Does Time-on-Market Measurement Matter?

Time-on-market is an important measure in residential real estate transaction and brokerage studies, such as Genesove and Mayer (1997), Glower et al. (1998), Knight (2002), Anglin et al. (2003), Rutherford et al.(2005), Turnbull et al. (2006), and Huang and Rutherford (2007).

Some recent research acknowledges a disconnect between MLS-reported “time-on-market” and the fact that many listings do not actually transact under their initial listing terms and period.

“Time-on-market” is used in this situation since it is the typical term used in a MLS.

In practice, agents often say that “time-on-market” is the average number of days a property is on the market. In addition to being somewhat nebulous, this definition is inherently inaccurate when all properties do not sell during their initial listing period. For example, in the present market, it is not uncommon for fewer than 50 % of listings to transact.

The problem is likely more acute in older data that was actually printed and then distributed. How the property listings were classified was very important in the pre-internet dissemination of information. For example, some multiple listing services had separate books for new listings and price changes. The noise in the older data is likely even greater since it is probable that researchers simply used a MLS generated time-on-market measure and did not even have actual listing, pending, and closed dates.

It is certainly possible for an existing listing to be extended or for a withdrawn listing to be placed back on the market. In practice, this is rarely the case. The MLS from which the relisting data are drawn requires that such requests be made before the expiration date of the original listing, and there are no penalties for failure to do so.

See also Munneke and Yavas (2001); Rutherford et al. (2005); Turnbull and Dombrow (2006); Turnbull et al. (2006); Huang and Rutherford (2007); and Rutherford et al.(2007). Levitt and Syverson (2008) correct for this flaw in their analysis, but they define a property as relisted if it comes back on the market within 180 days. Six full months seems a bit long to count a property as relisted, given how much market conditions can change in that period of time and the substantial changes that could be made to a residential property during such a time period. Through discussions with local practitioners, the common practice of excluding particular leads was identified. This refers to identifying by name particular individuals that had expressed an interest in the property during the expiring listing period. If that individual then enters negotiations during the exclusion period, the commission is earned by the listing broker who held the listing at the time of initial viewing. Locally, the generally accepted exclusion period seems to be 30 days, with the longest exclusion period set at 45 days. Given a natural break observed in the data at 48 days between prior listing expiration and relisting, the definition of a relisted property used in this study is 48 days.

In discussion with authors of several of these prior studies, it is apparent that many of the studies, especially the earliest ones when data were less extensive and less available, used time-on-market numbers without fully knowing how they were calculated because that number was the only one available.

Kiefer (1988) and Greene (1997) discuss the positive aspects of using non-linear distributional forms in modeling time-on-market. These researchers point out that the hazard techniques necessitated by the use of non-linear distributional forms, especially the Weibull or Exponential distributions, address issues with independent variables whose values vary over time and have non-normal error terms.

The attributes associated with the NOMKT variable have been shown to be important in the literature so we include discussion here and use the NOMKT variable in the analysis. The empirical results are robust to the use or non-use of this factor.

The sub-market data are available for review. A history of all listings for a specific property is required to create the relisting dataset. This data is normally not requested by, and therefore not provided to, academic researchers. The required data was available for the Miami-Dade MSA, so the focus of this section of the study turns to South Florida out of necessity.

A property is considered to be relisted if it is placed back on the market within 48 days of the end of the prior listing. As noted earlier, Levitt and Syverson (2008) use 180 days in a similar definition. Please refer to footnote 6 for a more complete explanation of the choice of 48 days. Results were qualitatively unchanged when obtained for periods of 30 days and 60 days; these are available from the authors upon request.

The DOP control differs from the control suggested by Anglin et al. (2003). Instead of hedonically suggested listing prices, the sales price is used to compute the overpricing control as 1–SP/ORIG_LP. This is the measure investigated by Allen and Dare (2006). It is also a control variable used by Herrin et al. (2004) and Turnbull and Dombrow (2006) in the same manner as in Eq. (2).

The goal of this research is not to critique early researchers, but to highlight limitations in the data to which they had access and which may have limited empirical results. Data limitations are a fact of life for all real estate researchers, and in the present case, such limitations are merely referenced as a potential weakness in prior research that limited researchers’ ability to tie empirical results to theoretical constructs. Few of the early studies mention definitions and TOM is ambiguous as applied.

The interested reader is referred to Johnson et al. (2007) for additional information on the probability of sale transaction metric.