On the Effects of Consumer Sentiment on House Permits: Asymmetric Evidence From State-Level Data in the United States

Abstract

Previous studies analyzing the link between the housing market and consumer sentiment focus on the impact of the sentiment on house prices. This paper assesses the impact of consumer sentiment on house permits as a proxy for housing production using a linear model that assumes symmetric effects. Additionally, we consider a nonlinear asymmetric model. We estimate both models using state-level data from each state of the United States and find short-run effects of consumer sentiment on house permits in 31 and long-run effects in 33 states. The comparable numbers by estimating a nonlinear asymmetric model were 42 and 41, respectively. In sum, increased consumer confidence has positive long-run effects on the issuance of house permits in most states.

Keywords:

• consumer sentiment
• house permits
• state level data
• the U.S.
• asymmetric
• analysis

Author's Note

Valuable comments of an anonymous reviewer and those of the editor are greatly appreciated. Any remaining error, however, is our own responsibility. All views and information expressed in this paper do not represent the opinions of any entity whatsoever with which the author has been, is now, or will be affiliated.

Notes

1 There is another literature on consumer sentiment indices in which researchers basically assess the predictive powers of different sentiment indices. For example, Marcato and Nanda (2016) find that the sentiment in real estate conveys valuable information that can help predict changes in real estate returns. They also review the related literature. Consumer sentiment has also been used to predict the turning point in the housing market (Croce & Haurin, 2009).

2 Considering the production side of the housing market, Bahmani-Oskooee, Ghodsi, et al. (2021) recently argued that increased issuance of housing permits could move house prices in either direction. They tested the supply-side versus the demand-side hypotheses using data from each state in the United States. Both hypotheses were supported in the short run for most states but only the demand hypothesis in the long run in some states. They uncover asymmetric short-run and long-run effects in 21 states.

3 This section closely follows Bahmani-Oskooee et al. (Bahmani-Oskooee, Azaryan, et al., 2021; Bahmani-Oskooee et al., 2022). In the first paper, authors investigate the impact of consumer sentiment on house prices by using data from each state of the United States. When a linear model was estimated, they find short-run effects of consumer sentiment on house prices in 34 states that lasted into the long run in only 13 states. The comparable numbers by estimating a nonlinear asymmetric model were 47 and 22, respectively. In the second paper, authors investigate the asymmetric effects of a news-based policy uncertainty measure on house prices, again, in each state. They report short-run asymmetric effects in 32 states and long-run asymmetric effects in 24 states. These were higher than the number of states reported by Bahmani-Oskooee and Ghodsi (2017), who only estimated a linear model. Comparable numbers in the later study were 24 and 17, respectively.

4 Examples of other studies that included income and mortgage rates are as follows: Chen and Patel (1998), Apergis (2003), Apergis and Rezitis (2003), Abelson et al. (2005), Gallin (2006), Chen et al. (2007), McQuinn and O’Reilly (2008), Rapach and Strauss (2009), Kim and Bhattacharya (2009), Mikhed and Zemcik (2009), Adams and Fuss (2010), Zhou (2010), Holly et al. (2010), Madsen (2012), Apergis et al. (2015), Antonakakis et al. (2015), Bahmani-Oskooee and Ghodsi (2016), Bahmani-Oskooee and Ghodsi (2018), Baghestani and Viriyavipart (2019), Alexiou and Vogiazas (2019), Alola (2020), De Albuquerquemello and Bessarria (2021), and Cuesta and Kukk (2021).

5 Note that the t test in this context is the same as the t test applied to test significance of lagged error-correction term in Engle and Granger (1987) two-step procedure.

6 Note that partial sums of a variable X are constructed as $X_t^{+}={\sum }_{j=1}^t\max \hspace{0.17em}\left(\Delta {\mathit{LnX}}_j,0\right),$ and $S_t^{-}={\sum }_{j=1}^t\min \hspace{0.17em}\left(\Delta {\mathit{LnS}}_j,0\right).$

7 Shin et al. (2014, p. 291) argue that two partial sum variables associated with each variable should be treated as a single variable so that the critical values of the F test stay at the same high level in both linear and nonlinear models.

8 Note that panel C reported not only cointegration tests results but also a few additional diagnostics. To make sure residuals in each model are autocorrelation-free, we report the Lagrange multiplier test as LM and to make sure the models are correctly specified, we report Ramsey’s RESET test. Both are insignificant almost in all models supporting autocorrelation-free residuals and correctly specified models. We also report the outcome of the CUSUM and CUSUMSQ tests for stability of all estimates. These are reported as CS and CS² where stable estimates are indicated by “S” and unstable estimates by “US.” Clearly, most estimates are stable. Finally, the goodness-of-fit in each model is reflected by the size of adjusted R².

9 The list includes Alabama, California, Colorado, Florida, Hawaii, Iowa, Idaho, Indiana, Kentucky, Maryland, Minnesota, Missouri, North Carolina, North Dakota, Nebraska, New Hampshire, New Jersey, New Mexico, Nevada, New York, Ohio, Pennsylvania, South Carolina, Tennessee, Virginia, and West Virginia.

10 Other diagnostics are similar to those in Table 1 and need no repeat.

11 In the absence of any other study on the asymmetric effects of consumer sentiment on housing permits, we are unable to engage in direct comparison. However, compared to Bahmani-Oskooee, Azaryan, et al. (2021), who assessed the asymmetric effects of consumer sentiment on house prices in each state, we find long-run effects in many more states. More precisely, while we find consumer sentiment to have long-run impact on house permits in 41 states, they find long-run impact of consumer sentiment on house prices in 21 states.

12 Future studies could use data from other countries to add to this part of the housing literature.