Florida Metropolitan Housing Markets: Examining Club Convergence and Geographical Market Segmentation

Abstract

This study explores the convergence of housing prices for 21 metropolitan areas within the state of Florida for the quarterly period 1987:2 to 2017:3. The examination of house price differentials between metropolitan and state-level house prices using a battery of univariate and panel unit root testing approaches yielded mixed results with respect to the presence of convergence. However, the Phillips-Sul (2007; 2009) club convergence approach identifies four distinct convergence clubs for metropolitan area house prices within Florida with a relatively clear geographical segmentation of the housing market.

Keywords:

• Florida
• housing prices
• unit roots
• club convergence

Notes

1 See Holtz-Eakin, D., & Winkler, A. (2012, November 13) Boom, Bust, and Beyond: A Look at Housing Market Data in Florida. American Action Forum. https://www.americanactionforum.org/research/boom-bust-and-beyond-a-look-at-housing-market-data-in-florida/.

2 As noted by Holmes et al. (2011), there is heterogeneity in house prices within states given the differences in housing markets, hence the use of census and state level data only will mask such heterogeneity.

3 According to Evans and Karras (1996), the convergence process is either absolute or conditional based on individual effects. If the individual effects are zero, then the convergence process is considered absolute since the metropolitan house price index converges in levels to the state-level house price index.

4 The early literature on the U.K. emphasized the role of Greater London and South East regions in driving house price dynamics. See Rosenthal (1986), Giussani and Hadjimatheou (1991), MacDonald and Taylor (1993), Alexander and Barrow (1994), Drake (1995), Ashworth and Parker (1997), Meen (1999), Cook (2003; 2005a;b), Holmes (2007), Holmes and Grimes (2008), and Abbott and Devita (2012).

5 Studies by Stevenson (2004) focused on Ireland; Luo et al. (2007) on Australia’s capital cities; Burger and Van Rensburg (2008) on metropolitan areas in South Africa; Larraz-Iribas and Alfaro-Navarro (2008) on regional housing prices in Spain; Chien (2010) on Taiwanese cities; Vandeenkiste and Hiebert (2011) for Euro-zone countries; Gupta et al. (2015) on eight European countries; and Holmes et al. (2017) on district-level house prices in Paris.

6 The simulation procedure first estimates the parameters under the null hypothesis in generating a panel data set using these parameters. The residuals are drawn from a Gaussian distribution with a variance equal to the estimated historical variance for each metropolitan area. Next, we apply the convergence procedure and compute the statistics t_ρ and F_δ. The empirical distributions of these two statistics are generated after 10,000 simulations. Critical threshold values correspond to the following quantiles: 1%, 5% and 10%.

7 Recall that the Lee and Strazicich (2003) model is based on (2).

8 The parameter α dictates the rate to which the cross-section variation over the transitions decays to zero over time.

9 Some clubs may be weakly divergent, i.e. $-1.65<{\mathrm{t}}_{\widehat{\mathrm{\gamma }}}<0.$

10 According to Phillips and Sul (2007), the choice of r = 0.3 is a satisfactory selection in terms of both size and power.

11 A challenge in this regard is matching several of the metropolitan areas defined in this study with the defined metropolitan areas from the U.S. Census Bureau and the U.S. Bureau of Economic Analysis. For example, our study defines the metropolitan area Miami-Miami Beach-Kendall while sources for such data as population and per capita income define Miami-Ft. Lauderdale-West Palm Beach, which cuts across two other metropolitan areas defined in our study: West Palm Beach-Boca Raton-Delray Beach and Ft. Lauderdale-Pompano Beach-Deerfield.