Initial Public Offerings and Local Housing Markets

Abstract

This study explores the link between corporations’ initial public offering (IPO) activities and the housing markets in the MSAs where the IPO firms are headquartered. Using a sample of 4,500 U.S. IPOs occurring from 1990 to 2018, a period covering the tech bubble years and the 2007–2008 financial crisis, we examine the existence and magnitude of the link between IPOs and housing, and related hypotheses including the short-term expectations effects, the wealth effects after the IPO lock-up expiration dates, and the long-term effects. Our results provide strong support to these hypotheses. The effects are stronger during the tech-bubble years. In addition, the results are to some extent driven by IPOs in traffic-congested and/or low-affordability cities, and the effects are generally amplified by local housing supply rigidity and IPO stock performance. We find that the overall impact of IPOs on the housing market is economically significant with the most prominent influence linked to the changes in short-term expectations.

Keywords:

• Initial public offering
• housing price
• expectations effects
• wealth effects

JEL classification:

• R30
• G30
• E30

Acknowledgments

We acknowledge helpful comments from Alexander Butler, Lu Fang, Ling Xiao Li, Pin-Te Lin, Zhenguo (Len) Lin, Michael LaCour-Little, Yingchun Liu, Wenlan Qian, Jarjisu Sa-Aadu, James Shilling, Sheridan Titman, Erdem Ucar, Ko Wang, Abdullah Yavas, Bektemir Ysmailov, Bing Zhu, two anonymous journal referees, and participants of the 2018 GCREC Conference in Qingdao, China, the 2019 Midwest Finance Association Annual Conference, the 2019 Asian Real Estate Society Annual Conference, the 2019 Eastern Finance Association Annual Conference, the 2019 Financial Management Association Annual Meeting, and the 2020 seminar at University of Northern Iowa.

Notes

1 The studies in the similar vein also include Borisov et al. (2017) and Babina et al. (2017), which provide varied findings and justifications for the relations between IPOs and local economic factors.

2 We also control for the population growth in untabulated results in Section “Alternate Variables” with similar inferences.

3 We thank Martin Kenney and Donald Patton for kindly sharing the data that serves as the main sample for our analysis. We also thank Jay Ritter for making his IPO data available for public use.

4 Emerging Growth IPOs Database by Kenney and Patton is also used in studies by Feuerriegel et al. (2014), Henry and Gregoriou (2013), and Vakrman and Kristoufek (2015).

5 See article titled “Advanced Industries Still Rule the U.S. Economy – But It’s an Advantage That’s Slipping” by Richard Florida, at Citylab on February 3, 2015, available from https://www.citylab.com/life/2015/02/advanced-industries-still-rule-the-us-economybut-its-an-advantage-thats-slipping/385084/.

6 https://site.warrington.ufl.edu/ritter/ipo-data/.

7 We also use the FHFA Purchase-Only Housing Price Indices (for 100 major MSAs) for robustness tests, which generate similar results as the FHFA All-Transactions Housing Price Indices do, as discussed in section “Alternate Variables.”

8 See Appendix A for the variable description.

9 IPO lockup refers to a pre-specified post-IPO period, typically 90 days to 180 days, that restricts the ability of the firm’s management, employees and extant shareholders to sell their pre-IPO shareholdings on the secondary market. Brav and Gompers (2003) document substantial absolute abnormal returns and trading volume at the expiration of the lockup period. This increase in trading volume suggests that some of the pre-IPO shareholders are converting their holdings into cash. In addition, Butler et al. (2019) report a positive association between IPO activity and mortgage origination.

10 Several studies found economic conditions to be associated with IPO activity and “hot issue” markets (Çolak & Günay, 2011; Ivanov & Lewis, 2008; Pástor & Veronesi, 2005). Furthermore, real estate prices are correlated with economic conditions as well (Titman et al., 2014).

11 One example of MSA economic activity change is that non-IPO firms might also relocate their headquarters to a particular MSA over the same time period when IPOs are occurring in this MSA, increasing employment and thus the demands for local housing.

12 Case and Shiller (1989), Campbell et al. (2009), and Titman et al. (2014) find that 1-year or 6-month lagged terms positively affect house price change rate, while reversal occurs usually between the 6th month and the 3rd year.

13 Some lockup periods are fewer than 180 days with the lower boundary usually at 90 days driving the selection of both 1-quarter and 2-quarter lags to test Hypothesis 2.

14 Specifically, we use a matching period of 0-day and 7-day for the returns corresponding to the current quarter IPO activity variable, 3-month for returns corresponding to the 1-quarter lagged IPO activity variable, and 6-month for returns corresponding to the 2-quarter lagged IPO activity variable.

15 Theoretically, a good IPO stock performance can be a double-edged sword to the IPO shareholders’ housing demand: it might increase house purchases if the shareholders sell the stocks to get cash, or reduce house purchases if shareholders are induced to delay selling stocks hoping for further stock price increase. The IPO literature leans towards the former hypothesis given the spike in trading volume and the drop in the stock price upon the lock-up expiration, suggesting that overall the wealth effect of IPO activities on the local housing markets should be strengthened by the IPO stock performance as we find in Section 4.2.

16 Similar approach is used in studies such as Butler et al. (2019).

17 We report the results without winsorizing. In untabulated robustness tests, we winsorize dependent, independent and control variables at 1 and 99 percentile and rerun our regressions. The results are essentially the same.

18 The table with the test statistics and their p-values is not included for the sake of brevity and is available upon request.

19 In the appendix, we describe how our variables are calculated.

20 The coefficient of iponuma is 1.5624, after dividing by 1,000. The median of MSA population in our sample is 0.226 million. Thus, the issuance of an IPO is associated with (1.5624 x 1,000) x 1/(0.226 x 10⁶)= 0.691% annual change in the local HPI level measured in the same quarter.

21 From Table 3, we no longer report the coefficients of control variables for the purpose of brevity, and the full tables are available upon request.

22 In the past, Texas A&M Transportation Institute published panel-data for the Commuter Stress Index. Thus, we were able to download the data for the period of 1982–2014. Now, however, the Institute only publishes the latest available annual data for the index (currently year 2017).

23 The estimates for the control variables are not reported in the published version of the paper for the sake of brevity and are available upon request.

24 https://en.wikipedia.org/wiki/List_of_regions_of_the_United_States. The U.S. Census Bureau defines four statistical regions. Each region consists of two or three divisions resulting in total of 9 divisions.

25 The small number of MSAs in most of states prevent us from dividing MSAs within each state into bins. We obtain similar results when we divide MSAs within each census bureau-designated regions into 9 or 25 bins. The results are also similar when we match by level of population and GMP instead of their growth rates.