Cross-Sectional Differences in Hotel Revenue Performance During the Covid-19 Pandemic

Abstract

This study investigates price elasticity of demand and its pricing effect on revenue performance of hotels during the COVID-19 pandemic. Using annual operating performance data on over 2,500 hotels from 2018 to 2021, this study provides empirical evidence of a relatively inelastic demand for the lodging sector. Price, income, cross-price, and lagged demand are positive and inelastic in their relationships with demand. Price has a significant and negative effect on lodging demand while a hotel’s competitive position has a significant influence on RevPAR performance with both negatively affected by COVID-19. More important, the results show a significantly greater positive impact of pricing on RevPAR penetration for hotels that raised rates relative to hotels that dropped rates. The findings suggest that it may be best for hotels to maintain or raise room rates to maximize revenue performance in the midst of an external shock.

Introduction

It is a well-known fact that external shocks such as the events of 9/11, the economic recession in 2008 and 2009, and the current COVID-19 pandemic, had significant negative effects on the performance of U.S. lodging industry (Musto, 2020). Although the factors that induced each crisis are different, the impacts on hotel operating performance have varied in severity with the COVID-19 pandemic dwarfing all previous external shocks. Operating performance data from STR shows industry-wide occupancy dropping from 66% in 2019 to 44% in 2020, a decline of 33%, whereas ADR declined 21% from US$131 to US$103 over the same period. The resulting freefall in RevPAR in 2020 was a staggering 48%, the steepest decline ever in STR’s 35-year history of tracking lodging performance (STR, 2021).

Assessing hotels’ operating performance, as measured by RevPAR, is of paramount importance to all stakeholders and participants for long-term planning in the lodging industry (Noone et al., 2013). Hotel owners seek profitability in part via revenue growth; hotel managers formulate competitive pricing strategies to achieve higher revenue performance relative to their competitive sets; and managers and franchisors earn fees tied to revenue performance. Clearly, these stakeholders’ interests are affected if revenue performance suffers. It is, therefore, important to analyze the impact of external shocks on revenue performance to inform the relevant stakeholders and enable future strategic decision-making.

Our main purpose in conducting this study is to investigate the price elasticity of demand for the lodging industry and to link its effect on annual revenue performance during the height of the COVID-19 pandemic in 2020. We conduct the study in two stages. In the first stage, we empirically measure the elasticity of demand for the lodging sector. Then we apply the elasticity estimate to assess the impact of the COVID-19 pandemic on lodging revenue performance in 2020. This is a distinguishing feature of our study from prior research. The second stage of the analysis compares hotel operating performance relative to the competitive set of hotels operating within the same submarket location and segment. This comparison permits us to assess changes in occupancy and pricing on a hotel’s competitive market RevPAR performance vis a vis hotel pricing strategies and their influence on a hotel’s competitive occupancy position and overall submarket RevPAR performance. Our analysis includes both absolute levels and percentage changes in lodging fundamentals at the property-level. Additional robustness tests ensure our results are not driven by alternative explanations.

Research assessing the impact of pricing strategies on revenue performance during external shocks is limited. Although prior research has provided some evidence regarding lodging demand and operating performance during 9/11 and the last economic recession (2008-2009), the extant research has produced mixed findings. Some studies find the relationship between price and demand to be significantly inelastic (Canina & Carvell, 2005; Hiemstra & Ismail, 1990; Lee & Jang, 2012; Wheaton & Rossoff, 1998), whereas others find insignificance (Damonte et al., 1998; Tsai et al., 2006). Even the relationship between income and demand is mixed with significance in some studies (Canina & Carvell, 2005; Wheaton & Rossoff, 1998) and lack of significance in others (Qu et al., 2002; Tsai et al., 2006). Furthermore, there is debate on which proxy is better at capturing income elasticity. Different measures of income have been used from disposable income (Corgel et al., 2012; Tsai et al., 2006) to GDP (Canina & Carvell, 2005; Wheaton & Rossoff, 1998). Critically, the assumption of constant elasticity over time that is held in prior studies may not be valid. For example, Damonte et al., (1998) document variation in aggregate price elasticities across time and Song et al. (2011) argue that this assumption may not hold in periods with high uncertainty such as an economic crisis.

Determining the impact of discounting on revenue performance remains an unresolved empirical issue. Whether discounting is an appropriate pricing strategy in the midst of an external shock is an empirical question. The evidence from the limited number of existing studies to-date has produced conflicting or mixed findings (Enz et al., 2004, 2009; O’Neill & Mattila, 2006; O’Neill & Yeon, 2022). For example, using monthly data, O’Neill and Yeon (2022) find that luxury and upper upscale hotels that continuously discounted rates during the 2008–2009 recession generated higher RevPAR relative to similar hotels that avoided discounting rates. In contrast, other studies (Enz et al., 2004, 2009) using annual performance data suggest that hotels should not discount but rather maintain or raise rates. These contrasting findings from a limited number of lodging performance research studies suggest that additional research is needed to facilitate and inform industry practitioners’ understanding of the influence of pricing on revenue performance. Knowing how these critical performance variables are related is essential to strategically formulating appropriate pricing during an external shock. The results should aid lenders, brokers, appraisers, investors, owners, and operators in future crises as they illuminate the impact of COVID-19 on revenue performance in ways past crisis-related research has not.

We focus on the lodging sector for a fundamental reason. The unique feature of the lodging sector, differentiating it from other property types, is the daily leasing of hotel rooms, making the hotel sector prone to external shocks such as a recession or exposure to event risk (deRoos et al., 2014; Wheeler et al., 2016). Consequently, the events of 9/11, economic recession, and the COVID-19 pandemic reverberate as “top-down” external shocks that contribute to a much higher level of revenue volatility in hotels than other major property types (Wheeler et al., 2016). Small negative changes in occupancy and ADR magnify the decline in RevPAR performance. The hotel sector is usually the first to exhibit signs of weakness when it is hit by these external shocks or when economic conditions deteriorate (Wheeler et al., 2016).

The current study differs from prior research and makes a significant contribution to the lodging academic literature via several distinguishing aspects. First, unlike prior research investigating the shocks from 9/11 (Enz et al., 2004, 2009; Noone et al., 2013) and the economic recession of 2008–2009 (O’Neill & Yeon, 2022), the focus here is on the impact of COVID-19 on annual hotel revenue performance from 2018 to 2021. Second, lodging demand studies tend to focus on macro factors at the aggregate level (Corgel et al., 2012; Fujii et al., 1985; Lee & Jang, 2012; Qu et al., 2002; Tsai et al., 2006; Wheaton & Rossoff, 1998) instead of a disaggregated level (Canina & Carvell, 2005; Hiemstra & Ismail, 1990). Accuracy of data and failure to uncover underlying behavioral relationships are issues that arise with aggregated data (Corgel et al., 2012). We focus our analysis at the property-level, and thus, extend prior research on lodging demand and external shocks to include the impact of COVID-19.

Third, with the exception of Hiemstra and Ismail (1990) who applied elasticity to the impact of room taxes, no research linking elasticity estimates with revenue performance at the property-level exists. We therefore extend the existing lodging demand literature by linking price elasticity of demand to revenue performance in our sample hotels to capture the impact of the COVID-19 pandemic. Fourth, we employ two different estimation techniques (two-stage least squares [2SLS] and feasible generalized least squares (FGLS)) within the same study. These estimation techniques are robust to various specification issues such as heteroskedasticity, serial correlation, and simultaneity. No previous study has employed both methodologies in the same study using a single data set. Even though considerable variation exists in variables and methodologies for estimating demand, O’Neill and Ouyang (2020) suggest a three-predictor economic model as appropriate for capturing lodging demand and note that models with more than three predictors suffer from multicollinearity issues. We measure price elasticity using annual data over a relatively short time period from 2018 to 2021 since elasticity may not be constant during a period of high uncertainty (Damonte et al., 1998; Song et al., 2011). It is important to note that sample periods in prior studies (Enz et al., 2004, 2009; Noone et al., 2013) cut across periods of economic upturn and downturn without isolating the specific impact of the external shock. In addition, the time series in prior work vary from monthly to annual data spanning a single year to multiple years. Such variation indicates a lack of consensus on the appropriate times series for modeling demand and revenue performance.

Finally, this study examines revenue performance using both absolute levels and percentage changes in RevPAR performance. Prior research assessing the impact of external shocks on revenue performance used only absolute RevPAR level (Noone et al., 2013; O’Neill & Yeon, 2022) or the average RevPAR position without assessing the change in that position (Enz et al., 2004, 2009). Although a levels analysis can provide some useful information, it lacks the insights that can be drawn from examining percentage changes. For example, the percent change in the submarket RevPAR index used here extends prior approaches by providing practitioners more useful information; it indicates whether a hotel’s pricing strategies led to improved or worsened revenue performance against the competition from a previous, non-crisis year.

Literature Review

Several lodging academic studies to-date have investigated lodging demand and operating performance of hotels with some mixed results. Nevertheless, these studies have provided valuable insights on lodging demand and revenue performance during prior economic shocks. Below, we review the theoretical background and discuss existing lodging demand, revenue performance, and the discounting literature that coincide with the external shocks of 9/11 and the economic recession of 2008 to 2009. Although none of the existing lodging demand and performance research focus on COVID-19, the predominant findings emerging from this literature suggest that lodging demand is inelastic, and external shocks have devastating negative effects on hotel revenue performance.

Theoretical Background on Demand Theory

Microeconomic theory provides the foundation for modeling lodging demand and linking it with revenue performance. The theory of the law of demand posits an inverse relationship between price and demand demanded (Samuelson & Nordhaus, 2009). Thus, an increase (decrease) in price is associated with a decrease (increase) in quantity demanded, all else equal. Aside from price, demand changes may be driven by prices of substitutes, consumer tastes and preferences, income, and future expectations (Canina & Carvell, 2005). External shocks such as the events of 9/11 and the global recession could also influence changes in demand (Song et al., 2011).

A key component of the law of demand is the concept of elasticity. Elasticity refers to the sensitivity of quantity demanded in response to price changes. Price elasticity of demand is measured as the percentage in quantity demanded divided by the percentage change in price. A small change in price followed by a large change in quantity demanded is characterized as elastic demand (elasticity coefficient greater than 1). Conversely, demand is inelastic (elasticity coefficient less than 1) when a large change in price is associated with a small change in quantity demanded (Corgel et al., 2012).

Income elasticity and cross-price elasticity are two additional concepts that emerge from the theory of the law of demand (Samuelson & Nordhaus, 2009). Income elasticity of demand is measured as the percentage in quantity demand divided by the percentage in income. When the price of a product decreases (increases), the quantity demanded will increase (decrease), driven by an increase (decrease) in purchasing power of the consumer. This implies a positive relationship between income and demand. Cross-price elasticity of demand is the percentage change in quantity demanded divided by the percentage in the price of a substitute. The substitution occurs in relation to a change in the price of a related product. For example, when the price of a related product increases, consumers will seek substitutes for it so its quantity demanded will decrease as consumers switch to alternatives. Thus, a positive relationship is implied between the quantity demanded and the price of the related product.

Empirical Research on Demand in Lodging and Tourism

Early studies on lodging demand by Fujii et al. (1985) and Hiemstra and Ismail (1990) focused on the impact of increasing room taxes on lodging demand. Fujii et al. (1985) used a time-series model to measure price elasticity of demand at –.953. Hiemstra and Ismail (1990) used survey data from 1989 to estimate the price elasticity of demand for lodging and then applied the estimate to determine the average amounts of room taxes paid. The results of the study showed an average price elasticity of –.44. Moreover, price elasticity of demand varied across hotel segments ranging from –.35 for low prices to –.57 for high-priced hotels.

In another study, Damonte et al. (1998) analyzed lodging demand in Charleston and Columbia County in South Carolina from 1992 to 1995. Their ordinary least squares (OLS) model with two predictors, a quarterly dummy variable and county average daily rate (ADR), provided mixed results. Price elasticity varied across the two counties, insignificant and inelastic in Charleston County and more elastic (between 0.8 and 1.8) in Columbia County. The overall results indicated negative and nonsignificant price elasticity.

At the national level, Wheaton and Rossoff (1998) used aggregated data to examine the demand and supply side of the lodging cycle and assess its relationship with the U.S. economy. The results suggest that lodging demand is closely correlated with the economy. To avoid simultaneity, the authors employed a single equation OLS model in which room demand was specified as a function of gross domestic product (GDP), lagged ADR, and lagged room demand. The results revealed that GDP was significantly positive and elastic at 1.8, whereas ADR was significantly inelastic at –0.48; the conclusion was that the lodging cycle remains an unexplained empirical issue. These results are consistent with earlier estimates of 1.3 for GDP and –0.4 for ADR (Coopers & Lybrand, 1995).

Qu et al. (2002) identified the determinants of hotel room demand and supply in the Hong Kong hotel industry from 1980 to 1998. Using 2SLS to avoid multicollinearity, simultaneity, and other specification issues, the results of their study revealed that ADR and tourist arrivals were significant determinants of demand. In addition, external shocks such as the 1990–91 recession and the 1997–98 Asian financial crisis had significant negative impacts on room demand. Their price elasticity coefficient of –0.15 is lower than earlier estimates. However, income was found to be an insignificant predictor of demand for lodging. In another study, Tsai et al. (2006) employed 2SLS to examine the relationship between hotel room demand and supply in Las Vegas. Their results showed an insignificant price elasticity coefficient of –.395. Moreover, lagged price and income among other predictors were also nonsignificant predictors. Lee and Jang (2012) also applied 2SLS in a log-difference equation to determine the effect of relative changes in supply and demand. Their results showed a significant and inelastic price coefficient of –0.34, consistent with theoretical expectations.

Canina and Carvell (2005) analyzed lodging demand for 481 hotels in 22 U.S. metro markets from 1989 to 2000. To avoid simultaneity, the authors employed lagged ADR for price and a market average ADR for substitutes and estimated the equation via FGLS. The study results showed various measures of income (GDP, personal disposable income), future expectations, ADR, and the price of substitutes (market ADR) to be significant in influencing lodging demand at the property level. Price elasticity of demand was significantly inelastic at –0.13, consistent with Qu et al. (2002)’s estimate of –0.15 for Hong Kong. Using a hypothetical example, the authors suggested that discounts would not enhance revenues.

Finally, Corgel et al. (2012) focused on a single year, 2012, to estimate a series of demand curves for the U.S. lodging industry. They estimated price elasticity for six hotel chain scales and selected metro markets using OLS model with a lagged dependent variable. The results of their study showed price (income) elasticity of –0.17 (0.56) for all hotels and –0.15 (0.30) for the top 50 U.S. markets. More important, both price and income elasticity varied across hotel chain scales, consistent with Canina and Carvell (2005).

While lodging demand studies focus on estimating elasticity, tourism demand studies focus on forecasting tourism (Song et al., 2011; Witt & Martin, 1987, 1988). For example, Witt and Martin (1987) investigated the elasticity of demand, measured by tourist arrivals, as the function of income, prices, substitute prices, and dummy variables to capture the impact of external events. Song et al. (2011) used a time-series approach to model the demand for hotel rooms in Hong Kong and generate forecasts assessing the impact of the external shocks including the events of 9/11 and the global financial crisis. Consistent with previous research, the results of their study showed income and relative prices of hotel rooms as the most important factors determining the demand for rooms in Hong Kong.

Summary of Empirical Research

We make several observations from the preceding literature on lodging and tourism demand. First, lodging demand studies tend to focus on macro factors at the aggregate level (Corgel et al., 2012; Fujii et al., 1985; 2; Qu et al., 2002; Tsai et al., 2006; Wheaton & Rossoff, 1998) instead of the property level (Canina & Carvell, 2005; Hiemstra & Ismail, 1990). Accuracy of data and failure to uncover underlying behavioral relationships are issues that arise with aggregated data (Corgel et al., 2012). Damonte et al. (1998) conclude that aggregated price elasticities are of limited use to public policy makers and lodging industry executives. We overcome this limitation by using disaggregated economic data at the state level instead of aggregated data at the national level.

Second, various macroeconomic variables and methodologies are employed to model hotel room demand. To avoid issues with simultaneity and serial correlation, lagged dependent variables (Corgel et al., 2012; Song et al., 2011; Wheaton & Rossoff, 1998; Witt & Martin, 1987, 1988) are used with various estimation techniques from a single equation OLS model (Corgel et al., 2012; Damonte et al., 1998; Hiemstra & Ismail, 1990; Song et al., 2011; Wheaton & Rossoff, 1998) and FGLS (Canina & Carvell, 2005) to simultaneous equation 2SLS (Lee & Jang, 2012; Qu et al., 2002; Tsai et al., 2006). Moreover, the sample data in the extant literature vary from monthly to annual and include single-year and multi-year data periods. Thus, considerable variation exists in variables and methodologies for estimating demand with no consensus emerging on the superiority of any particular methodology. O’Neill and Ouyang (2020) suggest that a three-predictor economic model appears to be best at capturing lodging demand, and any model with more than three predictors suffers from multicollinearity issues.

Third, the empirical evidence on price and income elasticity has produced mixed findings. While some studies found the relationship between price and demand to be significantly inelastic (Canina & Carvell, 2005; Hiemstra & Ismail, 1990; Lee & Jang, 2012; Wheaton & Rossoff, 1998), others failed to demonstrate significance (Damonte et al., 1998; Tsai et al., 2006). Similarly, the relationship between income and demand is also mixed with some studies finding significance (Canina & Carvell, 2005; Wheaton & Rossoff, 1998) and others finding no significance (Qu et al., 2002; Tsai et al., 2006). Furthermore, there is debate as to which proxy is better at capturing income since different measures have been used from disposable income (Corgel et al., 2012; Tsai et al., 2006) to GDP (Canina & Carvell, 2005; Wheaton & Rossoff, 1998). Fourth, prior studies assume estimates of elasticities to be constant over multiple years, an assumption that may not be valid. Damonte et al. (1998) documented variation in aggregate price elasticities across years, and Song et al. (2011) argued that such estimates may be insufficient in situations with a high degree of uncertainty such as an economic crisis. We overcome these limitations by focusing on measuring price elasticity over a relatively short-time period employing property-level data. The following hypotheses arise from the discussion above.

• Hypotheses 1 (H1): Lodging demand in the lodging sector from 2018 to 2021 was inelastic.

• Hypotheses 2 (H2): Income and lodging demand are significantly positively related.

To test these hypotheses, we use both disposable income and GDP as separate predictors in our empirical model and estimate the demand equation using two different estimation techniques (2SLS and FGLS) that are robust to various specification issues that some researchers failed to address in prior research. The use of both estimation techniques is another distinguishing feature of our study. Unlike Canina and Carvell (2005)’s measure of an aggregated market ADR for substitute prices, we use a disaggregated measure of the submarket ADR, controlling for both hotel location and segment. Flowing from demand theory and previous work in lodging we hypothesize:

• Hypotheses 3 (H3): There is a positive relationship between lodging demand and substitute prices.

Finally, with the exception of Hiemstra and Ismail (1990) who applied elasticity to the impact of room taxes, none of the previously mentioned studies linked elasticity estimates with revenue performance at the property level. Further distinguishing this study from the extant literature, we link price elasticity of demand to revenue performance in our sample of hotels to capture the impact of the COVID-19 pandemic. By so doing, we highlight the impact of discounting on revenue performance for a subsample of hotels in the midst of an external shock.

Revenue Performance and Discounting Studies

A series of these lodging demand and pricing-related studies have provided strong evidence of the impact of pricing behavior on hotel operating performance. For example, a review of lodging industry performance after the events of 9/11 (Enz & Canina, 2002) showed variations in operating performance across regions, states, cities, and chain scale segments. Hotel performance behaved in a cyclical manner, and not all regions, states, and cities were hit hard by the recession or by 9/11. This research was limited by focusing only on RevPAR performance without considering occupancy or ADR.

Over a period of years, empirical evidence demonstrated that discounting in the lodging sector does not work (Enz et al., 2004). Using annual performance data from more than 6,000 hotels between 2001 and 2003, the descriptive results revealed that hotels with lower prices relative to their competitive set achieved higher occupancies and captured market share from the competition but did not achieve higher RevPARs. When prices were reduced, given inelastic demand, price promotions were insufficient to stimulate demand to offset revenue losses. Conversely, those hotels with higher prices relative to their competitive sets recorded lower occupancies but higher RevPARs, in which case losses from occupancy declines were offset by higher RevPARs. The conclusion reached was that hotels in direct competition make more money when they maintain rate integrity and avoid discounting to fill rooms during a recession even when competitors do not.

These findings were subsequently corroborated in a study examining discounting during a recessionary period via a sample of more than 67,000 observations from more than 10,000 hotels spanning a 7-year period from 2001 to 2007 (Enz et al., 2009). The descriptive evidence demonstrated that hotels that priced higher than their competitive set had lower occupancies but higher RevPARs. For example, during the downturn, hotels that priced 20% to 30% below their competitive set achieved 15.5% higher occupancies but at the expense of lower RevPARs that were 12% below the competitive set. Conversely, hotels that maintained prices 20% to 30% above the competitive set, recorded 7.1% lower occupancy rates but achieved 8.7% higher comparative RevPARs. The authors noted that the pattern of demand and revenue behavior was consistent for hotels in all market segments, again concluding that the best way for a hotel to achieve higher revenue performance is to maintain higher rates and avoid discounting. Taken together, these studies reveal that hotels engaged in heavy discounting following the events of 9/11 failed to produce the desired results; discounting may increase occupancy, but it will decrease RevPAR (Enz et al., 2004, 2009). The authors argued the best strategy for hotels during a crisis is to maintain or raise rates. The lack of statistical testing is a limitation of these studies. Nevertheless, whether hotels engaged in discounting during the COVID-19 pandemic, and the impact of such a practice during this period, are legitimate empirical questions.

The effect of pricing (ADR) on RevPAR performance for almost 7,000 hotels over a 11-year period from 2000 through 2010 was the subject of another study (Noone et al., 2013). By testing the effect of relative price position and relative fluctuation on annual RevPAR performance, it was determined that revenue performance was strongest for hotels that priced higher than the competition and maintained a consistent relative price over time. Price position had a significant positive relationship with RevPAR, whereas price fluctuation had a significant negative relationship. The results indicated that pricing higher than competitors was associated with stronger RevPAR performance over time. Contrastingly, the greater the amount of price instability or shifting relative to the competitive set, the lower was the RevPAR performance. However, by hotel type, the results were mixed with price fluctuation in two out of five segments insignificant. Given the importance of pricing and its effect on hotel operating performance during a downturn, we assess the effect of pricing position on revenue performance during COVID-19.

A more recent study examined pricing, occupancy, and RevPAR relationships during the economic recession of 2008 to 2009 in a sample of 408 hotels in six submarkets across the United States (O’Neill & Yeon, 2022). Using monthly operating performance data from STR, the authors explored whether and how discounting room rates contributed to RevPAR performance in hotels of various classes. Specifically, the researchers examined the effects of pricing on RevPAR level. They classified the subject hotels into four groups depending on whether their ADR is above or below the competition and whether the ADR is lower or higher than a prior period. It was hypothesized that hotels that continuously discounted rates achieved a relatively higher RevPAR level during the economic recession with variation in this effect across hotel class. The results of the study showed that luxury and upper upscale hotels that discounted rates during the recession generated higher RevPAR levels, whereas hotels that maintained rates generated lower RevPAR levels. Mixed results were found for upscale hotels. In contrast, upper midscale, midscale, and economy hotels achieved higher RevPAR levels by maintaining rates, but their RevPAR suffered when they offered discounted rates. The authors recommended that luxury, upper upscale, and upscale hotel operators selectively offer discounted rates during a recession to increase occupancy and RevPAR. Alternatively, upper midscale, midscale, and economy hotel operators could still offer discounts but needed to exercise caution in offering them early in a recession. These results run contrary to prior findings (Enz et al., 2004, 2009) that hotels should not discount but rather maintain or raise rates.

Differences in the above findings may result from differences in methodologies relating to time periods being analyzed (different external shocks), use of monthly versus annual data, differences in classes of hotels analyzed, and differences in performance measures. While O’Neill and Yeon (2022) focus on only categorical differences in monthly RevPAR, we focus on relationships and the impact of these relationships on annual RevPAR performance. Unlike O’Neill and Yeon (2022), we compute a hotel’s relative price position or RevPAR position or the changes in these positions relative to the competitive set. Thus, our study is closest to Enz et al. (2004, 2009) in using annual revenue performance data to study the effects of discounting on revenue performance during a different external shock.

We overcome limitations in prior studies by assessing the magnitude of the discount and by capturing the effect of the pricing position and changes in that position on RevPAR performance. Our study, using annual performance data, contributes to the hotel discounting research literature by examining whether relative hotel pricing during COVID-19 led to an increase in RevPAR performance. We also explore how a hotel’s competitive market position is changed by its pricing. Our approach of analyzing the data in terms of absolute RevPAR levels and percent changes is an extension of prior research to yield greater insights on the pricing impact on RevPAR performance at the property level and relative to the competitive set. Therefore, the results of our study are relevant and informative for competitive and strategic pricing decisions in the midst of an external shock, by providing empirical evidence on the relationship between demand, pricing, and revenue performance. Based on previous work we expect that:

• Hypotheses 4 (H4): Maintaining rate integrity (not discounting) will result in higher RevPAR.

Methodology

Lodging Demand Estimation

Microeconomic theory and prior empirical research provide guidance for specifying the empirical model for estimating demand for lodging. In the first stage of our empirical analysis, we specify lodging demand as a function of room price, income, and the price of rooms at substitute hotels (Canina & Carvell, 2005). Lodging demand (Demand) is measured as the number of rooms sold at the individual hotel level. Price is measured as the hotel ADR (ADR), defined as the room revenue divided by the number of rooms sold. We expect a negative relationship between demand and price. Income is measured in two ways, personal disposable income (Dispinc), and real GDP (RGDP). Unlike prior research that used aggregated national data, we measure income at the state level. We expect a positive relationship between demand and income. The price of substitutes is measured using the submarket competitive set’s average ADR (SubADR). The submarket competitive set is defined by the location and chainscale in which the hotel competes with other similar hotels. To capture the effect of COVID-19 on demand, we include a dummy variable that is equal to one if the year is 2020, otherwise zero (Y2020). We posit a negative relationship between demand and the proxy for the COVID-19 pandemic. We include a lagged dependent variable (LDemand) in the empirical model assuming a partial adjustment process that reflects supply constraints, which takes time to adjust as well as to avoid simultaneity (Canina & Carvell, 2005; Corgel et al., 2012; Song et al., 2011; Wheaton & Rossoff, 1998; Witt & Martin, 1987, 1988). With the exception of the COVID-19 dummy, all variables are transformed to natural log (ln) form. Consequently, we employ a log-linear functional form for the empirical model so coefficients can be easily interpreted as elasticities. To mitigate heteroskedasticity and autocorrelation with a balanced panel, we estimate the regression equation using FGLS (Canina & Carvell, 2005). The empirical specification of the lodging demand model takes the following form:

$$\begin{array}{l}{\mathrm{Demand}}_{\mathrm{it}}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{\mathrm{ADR}}_{\mathrm{it}}+{\mathrm{\beta }}_2{\mathrm{Dispinc}}_{\mathrm{it}}\left({\mathrm{or}\mathrm{RGDP}}_{\mathrm{it}}\right)\\ +{\mathrm{\beta }}_3{\mathrm{SubADR}}_{\mathrm{it}}+{\mathrm{\beta }}_4{\mathrm{LDemand}}_{\mathrm{it-1}}\\ +{\mathrm{\beta }}_5{\mathrm{Y2020}}_{\mathrm{it}}+{\epsilon }_{\mathrm{it}}\end{array}$$ (1)

Estimating demand as a function of price does not take into account that price is also a function of demand. Therefore, consistent with the theory of demand and supply, both demand and price are considered to be endogenous variables, requiring the estimation of a simultaneous equation model via 2SLS to provide consistent parameter estimates (Lee & Jang, 2012; Qu et al., 2002; Tsai et al., 2006). The 2SLS approach is robust to various estimation problems including multicollinearity and specification errors, and therefore the appropriate model specification (Kennedy, 1998). In the first stage of the 2SLS approach, the endogenous variables are regressed on the exogenous variables to derive predicted values for the second stage to estimate the structural equations. We employ Equation 1 as the demand equation in the 2SLS approach to maintain consistency and comparison. For the second equation, we specify ADR as a function of demand (Demand), lagged ADR (LADR), hotel occupancy rate and market room supply. Occupancy (Occ) is defined as the number of rooms sold divided by the number of available rooms, whereas room supply (Supply) is defined as the total number of available room nights in the market in which the hotel operates. We use all exogenous variables in the system of equations as instruments for the endogenous variables. The specification of the second equation takes the following form:

$$\begin{array}{l}{\mathrm{ADR}}_{\mathrm{it}}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{\mathrm{Demand}}_{\mathrm{it}}+{\mathrm{\beta }}_2{\mathrm{LADR}}_{\mathrm{it-1}}+{\mathrm{\beta }}_3{\mathrm{Occ}}_{\mathrm{it}}\\ +{\mathrm{\beta }}_4{\mathrm{Supply}}_{\mathrm{it}}+{\epsilon }_{\mathrm{it}}\end{array}$$ (2)

We use the ADR coefficient from the 2SLS estimation to examine the impact of price elasticity on revenue performance. In addition, we estimate the impact of pricing during the COVID-19 pandemic using two hotel subsamples: hotels that raised rates and hotels that discounted rates. In this regard, we go beyond the empirical analysis to illuminate the practical application of our elasticity estimates on the two hotel subsamples.

Revenue Performance Model Specification

In the second stage of our empirical analysis, we investigate COVID-19’s impact on revenue performance. We assess the impact of each hotel’s submarket pricing position on hotel RevPAR performance along with controls for various property characteristics. Competitive sets are a key feature in the empirical analysis as a hotel’s performance is directly influenced by its competitors within its local market (Enz et al., 2004, 2009). Competitive set performance data are based on the submarket location and scale of the hotel as defined by STR. For example, the performance of an upper upscale hotel located in the Houston CBD submarket would be compared to other upper upscale hotels located within the same submarket. Penetration indices for occupancy, ADR and RevPAR for each hotel were computed by dividing each hotel’s measure by the submarket scale competitive set average for the same metric. The percentage change in the penetration index indicates the change in hotel performance relative to the competitive set. For example, a percent change in the RevPAR penetration index from 2019 to 2020 captures the overall impact on a hotel’s market position from the previous year. Assume that an upscale hotel achieved a RevPAR of US$126 in 2019 relative to US$120 for its upscale competitive set. Dividing US$126 by US$120 yields a RevPAR index of 105 for 2019, indicating that the hotel’s overall market position is 5% higher than its competitive set. Assuming during COVID-19 in 2020, the same upscale hotel’s RevPAR declined to US$115, whereas the competitive set RevPAR dropped to US$100, the RevPAR index for the upscale hotel for 2020 would be 115 (115/100 = 1.15) or 15% higher than its competitive set. The percent change in the RevPAR index is 9.52% (115/105 = 1.0952–1 = .0952), indicating the hotel’s overall market position improved relative to its competitive set from the previous year. The second stage multiple regression estimated via OLS in levels takes the following form with the variables defined in the next section:

$$\begin{array}{l}{\mathrm{RevPAR}}_{\mathrm{i}}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{\mathrm{Size}}_{\mathrm{i}}+{\mathrm{\beta }}_2{\mathrm{Age}}_{\mathrm{i}}+{\mathrm{\beta }}_3{\mathrm{SubADR}}_{\mathrm{i}}\\ +{\mathrm{\beta }}_4{\mathrm{Y2020}}_{\mathrm{i}}+\sum {\mathrm{\beta }}_5{\mathrm{Type}}_{\mathrm{i}}+\sum {\mathrm{\beta }}_6{\mathrm{Location}}_{\mathrm{i}}\\ +\sum {\mathrm{\beta }}_7{\mathrm{Scale}}_{\mathrm{i}}+\sum {\mathrm{\beta }}_8{\mathrm{State}}_{\mathrm{i}}+{\epsilon }_{\mathrm{i}}\end{array}$$ (3)

To provide further insights, we also assess the performance of the sample hotels relative to the competitive set and its impact on submarket occupancy and RevPAR penetration during the COVID-19 pandemic. Specifically, we regress changes in the hotel fundamentals (occupancy and ADR) on changes in the submarket occupancy and submarket RevPAR in 2020.

Revenue Performance Variables

Consistent with prior research, we employ RevPAR as the key dependent variable for our measure of revenue performance (Enz & Canina, 2002, Enz et al., 2004, 2009; Noone et al., 2013; O’Neill & Carlbäck, 2011; O’Neill & Yeon, 2022). RevPAR is defined as rooms revenue divided by total rooms available to control for the size effect and ensure consistency in comparison across hotels. Using available rooms as a scaling variable not only controls the size effect but also enables comparison of differences across hotels. The change in RevPAR is measured as the percentage change in RevPAR levels.

Hotel size (Size) is measured as the log of the number of rooms to control for size effects. Hotel age (Age) is measured as the natural log of the difference between the year built and the onset year (2020) of the pandemic. The COVID-19 effect is captured via a dummy variable designating 2020 (Y2020) as the COVID-19 year and as equal to one, otherwise zero. To capture the pricing position, we employ the submarket competitive set ADR (SubADR), defined as the hotel ADR divided by the submarket location and scale ADR (Enz et al., 2004, 2009; Noone et al., 2013). We also interact the pricing position and COVID-19 dummy (SubADR*Y2020) to determine the pandemic effect on the hotel pricing position in 2020.

Categorical variables captured hotel type (Type) as limited service (omitted group), full service, select service, and extended stay. Researchers have also used hotel location (Location) and chain scale segment (Scale) to explain variations in hotel market values (Noone et al., 2013). This study employs categorical variables to capture hotel location, and hotel segment following current STR definitions for these characteristics. Accordingly, location includes the following categories: small town, highway, resort, airport, suburban (omitted group), and urban/city. Similarly, hotel chain scale segment is categorized as economy, midscale, upper midscale (omitted group), upscale, upper upscale, and luxury. Finally, indicator variables designate the 50 U.S. states (State) where the hotels are located.

Sample Data

The sample data for the study span the period from 2018 to 2021, 2 years before the pandemic and the 2 years during the pandemic from 2020 through 2021. The property level annual revenue performance data used in this study are a subset of a larger data set comprised of three major pieces: performance data, financial statement information, and loan financing data, for hotels that were securitized in various publicly offered commercial mortgage-backed securities (CMBS) transactions. The CMBS hotel performance data were obtained from two main sources. First, the Bloomberg Professional Database was used to obtain property-level operating revenue performance data from CMBS servicer reports. These servicer reports provide annual operating performance, financial statement, and financing information on securitized hotels in various CMBS transactions. Second, we supplement the Bloomberg data with additional performance data from Trepp, LLC that was missing in the Bloomberg data. Trepp is the leading provider of CMBS data. For the scope of this study, we needed only the hotel performance data that included occupancy rates, ADR, and RevPAR from 2018 to 2021.

Submarket and market scale competitive set data including supply and demand, operating fundamentals, and hotel attributes such as operation, location, and scale were graciously provided by the STR Share Center. We matched these data to our performance data using the physical address of the hotel. We did not use performance data from STR because it was already included in our broader CMBS data set. Moreover, STR would have required “disguising” the performance data due to the firm’s confidentiality requirements to avoid identifying properties. This would have prevented us from using the STR performance data with our financial statement and loan data to investigate issues beyond the scope of this study. Finally, economic data (state level GDP and personal disposable income) for estimating our lodging demand are obtained from the Bureau of Economic Analysis (BEA). Our final balanced sample in levels (and percent changes) includes a total of 10,012 (7,509) annual hotel observations representing 2,503 hotel observations over 4 (3) years from 2018 to 2021.

Results

Frequency Statistics

Table 1 presents the frequency statistics for the overall sample data from 2019 to 2020. Out of 5,221 hotel observations, 94% of the sample is comprised of branded hotels, and the remaining 6% represent independent hotels. Of the branded hotels, 87% are franchised and 7% are chain managed. By hotel type, the results show that limited-service hotels represent 40% of the sample followed by extended stay hotels (22%), full-service (19%), and select service hotels (18%). By location, slightly more than half the sample hotels are located in suburban areas with 14% in small metro towns and another 11% in urban locations. Airport, interstate, and resort locations account for less than 10% of each of the overall sample hotel locations. The results also indicate that three-fourths of the sample hotels are represented by upper midscale and upscale hotels followed by upper upscale hotels (10%). Together, independents, economy, midscale, and luxury hotels comprise 15% of the overall sample hotel scale. By region, 30% of the sample hotels are South Atlantic, followed by East North Central (14%) and West South Central (14%), and Pacific (13%). The states with the most hotels in these regions include Texas, Florida, California, Georgia, and North Carolina, respectively.

Table 1.

Hotel Frequency Statistics (n = 2,503 Hotels).

Property characteristics	No.	Percent (%)
Branded:	2,377	95.0
Chain managed	187	7.5
Franchised	2,190	87.5
Independent	126	5.0
Hotel type:
Limited service	957	38.2
Full service	468	18.7
Select service	451	18.0
Extended stay	627	25.0
Location:
Interstate	216	8.6
Resort	158	6.3
Small metro/town	328	13.1
Airport	217	8.7
Suburban	1,305	52.1
Urban	279	11.1
Scale:
Economy	176	7.0
Midscale	86	3.4
Upper midscale	1,002	40.0
Upscale	850	34.0
Upper upscale	243	9.7
Luxury	20	0.8
Independents	126	5.0
Region:
East North Central	365	14.6
East South Central	153	6.1
Middle Atlantic	182	7.3
New England	52	2.1
Mountain	211	8.4
Pacific	314	12.5
South Atlantic	734	29.3
West North Central	140	5.6
West South Central	352	14.1

Descriptive Statistics

Descriptive statistics on hotel size and age are presented in Panel A of Table 2 while Panel B highlights the average KPIs for 2019 and 2020. Panel C shows the impact of the pandemic on the KPIs whereas Panel D reflects the room supply. The average hotel had 138 rooms and 23 years of age as of the end of 2020. In 2019, the average sample hotel operated at an occupancy rate of 74% and earned an ADR of US$131 with an average net impact on RevPAR of US$98. The impact of the pandemic shows a dramatic downfall in hotel fundamentals. Average occupancy rates dropped to 49%, whereas ADR settled at US$110. Taken together, the results show RevPAR for 2020 dropped to US$54. It is clear from the results in Panel C that occupancy rates declined (34%) to a greater extent than ADR (15%). Consequently, the results show hotel RevPAR performance plummeted 44% in 2020 due to COVID-19. Although the data show the average number of available room nights decreasing 3.7% from 787,000 in 2019 to 758,000 in 2020 (Table 2 Panel D), there is variation across hotel attributes as some submarkets faced an increase in room supply, whereas others experienced a decline.

Table 2.

Sample Descriptive Statistics

Panel A: Size and age	N	M	Median	SD	Min	Max
Size (rooms)	2,503	140	112	114	16	2860
Ln (size)	2,503	4.79	4.72	0.51	2.77	7.96
Age (years)	2,503	22.2	19	18.02	2	293
Ln (age)	2,503	2.88	2.94	0.67	0.69	5.68
	2019		2020		2021
Panel B: KPIs	M	Median	M	Median	M	Median
Occupancy	.744	.753	.497	.485	.647	.660
ADR ($)	129.36	119.63	109.16	101.15	123.06	111.48
RevPAR ($)	96.53	89.11	52.04	46.40	78.55	71.01
Occupancy recovery (%)	––	––	66.8	64.4	87.0	87.6
ADR recovery (%)	––	––	84.4	84.6	95.1	93.2
RevPAR recovery (%)	––	––	53.9	52.1	81.4	79.7
Panel C: % Change in KPIs	M	Median	M	Median	M	Median
Δ Occupancy	0	–.008	–.333	–.334	.388	.325
Δ ADR	.013	.004	–.142	–.143	.138	.111
Δ RevPAR	.009	–.005	–.426	–.444	.547	.491

Note. KPIs are key performance indicators including occupancy and average daily rate (ADR) and revenue per available room (RevPAR). Occupancy is the number of rooms sold divided by rooms available. ADR is defined as rooms revenue divided by rooms sold. RevPAR is defined as rooms revenue divided by rooms available. Change is measured as the percentage change over a year. Recovery in KPI is measured as the current year KPI divided by the 2019 which is the base year. ADR = average daily rate.

To provide further insights, we disaggregate the changes in KPI by various hotel attributes as shown in Table 3, Panels A–D. Variation in KPIs across various hotel characteristics is evident across both years. We performed tests of differences (t-tests, analysis of variance) across hotel attributes for each KPI, but in the interest of conserving space, highlight only a few significant differences. In Panel A, the data show that occupancy declines in branded hotels are significantly lower than independent hotels, (t = 2.95, p = .004). Although the decline in ADR is lower in independent hotels relative to branded hotels, the difference is not significant (t = 0.33, p = .744). The data suggest that independent hotels were more successful in maintaining rates than branded hotels. The impact of both operating metrics on RevPAR changes is significantly different between branded and independent hotels in 2020 (t = 2.26, p = .025). The above results are reversed in 2021 with independents significantly outperforming branded hotels in 2021 across all three metrics (occupancy t = 2.05**; ADR t = 3.57***; RevPAR t = 4.27***).

Table 3.

Average Percent Change in KPIs by Hotel Characteristics, 2019-2021.

Hotel Characteristics	N	Occupancy		ADR		RevPAR
Panel A: By operation		2020	2021	2020	2021	2020	2021
Branded	2,377	–.330	.379	–.142	.127	–.424	.532
Independent	126	–.394	.546	–.134	.332	–.477	.841
Panel B: By type
Limited service	957	–.317	.370	–.138	.176	–.415	.562
Full service	468	–.495	.562	–.149	.129	–.561	.718
Select service	451	–.412	.484	–.182	.147	–.515	.677
Extended stay	627	–.182	.216	–.112	.079	–.271	.303
Panel C: By location
Interstate	216	–.259	.312	–.115	.143	–.342	.492
Resort	158	–.356	.506	–.105	.254	–.432	.809
Small metro/town	328	–.254	.317	–.097	.157	–.341	.473
Airport	217	–.381	.439	–.158	.083	–.477	.534
Suburban	1,305	–.326	.370	–.144	.119	–.419	.513
Urban	279	–.472	.508	–.211	.173	–.583	.700
Panel D: By scale
Economy	176	–.016	.062	–.043	.130	–.061	.195
Midscale	86	–.216	.413	–.016	.110	–.297	.394
Upper midscale	1,002	–.321	.357	–.147	.149	–.420	.545
Upscale	850	–.353	.402	–.165	.109	–.457	.543
Upper Upscale	243	–.538	.610	–.152	.090	–.609	.716
Luxury	20	–.550	.344	–.185	.318	–.643	.699
Independents	126	–.394	.546	–.134	.332	–.477	.841

Note. ADR = average daily rate.

In Panel B, the results indicate that extended stay hotels outperformed all other hotel types in 2020. An 18% decline in occupancy combined with a 11% decline in rates reduced RevPAR by an average of 27% for extended stay hotels. In contrast, a 50% drop in occupancy coupled with a 15% decline in rates resulted in a 56% decline in RevPAR for full-service hotels. A test of differences (analysis of variance [ANOVA]) finds significantly lower RevPARs in 2020 for full-service hotels relative to all other hotel types. The lack of group and business travel during the pandemic took a heavy toll on full-service hotels. On the contrary, leisure travel fueled much of the demand in extended stay hotels, which also benefited from longer lengths of stay from health care workers and travelers facing restrictions and shutdowns. In 2021, we observe full-service hotels rebounding dramatically in RevPAR performance, driven by higher occupancies.

Panel C shows that hotel location was a factor in driving performance differences in 2020. The impact on revenue performance is lowest for hotels located on interstates and small metro towns. In contrast, urban hotels were the most severely affected by the pandemic. Urban locations, especially downtown where most full-service hotels are located, suffered a 58% drop in RevPAR performance compared to a 34% drop in RevPAR for interstate and small metro town locations. A test of differences finds significantly lower RevPARs for urban hotels relative to all other hotel locations in 2020. Even the performance of airport hotels suffered from the precipitous drop in air travel from the widespread restrictions and shutdowns across the country. Leisure travelers took to the highways, traveling across the country through many small towns, which benefited hotels in those locations. In 2021, urban and resort locations rebounded strongly driven by higher occupancies. Tests of differences indicate urban locations enjoyed significantly higher RevPARs than all other locations except resort hotels. The results in Panel D reveal that revenue performance also differs by hotel chain scale. More specifically, hotels in the economy segment had the lowest decline (6%) in RevPAR performance in 2020, driven largely by rate declines with occupancies almost unchanged for this segment. Meanwhile, upper upscale and luxury hotels suffered the greatest declines in RevPAR performance of at least 60%, fueled by substantially greater decline in occupancies. Full-service upper upscale and luxury hotels tend to be located in urban downtown areas and central business districts and are heavily dependent on group and business travel, which declined steeply during the pandemic. A test of differences finds the RevPARs of these two hotel scales significantly lower than all other segments in 2020. However, in 2021, we observe a dramatic reversal in RevPAR for upper upscale, luxury, and independent hotels, driven by occupancy in upper upscale and independents, and pricing in luxury and independent hotels. Tests of differences indicate RevPARs of upper upscale hotels are significantly higher than all other segments except luxury and independent hotels. Similarly, the RevPARs of independent hotels are significantly higher than all segments except upper upscale and luxury hotels.

Lodging Demand Analysis

The results of estimating Equations 1 and 2 for lodging demand are presented in Table 4 Panels A and B. The results estimated via FGLS are shown in Panel A, whereas Panel B provides the 2SLS results. Multicollinearity is found to be a nonissue because none of the variance inflation factors exceeded a value of 5. Adding additional economic predictors led to multicollinearity issues. The parameters are estimated with heteroskedasticity and autocorrelation robust standard errors (HAC). Models 1 (3) and 2 (4) in Panel A are estimated with personal disposable income (real GDP). In addition, we include an interaction between ADR and the COVID-19 dummy in Models 2 and 4. The positive and significant coefficients for substitutes in both models demonstrate a positive relationship between substitute pricing and demand, supporting H3.

Table 4.

Lodging Demand Estimation.

Panel A: FGLS	Model 1: Demand	Model 2: Demand	Model 3: Demand	Model 4: Demand
Constant	1.790***	1.346***	1.864***	1.429***
ADR	–.069***	.022**	–.068***	.023**
Dispinc	.013***	.014***
RGDP			.007***	.007***
SubADR	.051***	.048***	.055***	.052***
LDemand	.827***	.829***	.827***	.828***
Y2020	–.514***	1.086***	–.513***	1.088***
ADR*Y2020		–.345***		–.345***
N	7509	7509	7509	7509
Wald chi-square	66592***	67514***	66134***	66763***
DW stat	2.16	2.18	2.30	2.28
Ave VIF	2.03	2.54	2.03	2.54
Panel B: 2SLS	Demand	Demand
Constant	2.069***	2.135***
ADR	–.165***	–.165***
Dispinc	.012***
RGDP		.007**
SubADR	.173***	.176***
LDemand	.790***	.790***
Y2020	–.552***	–.551***
N	7509	7509
R 2	.769	.769
F value	3937***	3943***
DW stat	2.16	2.30
Ave VIF	2.03	2.03
Panel C:	ADR	ADR
Constant	.408***	.408***
Demand	.057***	.057***
LADR	.861***	.861***
Occ	.081***	.081***
Supply	–.021***	–.021***
R 2	.775	.774
F value	2924***	2919***
DW stat	2.06	2.06
Ave VIF	1.17	1.17

Note. All variables are expressed in natural logs except Y2020. Feasible generalized least squares (FGLS) regression in Panel A. Two-stage least squares (2SLS) regression in Panel B and C. Demand is the number of room nights sold. ADR is measured as rooms revenue divided by rooms sold. Dispinc is the personal disposable income at the U.S. state level. RGDP is the U.S. state real gross domestic product (GDP) chained to 2012 dollars. SubADR is the submarket ADR position measured as the (property ADR/submarket ADR) - 1. Y2020 is a dummy variable equal to one for year 2020 COVID-19 year otherwise zero. LDemand is lagged demand. LADR is the lagged ADR. Supply is the number of available rooms in the market. VIF is the variance inflation factor as an indicator of multicollinearity.

^*, **, *** significance at 10%, 5%, and 1% levels, respectively, based on heteroscedasticity and autocorrelation robust standard errors (HAC).

The results in Models 1 and 3 show that ADR is significant and negatively associated with demand consistent with theoretical and hypothesized expectations. Thus, a 1% increase (decrease) in ADR will reduce (raise) demand by 0.069%, indicating that lodging demand is relatively inelastic and providing support for H1. Income is also significantly and positively correlated with demand across all four models, consistent with theoretical expectations and supporting H2. Thus a 1% increase in disposable income (real GDP) will increase demand by 0.013% (0.007%), indicating relative income inelasticity for hotel demand. Although this finding is consistent with prior research, the magnitude of the income coefficient is lower relative to previous studies (Canina & Carvell, 2005; Corgel et al., 2012; Wheaton & Rossoff, 1998). It differs from the income nonsignificance by Qu et al. (2002) and Tsai et al. (2006).

The cross-price elasticity as measured by the submarket ADR is also found to be significant and positive. The implication is that a 1% increase (decrease) in submarket ADR will increase (decrease) a hotel’s demand by 0.05%. Thus, lodging demand is relatively inelastic relative to changes in competitors’ prices. Although consistent with Canina and Carvell (2005), this important finding is obtained by using a more refined submarket scale competitive set ADR than the aggregated market average ADR by Canina and Carvell (2005), the only other study to estimate the cross-price elasticity. The coefficient of the lagged dependent variable is positively significant and inelastic across all models, which suggests that a 1% increase in prior year demand will increase current demand by 0.83%. More importantly, we observe a significantly negative impact of COVID-19 on lodging demand in Models 1 and 3. When the COVID-19 dummy is interacted with ADR (ADR*Y2020), the results show a significant and negative interaction coefficient in Models 2 and 4. While the main effect of ADR is positive, the effect of ADR on demand due to COVID-19 is significantly lower in 2020 relative to other years.

In Panel B, the 2SLS results for the demand equation provide similar results to those in Panel B in terms of sign and significance of the coefficients. More important, the price elasticity coefficient of –0.165, although negatively significant and relatively inelastic, is more than double the coefficient in Panel A. The results suggest that a 1% increase (decrease) in ADR will decrease (increase) demand by 0.165%. This price elasticity estimate is similar in magnitude to some studies (Canina & Carvell, 2005; Corgel et al., 2012; Qu et al., 2002) and lower than others (Hiemstra & Ismail, 1990; Lee & Jang, 2012; Wheaton & Rossoff, 1998). It differs from the price elasticity nonsignificance by Damonte et al. (1998) and Tsai et al. (2006).

The coefficients for disposable income, lagged demand, and COVID-19 dummy, are similar in magnitude to those in Panel A, whereas the submarket ADR coefficient is substantially larger. In Panel C, we provide the results for the supply equation estimated using 2SLS. The results show that demand is significantly and positively associated with ADR. A 1% increase in demand will increase ADR by 0.06%. Similarly, the lagged ADR is significant and positive indicating the current ADR is significantly affected by the previous year’s ADR. Thus a 1% increase in previous year ADR is associated with a 0.86% increase in current ADR. Hotel occupancy rate has a significantly positive effect on room price. A 1% increase in the room occupancy rate will increase ADR by 0.08%. Finally, the supply of rooms has a significant and negative effect on ADR. A 1% increase in market room supply will reduce ADR by 0.02% consistent with the observation that excess supply has a depressing effect on lodging fundamentals when supply exceeds demand.

Elasticity Impact on Revenue Performance

In this section, we highlight the impact of the price elasticity of demand on revenue performance during the COVID-19 pandemic in 2020. Using the sample data means, we first present an example to demonstrate the impact of inelastic demand on room revenue and then provide the relevant empirical evidence for two specific subsamples: hotels that raised rates and hotels that discounted rates. We do not consider the impact of variable costs associated with selling rooms as this information is unavailable to us.

The average size of all hotels in our sample is 140 rooms. These hotels had an average occupancy of 74.42% in 2019 indicating demand of 38,029 rooms sold at an average ADR of US$129.40, which generated US$4,920,953 (38,029 * US$129.4) in room revenue for 2019. Given price inelasticity of –0.165 from our 2SLS regression, this would imply that a 10% price discount will increase demand by only 1.65% (10% * 0.165) and therefore, not enhance revenue. By reducing rates to US$116.46 (US$129.40—US$12.94) in 2020, room revenue will be US$4,501,878 (38,656 * US$116.46), a decrease of US$419,075 despite selling an additional 627 rooms (38,029 * 1.65%). Although the additional demand increased revenues by US$73,020 (627 * US$116.46), this gain in revenue from additional rooms sold is not offset by a loss of US$492,095 (38,029 * US$12.94) from reducing prices, resulting in a net loss of US$419,075. Hence, when demand is inelastic, a rate reduction is insufficient to enhance revenues because the change in quantity demanded rooms is insufficient to offset the change in price, resulting in a decline in room revenue. Consequently, the wise decision, given inelastic demand, is for hotel owners to raise rates.

The results of our subsample analysis are presented in Table 5. We refer to the subsample of hotels that raised rates as RR (100 hotels) and hotels that lowered rates as LR (2,403 hotels). The data in Table 5 Panel A indicate that RR hotels, on average, had 116 rooms, sold 31,872 rooms at an ADR of US$100.08 and generated US$3.190 million in room revenue. In 2020, this group of hotels sold 29,523 rooms at an average ADR of US$111.14 that produced US$3.282 million in room revenue, an increase of US$91,438 over the prior year. Thus, rooms sold decreased by 2,349 rooms, whereas rates rose by US$11.06, which implied an elasticity of –0.67([2,349/31,872]/(US$11.06/US$100.08]). We observe the revenue gain from raising rates (31,872 * US$11.06 = US$352,504) more than offset the revenue loss from selling fewer rooms (–2,349 * US$111.14 = –US$261,066), resulting in net increase in revenue of US$91,438 for RR hotels. Thus, in the midst of the pandemic, by raising rates, owners of RR hotels improved revenue performance when faced with inelastic demand. A closer analysis of the subsample of 100 hotels finds that 64% are extended stay hotels and the remaining 26% are limited-service hotels. By scale, more than half (52%) were economy hotels followed by upper midscale (17%), independent (14%), and upscale (12%) hotels.

Table 5.

Impact of Price Elasticity on Revenue Performance.

Panel A: RR Hotels	2019	2020	Change	% Change
Number of rooms	116	116
Occupancy	75.28%	69.73%
Demand	31,872	29,523	–2,349	–7.37
ADR ($)	100.08	111.14	11.06	11.05
Elasticity				–0.67
RevPAR ($)	75.34	77.50
Room revenue ($)	3,189,748	3,281,187	91,438	2.87
Panel B: LR Hotels
Number of rooms	141	141
Occupancy	74.63%	44.63%
Demand	38,407	22,967	–15,440	–40.20
ADR ($)	130.58	109.07	–21.51	–16.47
Elasticity				2.44
RevPAR ($)	97.45	48.67
Room revenue ($)	5,015,585	2,505,008	–2,510,177	–50.0

Note. Sample averages for all figures. Sample size is 100 hotels for RR subsample and 2,043 hotels for LR subsample. RR (LR) subsample represents hotels that raised (lowered) rates in 2020. Elasticity is computed as percent change in demand divided by percent change in ADR.

The average size of LR hotels in Table 5 Panel B is 141 rooms with demand for 38,407 rooms that were sold at an ADR of US$130.58 and generated US$5,015.6 million in room revenue in 2019. In 2020, demand for LR hotels dropped to 22,967 rooms at an ADR of US$109.07, which produced US$2.505 million in room revenue. Thus, demand declined by 15,440 rooms, whereas ADR dropped by US$21.51 resulting in revenues falling by US$2.511 million from the previous year. We observe a negative change in quantity demanded of –40.2% and a price discount of –16.5%, which implies an elastic demand of 2.44, inconsistent with our empirical estimates of an inelastic demand. Consequently, both changes in price and quantity demanded produced a negative effect on revenue performance for LR hotels. The change in demand produced a loss of US$1.684 million (–15,440 * US$109.07), whereas the change in price resulted in a loss of US$826,134 (–21.51 * 38,407). The average combined losses of a negative US$2.511 million suggest that reducing rates in the midst of a pandemic was not a wise decision for LR hotels as it only compounded what was already a bad situation and made it worse. These findings provide unequivocal support for H4.

Taken together, the results of our analysis provide empirical evidence of two different revenue performance outcomes for two groups of hotels during the COVID-19 pandemic. In one case, hotels that raised rates had superior revenue performance, whereas, in the other case, hotels that lowered rates during the COVID-19 pandemic achieved worse revenue performance. The takeaway from this analysis for hotel owners and managers is clear. Given an inelastic demand for lodging, hotel owners, and managers should not lower rates. Instead, hotel owners and managers should raise rates when confronted with adverse economic conditions or external shocks. The empirical evidence is clear that revenue gains from raising rates will more than offset the revenue loss from selling fewer rooms.

Impact of Submarket Position on RevPAR Levels

We employed an OLS regression in which RevPAR levels were regressed on a set of hotel characteristics and the submarket ADR position. Since multicollinearity is found to be a nonissue and a Woodridge test for autocorrelation is insignificant (F = 1.76, p = .19), we estimate the parameters with heteroskedasticity robust standard errors. The results of this analysis are presented in Table 6 Model 1. In Model 2, we also interact submarket ADR with the COVID-19 dummy variable to evaluate the effect of pricing on RevPAR performance in 2020 in Model 2. The results indicate the overall models are significant in explaining RevPAR performance with the predictors explaining 64% of the variation in RevPAR. Both hotel size and age have a significant and negative effect on RevPAR which implies that larger size and older hotels have lower RevPAR levels. The results reveal a significantly positive association between the relative price position and RevPAR performance, consistent with prior research (Noone et al., 2013). The positive price coefficient suggests that pricing higher than the competitive set will improve RevPAR performance. The COVID-19 coefficient in Model 1 suggests that the COVID-19 pandemic had a significant negative effect on RevPAR performance. The analysis reveals that hotel type, location, and scale are important predictors of RevPAR. For example, full service and extended stay hotels have significantly higher RevPARs than limited-service hotels. By location, with the exception of resorts, the RevPAR level of hotels in all other locations is significantly higher than suburban hotels. By scale, economy and midscale RevPAR performance is significantly lower than upper midscale hotels, whereas upscale, upper upscale, and luxury RevPAR performance is significantly higher than upper midscale hotels. The results in Model 2 are similar to those in Model 1. More importantly, the interaction between the price position and COVID-19 dummy is also negatively significant. It indicates the overall pricing effect on RevPAR performance was significantly and negatively impacted by COVID-19 in 2020 relative to other years.

Table 6.

Impact of Submarket Position on RevPAR Levels (2018 – 2021).

Variables	Model 1: RevPAR	Model 2: RevPAR
Constant	177.777***	176.469***
Size	–12.463***	–12.164***
Age	–4.991***	–5.043***
SubADR	.804***	.896***
Y2020	–38.292***	–37.588***
SubADR*Y2020		–.427***
Hotel type:
Full service	13.102***	13.190***
Select service	3.283*	3.109*
Extended stay	10.790***	11.027***
Location:
Interstate	5.187***	5.143***
Resort	19.384***	19.284***
Small metro/town	6.371***	6.476***
Airport	3.783***	3.843***
Urban	11.339***	11.236***
Scale:
Economy	–25.892***	–25.176***
Midscale	–19.618***	–18.667***
Upscale	11.620***	11.534***
Upper upscale	27.537***	27.078***
Luxury	68.063***	66.722***
Independent	42.270***	42.393***
State: 50 U.S. states	Included	Included
N	10,012	10,012
R 2	.636	.650
F-value	126***	153***
Ave VIF	1.44	1.44

Note. Dependent variable is RevPAR measured as rooms revenue/available rooms. Hotel size and age are logged. SubADR is the submarket ADR position measured as the (property ADR/submarket ADR)—1. Y2020 is a dummy variable equal to one for year 2020 COVID-19 year otherwise zero. Limited service is the omitted group for hotel type. Suburban is the omitted group for location. Upper Midscale is the omitted group for scale. Results for U.S. states are suppressed to conserve space. VIF is the variance inflation factor as an indicator of multicollinearity.

^*, **, *** significance at 10%, 5%, and 1% levels, respectively, based on heteroscedasticity-robust standard errors (HAC).

Impact of Pricing on Hotel Submarket Performance

In this section, we assess the performance of the sample hotels relative to the relative to the competitive set and its impact on market RevPAR penetration in 2020. For this purpose, we employed STR submarket scale competitive set data based on the submarket location and scale of each hotel in the data set. The results of this analysis are presented in Table 7 Panels A and B. Panel A presents the impact of pricing on relative market occupancy, whereas Panel B shows the net impact to market RevPAR position. To provide further insights, a subsample analysis of hotels with positive (Model 2) and negative changes (Model 3) in pricing is performed. The overall regression in Panel A shows a significantly positive relationship between changes in occupancy and changes in market occupancy across all models. In contrast, changes in pricing have a significantly negative impact on the market occupancy index except for hotels that raised rates in Model 2, which showed no significant difference in pricing. Thus, the overall results indicate a 1% increase in occupancy is associated with a 0.80% improvement in occupancy index, whereas a 1% increase in ADR is associated with –0.22% decrease in the occupancy index. These results suggest that positive pricing changes did not have a detrimental effect on occupancy penetration whereas negative pricing changes did (Panel A Model 3). A Wald chi-square comparison of the ADR coefficients between Models 2 and 3 is marginally significant. The evidence suggests that hotels would have been better off raising rates during the pandemic despite taking a hit to occupancy during the COVID-19 crisis.

Table 7.

Impact of Pricing on Submarket Performance in 2020.

	Model 1: Total	Model 2: Δ ADR > 0	Model 3: Δ ADR < 0
Panel A: Variables	Δ Submarket Occ	Δ Submarket Occ	Δ Submarket Occ
Constant	.252***	.142***	.264***
Δ Occ	.803***	.435***	.828***
Δ ADR	–.222**	–.005	–.216**
N	2503	100	2403
R 2	.510	.095	.523
F value	1028***	8***	1024***
Wald chi-square: Δ Occ			7.61***
Wald chi-square: Δ ADR			3.45*
Ave VIF	1.0	1.39	1.0
	Total	Δ ADR > 0	Δ ADR < 0
Panel B: Variables	Δ Submarket RevPAR	Δ Submarket RevPAR	Δ Submarket RevPAR
Constant	.273***	.197***	.266***
Δ Occ	.725***	.347	.728***
Δ ADR	.216**	1.224***	.173**
N	2503	100	2403
R 2	.380	.304	.377
F value	696***	109***	656***
Wald chi-square: Δ Occ			3.12*
Wald chi-square: Δ ADR			68.78***
Ave VIF	1.0	1.39	1.0

Note. Dependent variable in Model 1 is the percent change in submarket Occupancy (Panel A) and Submarket RevPAR (Panel B) index, respectively. Model 2 is a subsample of hotels whose percent change in ADR is greater than zero. Model 3 is a subsample of hotels whose percent change in ADR is less than or equal to zero. Percent change is measured from 2019 to 2020. The Wald chi-square test is used to assess the equality of the same KPI coefficients between regressions (Models 2 and 3).

^*, **, *** significance at 10%, 5%, and 1% levels, respectively, based on heteroscedasticity-robust standard errors.

In Panel B, the results for occupancy and ADR reveal a significant and positive effect on the market RevPAR index. A 1% increase in occupancy would improve the market RevPAR index by 0.73%. Similarly, a 1% increase in ADR would raise market RevPAR penetration by 0.22%. In Panel B, the results in Model 2 show no significant impact of occupancy on the market RevPAR index for hotels that raised rates. More importantly, we observe a significantly greater positive impact of ADR on RevPAR index for hotels that raised rates when compared to Model 3 results for hotels that reduced rates. Thus, hotels that raised rates during the COVID-19 pandemic significantly improved their market RevPAR position more than hotels that discounted rates. A comparison of the ADR coefficients between Models 2 and 3 shows a statistically significant difference (χ² = 68.78, p < .001). We conclude that hotels that raised rates improved their occupancy index and their overall RevPAR performance relative to the competitive set.

Sensitivity Analysis

To test the sensitivity of our results, we divided the sample into quintiles (five groups) based on changes in ADR. We then tracked and compared mean performance changes relative to the competitive set for both 2020 and 2021. The results of this analysis are presented in Table 8. These results provide some interesting insights into the impact of COVID-19 on hotel revenue performance in 2020. Hotels in the bottom 20% had a mean decline of 32% in ADR and experienced a 38% decline in occupancy. The net impact of these two metrics is a 58% decline in RevPAR performance. Although these hotels improved their occupancy index by 6%, their ADR index declined by 14% on average. Consequently, the overall market RevPAR penetration declined 10%, driven by reduced rates. In contrast, hotels in the top 20% experienced a decline in occupancy of 26% that drove all of the RevPAR performance despite a slight increase in ADR. More important, these hotels improved their ADR index by an average of 13%. The net impact of maintaining rates consequently led to a 10% enhancement to the market RevPAR penetration. For 2021, all groups of hotels showed increases in all three lodging fundamentals, with the bottom quintile showing the largest increase in RevPAR performance. However, performance relative to the competitive set differed across groups. The bottom 20% did not improve its occupancy penetration but increased pricing position by 12%, which led to a net increase in RevPAR penetration of 9%. Clearly, this group reversed course from 2020 and raised rates to a greater extent than the other groups in 2021, which contributed to its improved RevPAR penetration. In contrast, the top 20% improved its occupancy penetration by 4% but saw its pricing position decline by 5%, which led to a 6% decline in RevPAR penetration.

Table 8.

Pricing Effect on Hotel and Submarket Performance, 2020–2021.

Panel A: ADR Quintiles, 2020	Δ Occ	Δ ADR	Δ RevPAR	Δ Submarket Occ.	Δ Submarket ADR	Δ Submarket RevPAR
1 Bottom 20%	–.381	–.322	–.582	.062	–.138	–.097
2	–.368	–.196	–.492	.009	–.040	–.033
3	–.356	–.141	–.448	–.008	–.002	–.012
4	–.301	–.089	–.362	.018	.025	.044
5 Top 20%	–.261	.040	–.247	–.003	.129	.101
Panel B: ADR Quintiles, 2021	Δ Occ	Δ ADR	Δ RevPAR	Δ Submarket Occ.	Δ Submarket ADR	Δ Submarket RevPAR
1 Bottom 20%	.407	.280	.749	.000	.123	.093
2	.402	.133	.572	.011	–.008	–.006
3	.406	.106	.544	.018	–.027	–.016
4	.348	.097	.473	.001	–.031	–.030
5 Top 20%	.377	.071	.399	.038	–.051	–.061

Note. Total sample is divided into five quintiles ranging from top 20% of sample to bottom 20% of sample data based on the percent change in ADR in 2020 and the effect assessed on changes in metrics and penetration indices. In Panel B, the pricing effect in 2021 is tracked and observed for the same groups of hotels in Panel A.

As an additional test, we divided the sample into eight groups based on changes in ADR. Dummy variables designate each of the eight groups and were used in place of a single continuous ADR measure. These dummy variables along with changes in occupancy were regressed on the changes in submarket market occupancy and RevPAR indices for 2020 and 2021 for the same group of hotels. The results of this analysis are provided in Table 9 Panels A and B. The results in Panel A show hotels that discounted rates by 5% or more significantly improved their occupancy penetration relative to hotels that raised rates more than 5% (omitted group). For example, hotels that dropped rates more than 25% significantly improved their occupancy index by 20% relative to hotels that raised rates by more than 5%, which was very different from the impact of rate changes on the RevPAR index. The results indicate hotels that dropped rates to a greater extent experienced a significantly worse market RevPAR position relative to those hotels that raised rates. For example, the RevPAR index declined 25% for those hotels that dropped rates more than 25% relative to hotels that raised rates more than 5%. A Wald chi-square test of the equality of the ADR coefficients of the two groups (omitted group vs. Δ ADR < –.25) is statistically different at the 5% level for both submarket occupancy and submarket RevPAR position.

Table 9.

Sensitivity Analysis of Pricing Changes on Submarket Performance.

Panel A: Y2020	Δ Submarket Occ	Δ Submarket RevPAR
Constant	.198***	.420***
Δ Occ	.850***	.707***
0 ≤ Δ ADR < .05	.006	–.164***
–.05 ≤ Δ ADR < 0	.030	–.171***
–.10 ≤ Δ ADR < –.05	.070***	–.159***
–.15 ≤ Δ ADR < –.10	.089***	–.176***
–.20 ≤ Δ ADR < –.15	.112***	–.189***
–.25 ≤ Δ ADR < –.20	.138***	–.203***
Δ ADR < −.25	.201***	–.251***
N	2503	2503
R 2	.530	.384
F value	290***	186***
Wald chi2: Δ ADR	89.2***	40.2***
Ave VIF	3.51	3.51
Panel B: Y2021	Δ Submarket Occ	Δ Submarket RevPAR
Constant	–.188***	–.276***
Δ Occ	.657***	.310***
0 ≤ Δ ADR < .05	–.032	.127***
–.05 ≤ Δ ADR < 0	–.025	.119***
–.10 ≤ Δ ADR < –.05	–.031*	.137***
–.15 ≤ Δ ADR < –.10	–.055***	.132***
–.20 ≤ Δ ADR < –.15	–.061***	.141***
–.25 ≤ Δ ADR < –.20	–.073***	.162***
Δ ADR < –.25	–.080***	.254***
N	2503	2503
R 2	.903	.255
F value	49***	21***
Wald chi-square: Δ ADR	20.6***	37.8***
Ave VIF	3.53	3.53

Note. Dependent variable is the percent change in Submarket Occupancy and RevPAR index. Percentage change is measured from 2019 to 2020. The Wald chi-square test is used to assess the equality of the ADR coefficients between two groups of hotels: ADR changes more than 5% (omitted group) vs. ADR changes less than −25%. In Panel B, the pricing effect in 2021 is tracked and observed for the same groups of hotels in Panel A. ***, **, and * significance at 1%, 5%, and 10% levels, respectively, based on heteroscedasticity-robust standard errors (HAC).

The findings in Table 9 Panel A suggest that occupancy may improve with discounting rates; however, the decline in market RevPAR position is far worse for greater levels of rate reduction. Thus, greater changes in pricing led to greater declines in market revenue position in 2020. In Table 9 Panel B, we track the hotel performance in 2021 for the same group of hotels identified in Panel A. The results show the submarket occupancy position worsening significantly for those hotels that discounted rates in 2020 by more than 5%. Contrastingly, submarket RevPAR position improved significantly for all groups of hotels, driven by higher rates in 2021. Specifically, hotels that discounted rates to a greater extent in 2020, raised rates in 2021 significantly more than hotels that offered lower discounts. For example, hotels that discounted rates more than 25% in 2020 lost submarket occupancy position by –8% in 2021 while their RevPAR position improved significantly by 25%, driven by higher rates. The Wald chi-square test of coefficients between the two groups (omitted group vs. Δ ADR < –.25) is also statistically different at the 5% level for both submarket occupancy and submarket RevPAR position.

To provide further evidence of the pricing effect, we employed an alternative approach used in prior research (Enz et al., 2004, 2009) by grouping hotels into 14 categories based on their pricing position relative to the competitive set. The hotels’ occupancy and RevPAR position relative to the competitive set were then assessed. The results in Figure 1 show the average occupancy and RevPAR position of the hotels relative to the competitive set at various levels of ADR position for the 2020 COVID-19 year.

Figure 1.

Impact of Pricing Changes on Occupancy and RevPAR Position in 2020.

Although these findings reveal occupancy position rose for hotels that reduced rates, their market RevPAR position fell relative to the competitive set. For example, hotels with rates between 25% and 30% lower than their competitive sets achieved average occupancies that were 20% higher than their competitive sets. However, their RevPAR position was 13% below the competitive set, consistent with the observation that discounting did not prove fruitful for this group of hotels during the COVID-19 pandemic. Dropping rates to increase occupancy did not pay off in terms of RevPAR performance. The revenue gains from a change in quantity demanded (occupancy) were simply insufficient to offset the revenue loss from reducing rates, resulting in lower RevPAR performance relative to the competitive set, which corroborates our empirical findings of an inelastic demand for the lodging sector.

In contrast and to our surprise, hotels with rate changes greater than zero improved their occupancy position, which led to an increase in their RevPAR position. For example, hotels with rate increases between 20% and 25% achieved average occupancies that were 10% above the competitive set, which improved their market RevPAR position to 34% above the competitive set. Thus, hotels that raised rates achieved higher occupancy and RevPAR gains relative to their competitive sets. In summary, we observe two contrasting effects of pricing outcomes due to COVID-19. On one hand, the percent difference in ADR position for hotels that reduced rates is substantially lower than the relative improvement in occupancy, resulting in a lower relative market RevPAR position. On the other hand, the percent difference in ADR position for hotels that raised rates is substantially greater than the percent difference in relative occupancy, which contributed to a substantial improvement in RevPAR performance relative to the competitive set.

Finally, we perform the regression analysis in Table 6 across hotel attributes to determine whether there is variation in the effect of price position on RevPAR performance. We find cross-sectional variation in the impact of price position on RevPAR performance across various hotel attributes. The price position (relative to the competition) coefficient is significant and positive at the 5% level across brand, hotel type, location, and scale, except for the urban location, midscale, and upper midscale hotels. These results still suggest that pricing higher than the competitive set will yield higher revenues. In addition, the coefficient of the COVID-19 dummy variable is negative and significant across all hotel attributes, implying a negative effect of the pandemic on RevPAR performance. Our comprehensive analysis of the pricing impact on RevPAR performance during the COVID-19 pandemic attests to the robustness of the results presented herein and provides strong empirical support for our conclusions.

Conclusions

This study investigated the price elasticity of demand and linked it to revenue performance in a sample of 2,503 hotels from 2018 to 2021. More importantly, this study highlights the impact of the COVID-19 pandemic on hotel RevPAR performance. In the first stage of our analysis, the results show lodging demand to be relatively inelastic at –0.17, demonstrating that dropping rates in the midst of an external shock will not increase revenue performance. Applying the elasticity estimate suggests that a 10% discount would have resulted in a net loss of US$419,075. A revenue performance analysis of two-subsample hotels during the COVID-19 pandemic in 2020 reveals that those hotels that raised rates in 2020 increased revenues by an average of US$91,438. In contrast, hotels that reduced rates incurred an average net loss of US$2.511 million in revenue performance in 2020 driven by negative changes in both rooms sold and discounting. These results suggest that reducing rates in the midst of a pandemic was not a wise decision. Consequently, the lesson here, given inelastic demand, is for hotel owners to raise rates in the midst of an external shock and, at the very least, avoid discounting.

The second stage of analysis revealed significant effects of a hotel’s competitive market position on RevPAR performance. The significant positive relationship between a hotel’s submarket ADR position and RevPAR performance suggests that raising rates will enhance revenue performance. Moreover, the results indicate a significant negative impact of the COVID-19 pandemic on RevPAR performance. An assessment of the pricing impact on submarket performance revealed no significant impact on occupancy penetration for hotels that raised rates. This finding suggests that hotels would have been better off raising rates during the pandemic. More important, the results show a significantly greater positive impact of ADR on RevPAR penetration for hotels that raised rates relative to hotels that dropped rates. Thus, hotels that raised rates during the COVID-19 pandemic significantly improved their submarket RevPAR position more than hotels that lowered rates. Additional sensitivity analyses provide corroborating evidence to support these empirical results.

Our comprehensive analysis linking price elasticity with revenue performance and the examination of pricing effect revenue performance provides strong and robust evidence of the negative consequences of discounting during the COVID-19 pandemic. Based on our assessment, we conclude that discounting did not work and therefore, was a losing strategy for those hotels that discounted rates during the COVID-19 pandemic in 2020. Although these results are consistent with prior research by Enz et al. (2004, 2009) that used annual performance data, they differ from O’Neill and Yeon (2022) who used monthly data to show that luxury and upper upscale hotels that discounted rates generate higher revenue performance. Instead, we find that luxury and upper upscale hotels were the worst revenue performers during the COVID-19 pandemic.

Practical Implications

The results indicate that economy extended stay hotels were top performers during the COVID-19 crisis in 2020. Demand in these hotels was fueled by leisure guests fleeing the restrictions and lockdowns in the major urban areas to “drive-to” resorts. In addition, these hotels were open and benefited from long-term stays of health care workers, first responders, and displaced workers given their apartment-style home accommodations with hotel services that are suited to the needs of long-term guests as well as for those moving out of apartments and seeking affordable alternatives.

Extended stay hotels have drawn the interest of owners, developers, and investors because of their lower variability and higher occupancy levels from longer stays, low cost and faster development, efficiency in operation from the lower cost of labor and operating expenses and therefore, high profitability (Hart, 2020; Karmin, 2022). Extended stay hotels have emerged as top performers and proven to be a resilient hotel type amid the COVID-19 pandemic and one may argue, to be recession-proof. For example, our results show the average occupancy rate in the extended stay segment for 2020 was 62.3%, a premium of 17% points over all other hotel types, and specifically, 25% points over full-service hotels.

With lodging recovery well underway in 2023 driven by strong leisure demand, the extended stay segment is likely to expand and attract greater developer and investment interest from private equity groups and institutional buyers to capitalize upon the segment’s resilience, fueled by strong fundamentals and the favorable growth potential of this sector. The purchase of 110 extended-stay WoodSpring Suites by Blackstone/Starwood Capital Group in March 2022 reflects investor expectation and the confidence that this segment will outperform others (Heschmeyer, 2022; Karmin, 2022). Although prior research excluded the extended stay segment (Enz et al., 2004, 2009; Noone et al., 2013), given the rapid growth and involvement of all the major hotel brands in this space in recent years, we believe this hotel segment should not be excluded and, instead, be a focus of any empirical analysis.

In contrast to extended stay hotels, our findings indicate that full-service hotels suffered the steepest decline in performance during the pandemic. Clearly, business and convention travel, a major demand generator for these hotels, disappeared in the face of restrictions and lockdowns. Many of these full-service hotels, comprised of upper upscale and luxury hotels, located in major urban areas or metropolitan markets and airport locations suffered a severe blow during the pandemic. Despite the setbacks, these hotel segments are expected to bounce back fueled by momentum in occupancy and pricing power. Although leisure travel will continue to be a strong driver for these hotel segments in 2023, full recovery in business and convention travel for these hotel segments is not expected until 2024.

Our study results reveal a steep decline in average occupancy rates during the COVID-19 crisis. Average occupancy rates fell by more than twice the decline in ADR, which led to overall RevPAR falling 44%. With a freefall in occupancy, one has to wonder whether it was a reasonable strategy for a hotel operator to discount rates in the midst of the pandemic to boost occupancy. It appears that fateful decision was made by many hotel operators. Given an inelastic demand for the lodging sector, discounting may improve occupancy, but the increase in occupancy is insufficient to offset the decline in revenue from dropping rates with the overall consequence of lower RevPAR performance at the hotel and relative to the competitive set. The results of our study on the effects of price elasticity and pricing on revenue performance are important, relevant, and informative for hotel operators in their competitive pricing decisions when faced with an unexpected external shock like COVID-19.

In light of our findings, hotel operators should resist the temptation to drop rates when the sector is hit by an external shock. Hotel operators often face pressure to follow the lead of direct competitors in reducing prices to maintain parity and avoid losing demand share (Enz et al., 2009). Instead, operators should strive to maintain or raise rates even when others are reacting to the shock by dropping rates. Our subsample analysis of hotels that raised rates during the pandemic is confirmation that pricing higher than competitors and maintaining that position will lead to stronger RevPAR performance. These findings show that raising rates has a net positive impact on the market RevPAR position because the revenue gains from raising rates more than offset the revenue losses from occupancy declines. This observation is consistent with an inelastic market where raising rates will increase RevPAR performance whereas discounting will decrease revenue performance (Canina & Carvell, 2005). Although occupancy penetration may improve with discounting, the decline in market RevPAR penetration will be far worse for greater rate reductions. This is so because revenue losses from reducing rates will be greater than the revenue gains from occupancy increases, which will worsen the RevPAR position relative to the competitive set. Therefore, based on our comprehensive analysis, the best strategy for a hotel to adopt in the midst of an external shock is for the hotel to maintain or raise rates to maximize room revenue rather than dropping prices relative to the competition, which is consistent with previous findings using annual performance data (Enz et al., 2004, 2009).

Limitations and Directions for Future Research

The COVID-19 pandemic did not begin to take hold until the end of March 2020. The annual data used for this study include the first quarter of 2020. To the extent that the first quarter operations are normal, the estimates in this study could be overstated. Moreover, there is usually a lag between pricing changes and the impact on demand so the annual performance data used in this study may not adequately capture the impact. As such, the results of our study using annual performance data will differ from prior studies that used monthly performance data to arrive at a different conclusion. A more accurate assessment for future research would be to use monthly data with a longer time span that excludes the first quarter of 2020. This study has focused only on the top-line operating fundamentals and ignored the impact on profitability, which is beyond its scope. Future research could extend this study by investigating the impact of COVID-19 on hotel profitability, particularly on how hotels managed operating expenses in response to the crisis. It should also be noted that the sample data is comprised of hotels securitized by CMBS, which tend to be higher quality hotels. Therefore, the findings may not be applicable to all hotels. In some cases, the empirical analysis is based on a relatively small subsample. A larger subsample would provide greater precision and reliability of the estimates. Finally, future research would benefit from additional studies on the impact of COVID-19 to validate the findings of this study.

Author Biographies

Amrik Singh is an Associate Professor in the Fritz Knoebel School of Hospitality Management in the Daniels College of Business at the University of Denver where he teaches courses in advanced revenue management, cost management, lodging valuation and hotel development and feasibility studies. His current research interests focus on commercial mortgage-backed securities (CMBS), financial performance and financial distress in the lodging industry.

Dr. David L. Corsun is an associate professor in and director of the Fritz Knoebel School of Hospitality Management in the Daniels College of Business at the University of Denver. Dr. Corsun’s current research is focused on several of the industry’s challenges, including the shared economy, pricing, and successful leadership development for multi-unit managers in lodging.