, when transaction frequency skyrocketed. Previous research has only looked at NPI sales prior to 2002.
  6. Literature Review: Cole, Rebel. 1988. A New Look at Commercial Real Estate Returns. PhD Dissertation, University of North Carolina at Chapel Hill. Guilkey, David, Mike Miles and Rebel Cole. 1989. The Motivations for Institutional Real Estate Sales and Implications for Generalizing from Specific Property Sales to Asset Class Returns. AREUEA Journal 17, 70-86. Miles, Mike, Rebel Cole and David Guilkey. 1990. A Different Look at Commercial Real Estate Returns. AREUEA Journal 18, 403 – 430. Webb, Brian, Mike Miles, and David Guilkey. 1992. Transactions-Driven Commercial Real Estate Returns: The Panacea to Asset Allocation Models? AREUEA Journal 20, 325 – 357. Fisher, Jeff, Dean Gatzlaff, David Geltner and Donald Haurin. 2004. An Analysis of the Determinants of Transaction Frequency of Institutional Commercial Real Estate Investment Property. Real Estate Economics 32, 239 – 264.
  7. Literature Review:Guilkey, Cole and Miles (1989)
  8. Literature Review: Miles, Cole, and Guilkey (1990)
  9. The data used in these early studies was hand-collected by the authors from hard-copy appraisals in the offices of NCREIF members around the country. After publication of these early studies, NCREIF formalized the data collection process to include many of these additional property characteristics (such as square footage and percentage leased), beyond just income, appraised value, partial sales, and capital improvements.
  10. Literature Review:Fisher, Gatzlaff, Geltner and Haurin (2004)
  11. Literature Review: Fisher, Gatzlaff, Geltner and Haurin (2004)
  12. Literature Review: Fisher, Gatzlaff, Geltner and Haurin (2004)
  13. Data:NCREIF NPI Research Database 105,960 property-year observations over the 1978 – 2012 period. 9,053 properties sold during 1978 – 2012. 96,907 unsold property-year observations over the 1978 – 2012 period.
  14. Data:Explanatory Variables I generally follow Cole (1988), GCM (1989), MCG(1990), and Fisher et al. (2004) in choosing explanatory variables: Property Characteristics Owner Characteristics National / Local Market Conditions
  15. Data:Explanatory Variables Property Characteristics Square Footage Building Age Percentage Leased Income Return
  16. Data:Explanatory Variables Owner Characteristics Fund Type Open, ODCE (Open Diversified Core Equity), Closed, Separate Account Joint Venture (yes/no) Uses Leverage (yes/no) Holding Period (years)
  17. Data:Explanatory Variables National Market Conditions NPI Capital Appreciation Return S&P 500 Return 10-Year T-Bond Yield % Change in Unemployment Rate 1986-1987 Dummy for Tax Reform
  18. Methodology:Descriptive Statistics and Univariate Tests I calculate means and standard errors for: Sold and Unsold properties In Hot and Cold Markets Cold Markets are defined as years where: NPI Capital Appreciation Return < 1.00% 1990 - 1996 2001 – 2002 2009 – 2010 Hot Markets: All other years during 1978 - 2012
  19. Methodology:Imputation of Missing Values For several key property-level variables, values are missing. Out of 21,003 properties: Square Footage: 1,579 Year Built: 5,045 % Leased: 3,192 At least one missing: 7,257 I can either: Delete more than 1/3rd of or sample, or Impute the missing values.
  20. Methodology:Imputation of Missing Values I choose to impute, using a regression model populated by all of my potential explanatory variables. Imputation is standard practice for survey data. My imputation methodology was developed by Arthur Kennickell (1991) for the Fed’s Surveys of Consumer Finance, and has been used for most Fed surveys conducted during the past 20 years. For each variable, it estimates a variance-covariance matrix based upon available pair-wise observations, then calculates a unique regression model for predicting each missing observation.
  21. Data:NPI is an Unbalanced Panel Consider three properties: One bought in 2003, sold in 2007 One bought in 2004, not sold through 2012 One bought in 2005, sold in 2011 From this, it is obvious that time-series observations for a particular property are notindependent, and that censoring is an issue.
  22. Methodology:Multivariate Tests Discrete-time hazard model (Shumway 2001) Algebraically, equivalent to a multiple-period (panel) logistic regression. However, logit(and probit) assumeproperty-year observations for a particular property are independent; andassume no censoring. But a property can only be sold once by a manager, So each property constitutes only a singleobservation. Should calculate std. errors using # properties, not # property-years.
  23. From Table 1:Properties Sold from the NPI by Year Number of properties rises from 264 in 1978 to 7,805 in 2012. Number sold rises from 2 in 1979 to 774 in 2005. Percentage sold ranges from 1% in 1979 to 15% in 2005; we clearly see the three down markets.
  24. From Table1:Percentage Sold Tracks NPI Returns Correlation for 1978-2012 is only 0.23. However, correlation for 1985-2012 is 0.75. During early years, managers were buying, not selling.
  25. From Table 2:Univariate Tests for Full Sample Sold properties are significantly more likely to be Apartment, Office, older, longer held, smaller, and more profitable.; less likely to be Retail or Industrial.
  26. From Table 2:Univariate Tests for Full Sample Sold properties are significantly less likely to be owned by an Op-end Fund, or Joint Venture, or to be levered; Sold properties are more likely to be owned by a closed-end fund.
  27. From Table 2:Univariate Tests for Full Sample Properties are more likely to be sold when NPI capital appreciation is higher , when stock returns are higher, when unemployment is falling, when Inflation is lower and when T-Bond yields are falling less fast.
  28. From Table 5:Univariate Tests for Full Sample - Hot vs. Cold Hotel properties are significantly less likely to be sold in hot markets. Retailproperties are significantly less likely to be sold in cold markets. Older properties are significantly more likely to be sold in hot markets.
  29. From Table 5:Univariate Tests for Full Sample – Hot vs. Cold Open-end funds are significantly less likely to sell in hot markets. Closed-end funds are significantly more likely to sell in hot markets. Separate Accounts are significantly more likely to sell in cold markets. Levered properties are significantly less likely to see in cold markets.
  30. From Table 5:Univariate Tests for Full Sample – Hot vs. Cold Sold properties sell when NPI Appreciation is higher, but this effect is greater in cold markets. Properties sell when Stock Returns are higher in hot markets but not in cold markets. Properties sell when Inflation is rising in hot markets but when inflation is falling in cold markets.
  31. From Table 10:Multivariate Tests for Hot vs. Cold Markets Older properties are significantly more likely to sell in hot, but not in cold markets. Other property characteristics are not significant in either type of market.
  32. From Table 10:Multivariate Tests for Hot vs. Cold Markets Relative to properties owned by Closed-End funds, properties owed by Separate Accounts are less likely to sell hot and cold markets, but the effect is significantly stronger in cold markets. JVs are less likely to sell in hot markets, but more likely to sell in cold markets.
  33. From Table 10:Multivariate Tests for Hot vs. Cold Markets The influence of the NPI Appreciation Return is significantly stronger in cold markets. 1% change increases odds of sale by 8.3% in cold market but by only 3.0% in hot market. The change in Unemployment Rate reduces odds of sale in hot market but increases odds of sale in cold market. Stock Return is not significant in either market. T-Bond Yield is significant in both markets.
  34. Summary and Conclusions In this study, I extend the literature examining sold vs. unsold commercial properties. I analyze data that covers sales from 1978 – 2012; previous research covered only the 1985 – 2011 period. I analyze a much fuller dataset for 1985 – 2011 by imputing missing values for key property-specific variables and including new NCREIF properties. I analyze hot and cold markets separately to test whether determinant of sales differ in the two types of markets. I utilize a hazard model that correctly treats the non-independence of time-series observations on an individual property and correctly treats censoring.
  35. Summary and Conclusions My findings largely confirm the results reported by Fisher et al. (2004) for the 1985-2001 period. Key determinants of property sales are: building age, fund type, holding period. Key macro factors are: National property performance, returns on alternative investments (T-bonds, but not Stocks), unemployment and inflation.
  36. Summary and Conclusions I also find significant differences in determinants of sales in hot and cold markets. Older properties are more likely to sell, but only in hot markets Joint ventures are more likely to sell in cold than in hot markets Separate accounts are much less likely to sell in cold markets National property returns are more important in cold markets. Unemployment changes sign in hot vs. cold markets. Rising UE reduces odds of sale in hot market, increases odds of sale in cold market.
  37. Thank you!
More Related