BUS 101-SLP-3

All documents are reproduced with the permission of the copyright owner. Further reproduction or distribution is prohibited without permission. Please do not reply directly to this email.

annette tyler sent you the following:

Email 1 of 1

Table of contents

1. US hotel industry revenue: an ARDL bounds testing approach

Bibliography

Document 1 of 1

US hotel industry revenue: an ARDL bounds testing approach

Author:  Chen, Han1; Chen, Rui2; Shaniel Bernard3; Rahman, Imran31 Lester E. Kabacoff School of Hotel, Restaurant and Tourism Administration, University of New Orleans, New Orleans, Louisiana, USA2 Department of Agricultural Economics and Rural Sociology, Auburn University, Auburn, USA3 Department of Nutrition, Dietetics and Hospitality Management, Auburn University, Auburn, USA

Publication info:  International Journal of Contemporary Hospitality Management ; Bradford  Vol. 31, Iss. 4,  (2019): 1720-1743.

ProQuest document link

Abstract: 

Purpose

This study aims to develop a parsimonious model to estimate US aggregate hotel industry revenue using domestic trips, consumer confidence index, international inbound trips, personal consumption expenditure and number of hotel rooms as predictor variables. Additionally, the study applied the model in six sub-segments of the hotel industry – luxury, upper upscale, upscale, upper midscale, midscale and economy.

Design/methodology/approach

Using monthly aggregate data from the past 22 years, the study adopted the auto-regressive distribute lags (ARDL) approach in developing the estimation model. Unit root analysis and cointegration test were further utilized. The model showed significant utility in accurately estimating aggregate hotel industry and sub-segment revenue.

Findings

All predictor variables except number of rooms showed significant positive influences on aggregate hotel industry revenue. Substantial variations were noted regarding estimating sub-segment revenue. Consumer confidence index positively affected all sub-segment revenues, except for upper upscale hotels. Inbound trips by international tourists and personal consumption expenditure positively influenced revenue for all sub-segments but economy hotels. Domestic trips by US residents added significant explanatory power to only upper upscale, upscale and economy hotel revenue. Number of hotel rooms only had significant negative effect on luxury and upper upscale hotel sub-segment revenues.

Practical implications

Hotel operators can make marketing and operating decisions regarding pricing, inventory allocation and strategic management based on the revenue estimation models specific to their segments.

Originality/value

It is the first study that adopted the ARDL bound approach and analyzed the predictive capacity of macroeconomic variables on aggregate hotel industry and sub-segment revenue.

Links:  Full Text

Full text: 

1. Introduction

The tourism industry has positioned itself as one of the nation’s largest employers representing 8.0 per cent of US GDP (World Travel and Tourism Council (WTTC), 2015). The lodging industry, being one of its major segments, contributed almost $600bn to the US GDP and created more than eight million jobs in different hotel segments (American Hotel and Lodging Association (AH&LA) American Hotel and Lodging Association, 2017). In terms of total output, lodging represents the largest segment in the broader tourism industry, with travelers spending more than $393bn per year on accommodations (Select USA, 2016). The lodging industry has outperformed the US economy over the past five years with industry revenue growing at an annual rate of 3.7 per cent reaching $189.5bn in 2015 (AH&LA, 2015). This rapid growth is driven by an escalating tourism demand, stemming from an influx of domestic and international leisure and business travelers. In addition, the growing demand of hotel service is influenced by the broad economy, where changes in consumer confidence and consumption expenditures can affect decisions on entertainment, travel and lodging (Alvarez, 2015). Moreover, the growth of hotel industry depends on the interplay between supply and demand (HOSPA, 2013).

There is no doubt that the lodging industry plays an important role in the US economy. However, it is not easy to plan strategically based on the macroeconomic environment. In essence, a three-step process needs to be followed. The first step is to identify the macroeconomic predictors or the forces of change that can significantly influence industry revenue. The second step is to see whether these macroeconomic predictors can accurately estimate industry revenue using a parsimonious estimation model. The last step is to test the utility of this model in various industry sub-segments and explain any ensuing variations.

Hoteliers need to accurately estimate hotel sales using key external drivers at both micro and macro levels to carry out strategic planning and management. At the micro level, sales estimation is a key tool for managers’ decision-making involving functional areas such as marketing, sales, finance, and accounting (Mentzer and Bienstock, 1998). In the hotel industry, revenue estimation is considered as an indispensable part of hotels’ marketing and operations especially as it relates to pricing and inventory management (Talluri and Van Ryzin, 2004). At the macro level, estimating expected growth in aggregate hotel revenue based on specific economic drivers helps hotel corporations to make large-scale investment related decisions on segment-based expansion, mergers and acquisitions. However, there is very little research that delves into hotel revenue estimation models at the macro level (Anderson et al., 2000; Bojanic, 1996). Furthermore, none of the relevant research studies employed the autoregressive distributive lag (ARDL) bounds testing procedure, which in the past had been popularly applied to examine the impact of macroeconomic factors on economic growth and tourism demand (Narayan, 2004; Srinivasan et al., 2012). Additionally, to the best of our knowledge, there is no single study that undertakes estimation of sub-segment hotel sales using relevant macroeconomic indicators. It is essential to understand the different ways in which macroeconomic predictors affect sales for each sub-segment so that practitioners can carry out more segment-specific strategic planning.

The purpose of this study, then, is to develop a predictive model to estimate monthly hotel industry revenue using five macroeconomic variables – domestic trips by US residents, consumer confidence index (CCI), inbound trips by non-US residents, personal consumption expenditure (PCE), and number of hotel rooms (NOHR). In addition, the model will be tested across industry sub-segments, which include luxury, upper upscale, upscale, upper midscale, midscale, and economy hotels. The study employs the five most relevant and accessible macroeconomic predictors to develop a working estimation model that is simple, user-friendly, and efficient. The goal, therefore, is to develop a parsimonious model with maximum explanatory power. Theoretically, the study is expected to fill the gap in current literature by proposing a hotel industry and sub-segment revenue estimation model using macroeconomic predictors and the ARDL approach. The estimation model and comparisons across sub-segments will help various stakeholders of the hotel industry such as operators, suppliers, investors, policy-makers, researchers, and various independent organizations better understand the influential factors in each sub-segment.

2. Literature review

2.1 Estimation of hotel industry revenue

Estimating hotel industry revenue can address external factors that influence industry revenue at the micro and macro level. At the micro level, it is typical to include arrival or booking date, segment price and duration of use, whereas at the macro level total demand may influence revenue to a great extent (Lee, 1990). The lodging industry is sensitive to fluctuations in demand; however, forecasting demand is pertinent due to the nature of the industry and its operational characteristics (Yüksel, 2007). Moreover, since estimation differs for transient and group customers (Yüksel, 2007), it is natural for it to differ across hotel sub-segments. Examining hotel segment performances in relation to economic factors is relevant since the results often differ when compared to overall industry performance (Canter and Maher, 1998). Making a distinction promotes a better understanding of how the varying demands across sub-segments relate to economic measures.

Extant literature regarding hotel revenue estimation is still in its infancy. Little prior research has estimated hotel revenue at the macro level, with most estimation models directed at the micro level that utilized techniques such as choice models (Talluri and Van Ryzin, 2004). Other studies applied several revenue management techniques to increase operational efficiency in the hotel sector (Solnet et al., 2016). There are also a handful of studies that estimated hotel demand in terms of guest arrivals and occupancy. For instance, Damonte et al. (1998) estimated hotel demand using a cross sectional sample of 310 properties. The factors in their analysis were average daily rate (ADR), number of rooms available, number of employees, food and beverage revenue and number of tourists attending conferences. The results indicated that price elasticity of demand varied across hotel segments. Consequently, Canina and Carvell (2005) expanded on this study to include consumer confidence index (CCI), income, expectations of income (corporate income and disposable personal income), ADR and local market’s ADR to estimate demand for urban hotels in a metropolitan market. The results, utilizing data spanning from 1989 to 2000, found all predictors as significant influencers of lodging demand.

2.2 Hotel segmentation

Hotels are categorized based on several factors including but not limited to hotel size, location, target markets, levels of service, number of rooms etc. Prior hotel segmentation studies in an econometric context have predominantly used micro-economic variables in their estimation of pricing, growth, and consumer demand. For example, Falk and Hagsten (2015) used two stage least absolute deviation estimators to predict growth and revenue for Swedish hotel establishments. They found that growth rate of overnight stays was significantly and positively related to the price segment of the hotel at the beginning of the same period; however, the relationship became negative as the price increased indicating that high-end hotels do not have better growth prospects than hotels in medium price segments. Similarly, Damonte et al. (1998) indicated that price elasticity of demand varied across different hotel segments when estimating hotel demand using ADR, number of rooms available, and number of tourists.

Luxury hotels were perceived by travelers as experiences, rather than products (Chu, 2014). Access to more disposable income may increase the frequency of consumer’s stay in such hotels. For example, Graf’s (2011) study on 2,824 hotel properties found that consumers who usually stay in lower class hotel segments, may switch to higher ones once their income increases. Tran (2015) estimated the effects of economic factors on the demand for luxury hotel sub-segment in the USA between 1998 and 2013. Results of their study indicated that US residents would extend their length of stay in luxury hotels when their income rises. German, Chinese, Japanese and Korean visitors would stay in luxury hotels when their income increases even if the hotel price goes up.

To the best of our knowledge, there is a lack of research regarding macroeconomic variables’ influence on the aggregate US hotel industry and sub-segments revenue, wherein the model that works for the aggregate hotel industry revenue may not work for different sub-segments. The study analyzes six sub-segments of the hotel industry to test the accuracy of the estimation model. In order to divide the hotels in different segments, the nomenclature used by Smith Travel Research (2014) (STR) was followed. STR positions hotels in classes based on their historical ADR, not on subjective criteria such as features or amenities. Both chain and independent hotels use the same ADR categorization.

2.3 Identifying predictors

Previous research shows the impact of economic predictors on the lodging industry (Choi, 2003; Zarnowitz, 1992). Given the paucity of research in hotel industry revenue estimation in the extant hospitality literature, the current study selected variables suggested by Canina and Carvell (2005), Chen et al. (2007) and Alvarez (2015), which include domestic trips by US residents, consumer confidence index (CCI), inbound trips by non-US residents, personal consumption expenditure (PCE), and number of hotel rooms (NOHR) in the industry.

2.3.1 Domestic trips by US residents.

Domestic tourists are defined as individuals taking overnight trips or longer away from the place of their residence (International Union of Official Travel Organizations, 1974). Consistent with Alvarez (2015), the current study measures domestic trips of US residents by the number of domestic flights within the USA for both leisure and business travel. Domestic trips by US residents increased from 660.9 million in 2006 to 682.1 million in 2015 (Alvarez, 2015), exhibiting an increasing trend that might explain the variation of the US hotel industry revenue in the same period. Close to 80 per cent of domestic trips were taken by leisure travelers in 2016 (USA Travel Association, 2016), who are known to be price sensitive and less willing to pay for a higher room rate (Masiero et al., 2015).

Witt and Witt (1995) chose domestic and international trips to forecast tourism demand and both variables added significant explanatory power in their study. These findings underscore the importance of the number of domestic trips in predicting hotel industry revenue. Economic growth leads to the increase in number of air passengers and business activities – hence more domestic business and leisure travel (Chi and Baek, 2013). As the number of domestic trips increases, more people will be staying in hotels increasing hotel revenue. Thus, domestic trips by US residents can be a significant predictor of US hotel revenue.  H1.

The number of domestic trips made by US residents has a significant positive influence on aggregate hotel industry revenue; and revenues for all six sub-segments.

2.3.2 Consumer confidence index.

Consumer confidence refers to the degree of optimism consumers feel about the state of the economy and their personal finances, which guides their decisions on spending and saving. Consumer confidence is measured by two indices – The Conference Board’s Consumer Confidence Index (CCI) and University of Michigan’s Index of Consumer Sentiment (ICS). Vuchelen (2004) contended that consumer sentiment is an efficient variable to use for forecasters to avoid some errors, since many economic and financial variables can significantly influence the consumer sentiment. Since both CCI and ICS essentially measures consumer confidence but with a different methodology, it is safe to use either one in estimation models.

In this study, CCI is used to predict US hotel revenue. The US CCI is calculated by The Conference Board using a monthly survey. The monthly survey includes questions related to consumers’ household finances, employment, income, business conditions and economic outlook (The Conference Board, 2011). CCI has a good forecasting power for consumer spending as it influences individual expectations and consumption preferences. High consumer confidence can decrease uncertainty in the future, thereby reducing precautionary savings and increasing present consumption (Ludvigson, 2004). In addition, an increase in consumer confidence can boost future income and wealth expectations (Ludvigson, 2004). Thus, CCI can influence the real economy by increasing consumer expenditure on entertainment, travel, lodging, etc. Hence, higher CCI results in direct future consumption growth.

Prior research provides empirical evidence that CCI increases labor income (Carroll et al., 1994), which eventually translates into more expenditure. Singal (2012) also demonstrated that CCI can explain a significant part of variation in consumer expenditure on hotel industry. In hindsight, CCI can influence the real economy by increasing or decreasing consumer expenditure. Increased expenditure on lodging related services would increase hotel revenue. Thus, the following hypothesis is proposed:  H2.

CCI has a significant positive influence on aggregate hotel industry revenue; and revenues for all six sub-segments.

2.3.3 Inbound trips by non-US residents.

International tourists are defined as “tourists who stay at least one night in a country where they are not residents”, where a resident is “a person who has lived for most of the past year in a country” (Eilat and Einav, 2004, p. 1319). The study measures inbound trips of non-US residents by the number of international arrivals to the USA for both leisure and business purposes. The number of international visitors to the USA is found in three US and international government sources: the USA Department of Homeland Security/USA Customs and Border Protection I-94 arrivals program data, Statistics Canada’s International Travel Survey and Banco de Mexico travel data (National Travel and Tourism Office, 2016).

International visitors have significant influence on the US hotel industry (Tran, 2015). USA received the largest share of world tourism (14.2 per cent) in year 2014 (AH&LA, 2015). The international visitor arrivals and their lengths of stay can influence the demand for hotel rooms. Each international visitor stayed an average of 18 nights in the USA (USA Travel Association, 2015). Proceeds from international visitor arrivals in the USA accounts for as much as 20 per cent of US hotel revenue (AH&LA, 2015). The top five countries with the most international visitor arrivals are Canada, Mexico, UK, Japan and China (National Travel and Tourism Office, 2015). The variable international visitor arrivals has received overwhelming attention in the tourism literature especially in estimating tourism demand, with the vast majority of these studies testing the utility of various estimation models specific to a country/region (Kraipornsak, 2011; Peng et al., 2015; Yang et al., 2010).

Travelers’ demographic characteristics matter in international trips. Income, exchange rates, and transportation costs are the three critical factors affecting international tourism demand (Lim, 1997). Access to financial resources positively influence tourist demand and travel frequency (Davies and Mangan, 1992). The general assumption is that as peoples’ disposable income increases, so will their tendency to engage in leisure travel, extend their length of stay in a destination and spend more in travel related services (Alegre et al., 2011). In essence, wealthy families are known to engage in international travel, whereas families in lower income brackets tend to travel domestically (Fang Bao and McKercher, 2008). Previous studies show that most international tourists are considered as affluent travelers, indicating that income in their country of origin is the most important explanatory variable in generating those trips (Crouch, 1995; Lim, 1997). Overseas travelers spend almost $4,400 during their visit to the USA (USA Travel Association, 2015). As more international tourists arrive in the USA, demand for hotel rooms and other lodging-related services goes up, contributing directly to hotel sales. Therefore, a significant positive relationship between inbound trips by non-US residents and lodging industry and its sub-segment revenue is expected.  H3.

Inbound trips by non-US residents have a significant positive influence on aggregate hotel industry revenue; and revenues for all six sub-segments.

2.2.4 Personal consumption expenditure.

Personal consumption expenditures (PCE), “measures the goods and services purchased by persons – that is, by households and by nonprofit institutions serving households which are residents in the USA” (Bureau of Economic Analysis, 2014, p. 5-2). Consumer spending on goods and services in the US economy is measured primarily by the PCE, which accounts for approximately two-thirds of domestic spending (Bureau of Economic Analysis, 2014). PCE is known as the primary driver of future economic growth. It implies how much of the household income is spent on current consumption as opposed to being saved for future consumption.

Households engaging in more consumption outside (e.g. dining out and shopping) will “directly lead to more activities and travel consistent with the behavioral paradigm that travel demand is a derived demand” (Ferdous et al., 2010, p.1). Therefore, PCE is a good indicator of US consumer spending, which can be used to estimate tourism demand (Chen et al., 2007). As discussed above, the more the tourism demand, the higher is the expenditure on lodging and related service. In a hotel context, as perceived value of the product or service increases, consumers’ intention to purchase grows (Ashton et al., 2010), which contributes to hotel revenue growth. Similarly, Corgel et al. (2012) suggested that the increase of income generates higher demand for higher priced hotel segments such as luxury and upper upscale hotels than for lower priced ones such as midscale and economy hotels. Hence, a significant positive relationship between PCE and US hotel industry and sub-segment revenue is expected.  H4.

PCE has a significant positive influence on aggregate hotel industry revenue; and revenues for all six sub-segments.

2.2.5 Number of hotel rooms.

Number of hotel rooms (NOHR) variable measures the number of available hotel rooms in the aggregate hotel industry. Similarly, NOHR in each sub-segment is used in the estimation of its corresponding hotel sub-segment revenue. The increases of available hotel rooms in the industry will lead to more supply in the industry, increasing competition, and resulting in lower ADR for some sub-segments. Hotel revenue is the result of both ADR and number of rooms sold, which is the demand from the consumers’ side. The growth of hotel revenue depends on the price elasticity of demand (Canina and Carvell, 2005). Price elasticity is defined as the per cent change in demand divided by the per cent change in price, which measures the degree to which demand is sensitive to changes in price (Corgel et al., 2012; Trans, 2011). If the demand is price elastic, hotel revenue may increase if room rates reduce. On the other hand, if the demand is price inelastic, hotel revenue will reduce as room rates fall (Canina and Carvell, 2005). Prior studies demonstrated that all US hotel sub-segments had inelastic demand, which showed that the growth in room rate is much greater than the growth in demand (Hiemstra and Ismail, 1993). Tran (2011, 2015) supported this view by demonstrating that the price is inelastic in US luxury hotel sub-segment and suggesting that consumers are not sensitive to price changes. Similarly, research at the property level also showed that price elasticity of demand varied across hotel sub-segments with higher priced hotels having lower price elasticity than lower priced hotels (Canina and Carvell, 2005). Hence, the increase of hotel supply will result in reduced ADR, which further results in increased demand, leading to a reduced level of hotel industry revenue (price inelasticity).

Therefore, the study proposes the following hypotheses:  H5.

NOHR has a significant negative influence on aggregate hotel industry revenue; NOHR for each sub-segment has a significant negative influence on sub-segment revenues.

3. Methodology

Monthly data from January 1996 to September 2017 were collected from a variety of sources. Monthly aggregate hotel industry and sub-segment revenue data were collected from STR. Revenue data for chain hotels and independent hotels as well as number of hotel rooms in each sub-segment were included in this dataset. These data were not publicly available and required subscription to STR. Domestic trips by US residents’ data were gathered from the Bureau of Transportation Statistics (2018) website. Consumer confidence index (CCI) data were compiled from The Conference Board (2018) website. Inbound trips by non-US residents’ data were collected from the National Travel and Tourism Office (2018). Seasonally adjusted personal consumption expenditure (PCE) data were collected from the Federal Reserve Economic Data (2018) (FRED). All data sets apart from hotel industry revenue and number of hotel rooms were publicly available. In addition, all data sets were not seasonally adjusted except PCE.

Seasonality is a known characteristic of tourist demand, which also influences accuracy of hotel revenue estimation (Chen et al., 2015). Therefore, it cannot be overlooked in the modeling process. Seasonal adjustments of all variables except PCE were conducted using the X-12-ARIMA program, which is developed by the US Census Bureau. The X-12-ARIMA procedure makes adjustment for monthly or quarterly series. It is the primary method for seasonal adjustment of government and economic time series in USA, Canada and the EU (Miller and Williams, 2004).

3.1 Model specification and estimation procedure

The current study examines the five macroeconomic predictors’ utility in estimating aggregate US hotel industry revenue. It also assesses how the causal model works for potentially dissimilar industry sub-segments. As seen in Table I, the sub-segment classification followed the nomenclature used by STR. The auto-regressive distribute lags (ARDL) approach is used via EViews 10. Unit root analysis is first conducted to confirm that the data are stationary, i.e. I (0), or non-stationary, i.e. I (1). Then the cointegration test with the ARDL approach is utilized.

The US hotel revenue is determined by variables that measure the hotel supply – NOHR and hotel demand – namely, domestic trips by US residents, CCI, inbound trips by non-US residents and PCE. US nominal hotel renvenue and PCE have been deflated by consumer price index to remove the effects of inflation. Therefore, the determinats of US hotel revenue take the specification form: (1) ln⁡Revenuet=α+β ln⁡(⁡DTt)+γ ln⁡(⁡CCIt)+δln⁡(⁡IAt)+θ ln⁡(PCEt)+ψ ln⁡(NOHRt)+εtwhere Revenuet denotes the aggregate US hotel industry revenue and sub-segment revenues for luxury, upper upscale, upscale, upper midscale, midscale and economy in year t. DTt and IAt are defined as domestic trips by US residents and inbound trips by non-US residents in year t.

Many empirical studies have proved that most aggregate data have a unit root (Kwiatkowski et al., 1992). Variables that contain a unit root were generated by a non-stationary data process of taking the first difference of the variables, resulting in spurious regression. When working with time series datasets, it is important to look for a unit root. If a unit root is found in a series, it means that more than one trend is present in the series. Since the data are at the macroeconomic level and time series, whether all variables are stationary or not are detected to avoid spurious regression using stationary test. There are different ways to test whether data is stationary or not, such as Lagrange multiplier (LM test) (Kwiatkowski et al., 1992), Augmented Dickey-Fuller (ADF) test (Becketti, 2013; Dickey and Fuller, 1979; Hamilton, 1994) and PP test (Phillips and Perron, 1988). This paper employed ADF and PP tests to determine a time series is stationatary or not.

However, the regression specification of differencing could only provide the short-run estimates not the long-run estimates. ARDLs are standard least squares regressions which include lags of both the dependent variable and independent variables as regressors (Greene, 2008). Although ARDL models (the bound test approach) have been used in econometrics for decades, they have gained popularity in recent years as a method of examining long-run and cointegrating relationships between variables (Pesaran and Shin, 1998). Therefore, this problem could be solved by considering the cointegration and the error correction model (ECM) and obtaining both short- and long-run information (Nkoro and Uko, 2016). Additionally, ARDL models could yield consistent estimates of long run relationship, irrespective of whether the regressors are purely I(0) or I(1) or a mixture of both.

To examine the long-run relationship between US hotel revenue and its determinats, the ARDL cointegration procedure is applied (Pesaran et al., 1996; Pesaran and Shin, 1998). The model with lower Akaike information criterion (AIC), Schwarz Bayesian criterion (SBC), Hannan-Quinn Criterion (HQ) and higher adjusted R2 performs better than the other models. Lagrange multiplier (LM) test was used to test the residual’s autocorrelation. The bounds procedure for testing the existence of a long-run/cointegration relationship is implemented regardless of I(0) or I(1) as a post estimation command (Pesaran et al., 2001). If the existence of a long-run relationship (cointegration) is affirmed, the second step is to construct the conditional ARDL specification for ln⁡(revenue). Following Pesaran et al. (2001), the conditional unrestricted equilibrium correction model for US hotel revenue can be specificed as: (2) Δln⁡Revenuet=α2,0+π1ln⁡(Revenue)t-1+πxXt-1+∑i=1p-1ψi′ΔECMt-i+w′ΔXt+εtwhere: Xt=(ln⁡(DTt),ln⁡CCIt,ln⁡IAt,ln⁡PCEt,ln⁡(NOHRt)′,ECMt=(ln⁡(Revenuet),Xt),ECMt is the speed of adjustement parameter, which is derived as error term from long-run model.

The standard long-run ARDL (p,p1,p2,p3,p4,p5) specification can be expressed as: (3) ln⁡(Revenuet)=α1,0+∑i=1pπ1,iln⁡Revenuet-i+∑i=0p1β1,iln⁡DTt-i+∑i=0p2γ1,iln⁡CCIt-i+∑i=0p3δ1,iln⁡(IAt-i)+∑i=0p4θ1,iln⁡PCEt-i+∑i=0p5ψ2,iΔln⁡(NOHRt-i)+ε1,twhere t = max⁡(p,p1,p2,p3,p4,p5,…,T);p,p1,p2,p3,p4andp5 are the number of optimal lag order, which will be obatined based on the minimization of AIC or Bayesian information criterion (BIC).

The short-run specification dynamics include one period lagged error correction version of the ARDL model, (4) Δln⁡(Revenuet)=α2,0+∑i=1pπ2,iΔln⁡Revenuet-i+∑i=0p1β2,iΔln⁡DTt-i+∑i=0p2γ2,iΔln⁡(CCIt-i)+∑i=0p3δ2,iΔln⁡(IAt-i)+∑i=0p4θ2,iln⁡ΔPCEt-i+∑i=0p5ψ2,iΔln⁡(NOHRt-i)+λECMt-1+ε2,twhere ECMt-1=ln⁡(Revenuet-1)-α+βln⁡(DTt-1)+γln⁡(CCIt-1)+δln(⁡IAt-1)+θln⁡(PCEt-1)+ψln⁡(NOHRt-1)is the ordinary least square (OLS) residual series from the long-run cointegrating regression, which means, α,β,γ,δ,θ and ψ can be OLS estimates from Equation (1). ECMt-1 demonstrates how much of the disequilibrium in the previous period (ln⁡Revenuet-1) is being adjusted in current period (ln⁡Revenuet). A significantly postive estimate of ECMt-1 indicates divergence, and a significantly negative coefficient means convergence. This ARDL model with its ECM can be regressed by the OLS approach.

Thus, the conditional error correction representation of ARDL model for US hotel revenue can be reparameterised as (Bahmani-Oskooee and Ng, 2002; Pesaran et al., 2001): (5) Δln⁡(Revenue)t =α0+∑i=1pπ2,iΔln⁡Revenuet-i+∑i=0p1β2,iΔln⁡DTt-i+ ∑i=0p2γ2,iΔln⁡(CCIt-i)+∑i=0p3δ2,iΔln⁡(IAt-i)+∑i=0p4θ2,iΔln⁡PCEt-i+ ∑i=0p5ψ2,iΔln⁡NOHRt-i+πln(Revenuet-1)+βln(⁡DTt-1)+ γln⁡(CCIt-1)+δln⁡(IAt-1)+θln⁡(PCEt-1)+ψln⁡(NOHRt-1)+εt

4. Results

4.1 Unit root test results

This study uses the logarithms of the data in order to eliminate the effects of potential heteroscedasticity. Before an appropriate ARDL model, the first step is to determine whether data are I(0) or I(1). Table II indicates all variables are I (1), except PCE, NOHR[1], midscale, upper upscale sub-segments revenue and aggerate industry revenues that are I(0). Therefore, there is a mixture of I(0) and I(1) for all variables, indicating the ARDL model can be proceeded further. Figure 1 implies there is no obvious trend and structural break of revenue data over periods. In addition, Table II shows that all variables are stationery with or without trend. However, variables are more stationary with trend relative to without trend. Therefore, time trend will be included in the estimation of the ARDL model.

4.2 Results from auto-regressive distribute lags bound tests for cointegration

To test the null hypothesis of cointegration, the first step is to determine whether there is a relationship between the variables over the long term using bound tests (Bahmani-Oskooee and Ng, 2002). All of the computed F statistics are much greater than the upper critical value for 1 per cent from Table III. Hence, the null hypothesis was rejected, confirming the existence of co-integration among US hotel revenue, domestic travel, CCI, international arrival, PCE, and NOHR in these seven models.

After validating the application of ARDL model, the next step is to determine the lag length for the dependent variable. To implement the choice of lag length, an unrestricted VAR model was applied for Δln⁡(Revenuet) and the constant,ln⁡(Revenuet-1), ln⁡(DTt-1), ln⁡(CCIt-1), ln⁡(IAt-1), ln⁡(PCEt-1), ln(NOHRt-1) and a fixed number of lags of Δindepdentvariables as exogenous regressors. According to the VAR lag order selection criteria (the lower value of AIC, SC, HQ etc., the better results), the optimal lag length for dependent variable is three or eight. The same procedure was applied to all six hotel sub-segments and the optimal lag length for each hotel sub-segment revenue can be found in Table IV.

4.3 Results of the long- and short-run effects

The augmented ARDL model with appropriate lag orders are obtained for the equation, all criteria cited above are used and after that the smallest lag length among them was taken. The regression results of ARDL model for revenue of US hotel industry and six different hotel sub-segments are reported in Table V. Until now, the study analyzed both the long- and short-run relationships among the US hotel revenue, domestic trip, CCI, international arrival, PCE and NOHR.

The results reveal that the estimated coefficient of domestic trips by US residents is significantly positive for the aggregate hotel industry revenue, upper upscale sub-segment revenue, upscale sub-segment revenue and economy sub-segment revenue only. It shows that in the long run, one per cent increase in the domestic trips leads to approximately 0.20 per cent increase in the aggregate US hotel industry revenue, 0.31 per cent increase in the upper upscale sub-segment revenue, 0.15 per cent increase in the upscale sub-segment revenue and 0.39 per cent increase in the economy sub-segment revenue. H1 was partially supported.

CCI positively and significantly influenced aggregate hotel industry revenue. The same relationship also exists throughout the hotel sub-segments except for the upper upscale sub-segment. The results imply that 1 per cent increase in CCI will lead to 0.18 per cent increase in aggregate hotel industry revenue and 0.05-0.26 per cent increase in revenue for different hotel sub-segments except the upper upscale sub-segment. This empirical evidence confirms that CCI has a positive impact on majority of US hotel sub-segment revenue in the long run. Hence, H2 was partially supported.

The inbound trips by non-US residents variable has a significant positive influence on aggregate hotel industry revenue. The same relationship also applies to all hotel sub-segments except for economy, partially supporting H3. As for PCE, it has a significant positive relationship with aggregate revenue and throughout all hotel sub-segments except for economy sub-segment, partially supporting H4. NOHR has a significant negative influence only on luxury and upper upscale hotel sub-segment revenues. However, this variable is not significantly associated with revenue from all other hotel sub-segments or the aggregate hotel industry revenue. H5 is partially supported.

The estimates of the short-run dynamics associated with the long-run relation from ECM are presented in Table VI. ECM is statistically negative for all seven models, implying there is adjustment from disequilibrium into long-run equilibrium, which helps reinforce the long-run relationship (co-integration) between hotel revenue and its determinants. More specifically, the estimate of ECM indicates 70.9 per cent of the disequilibrium in aggregate hotel revenue from previous period which will be converged back to the long-run equilibrium in the current period. The results also imply a range of 54.4-99.0 per cent of the disequilibrium in sub-segment revenue from previous period which will be converged back to the long-run equilibrium in the current period across the six hotel sub-segments.

Although domestic trips do not have a significant positive effect on luxury, upper midscale, and midscale sub-segment revenue in the long run, it does have a significant impact on the aggregate US hotel revenue growth rate and all six sub-segments revenue growth rate in the short term. The growth rate of CCI and PCE do not have a significant impact on either aggregate hotel industry revenue growth rate or any sub-segment hotel revenue growth rates in the short run. Inbound trip growth rate has significant positive short-term effects on the growth rate of aggregate hotel industry revenue and all six hotel sub-segments revenue in the short term.

As the ARDL model and its associated ECM have been estimated by the OLS, the assumptions of OLS−the normality, heteroscedasticity, and the serial correlation have been tested, which are reported in Table VI. From Table VI, DW (Durbin-Watson) values for all seven models are higher than the upper bound critical value (dU = 1.726), indicating that the study failed to reject the null hypothesis. Thus, there is no positive serial correlation/autocorrelation of residuals. There is no heteroscedasticity issue in the model from Breusch-Pagan Lagrange Multiplier (BPG LM) test either. In addition, data follow the normal distribution. It is important that the error of this model is serially independent. If not, the parameter estimates will not be consistent, because of the lagged values of dependent variable that appear as regressors in the model. All errors are serially independent.

Finally, the study tested the stability of the long-run and short-run estimates of the ARDL models. Following Bahmani-Oskooee and Ng (2002) and Pesaran and Shin (1998), the stability test – the cumulative sum (CUSUM) and the cumulative sum of squares (CUSUMSQ) – is undertaken to assess the parameter consistence based on the AIC from the ECM. According to Figure 2, both the plots of CUSUM and CUSUMSQ statistics stay within the critical bounds of 5 per cent significance level, which applies to aggregate model as well as all hotel sub-segments, showing that there is no instability for all sub-segment revenue estimation models.

5. Discussion

The findings showed the overwhelming support of CCI in estimating aggregate hotel and sub-segment revenues except for the upper upscale sub-segment. CCI’s significant positive support of US hotel revenue is consistent with Singal’s (2012) finding that CCI can explain a significant part of variation in consumer expenditure in the hotel industry. The more confidence consumers have in the US economy, the more they will spend on travel boosting hotel revenue. Low consumer confidence results in lower consumption as consumers often postpone their trips and/or reduce their frequency of travel (Ludvigson, 2004; Singal, 2012). The insignificant relationship between CCI and upper upscale segment revenue can be explained by the dichotomy properties of price: objective price and perceived price. Objective price is the actual price of a product, while perceived price appeals to the subjective internal impressions derived from the perception of price (Dodds et al., 1991). Prior research shows that consumers do not rely on objective price, but rather interpret perceived price in way that are meaningful to them (Zeithaml, 1988). Therefore, even though tangible price differences exist between these two classes, consumers may end up choosing to stay in luxury hotel instead of upper upscale due to the higher perceived value.

PCE positively affects aggregate hotel revenue and this relationship is consistent across all sub-segments except for the economy sub-segment in the long run. The more household income is being spent on current consumption, the more hotel revenues will grow. The result echoes with findings of previous research that PCE is a good indicator of the tourism demand which can generate hotel revenue (Chen et al., 2007). The findings are consistent with Corgel et al. (2012) that the increase of income generates more demand for upper priced hotel segments than for lower priced ones such as economy hotel sub-segment. As consumers have more spendable income, they would most likely choose better hotels to stay in, which somewhat explains the insignificant relationship of PCE with economy sub segment hotel revenue. The economy sub segment is popular among relatively low-income consumers who are in need of accommodations both for short and long term. This is a popular segment among blue-collar workers looking to relocate, temporarily move to other destinations, or travel for work. Many people also use this segment as an alternative to rent out an apartment. Revenue for this segment, therefore, does not fluctuate much based on changes in PCE as this segment is used by people who are in indispensable need of accommodation. This segment is also very popular among motorists looking for an overnight accommodation near a highway, who might opt for better hotels to stay in if they have more spendable income.

Besides CCI and PCE, domestic trips showed significant positive influence on aggregate US hotel industry revenue. The significant positive effect of domestic trips is supported by the fact that lodging industry is the largest segment of the tourism industry. Hotel industry revenue accounts for nearly 19 per cent of total travel and tourism related spending (Select USA, 2016). In addition, domestic travelers are actually spending more in recent years. In 2014, the typical business traveler spent about 3 per cent more per night, and the typical leisure traveler spent about 6 per cent more per night compared to the previous year (AH&LA, 2015), which contributes to the increase in US hotel revenue as a whole. Further analyses revealed that domestic trips by US residents only added significant explanatory power to aggregate, upper upscale, upscale and economy sub-segment revenue. It is also observed that domestic trips by US residents are significant in short term throughout all sub-segments. In 2016, close to eighty per cent of domestic trips were taken by leisure travelers (USA Travel Association, 2016). Leisure travelers are known to be price sensitive compared to business travelers whose travel expenses are covered by their companies most of the time (Kashyap and Bojanic, 2000; Noone and McGuire, 2013). For example, Masiero et al. (2015) found difference in willingness to pay (WTP) for certain hotel attributes among leisure and business travelers. They found that business travelers were willing to pay up to 25 per cent higher for certain hotel attributes than leisure travelers. In addition, Trejos (2018) found that business travelers prefer to stay in hotel brands such as Hilton, Hyatt, Embassy Suites, Courtyard, Doubletree, and Hilton Garden Inn etc., which mostly fall into either upper upscale or upscale categories. The upper upscale segment especially is known as the sweet spot for corporate travel (Business Travel News, 2013). These two sub-segments also registered the highest occupancy rates among the six sub-segments in recent years (Smith Travel Research, 2018). Therefore, the significant relationship between domestic trips and upper upscale and upscale hotel sub-segments revenue makes sense as most business travelers prefer to stay in these two sub-segments.

According to Yesawich et al. (2000) 60 per cent of leisure travelers were actively searching for the “lowest possible price” for travel-related products. The price sensitive leisure travelers tend to target lower cost lodging segments, which explains why domestic trips by US residents has significant positive effect on economy sub-segment revenue. About 39 per cent vacations taken by leisure travelers in 2017 were road trips (MMGY Global, 2018). Even when flying, many domestic travelers on leisure trips prefer to rent cars from the airport to reach their final destinations. Among motorists economy sub-segment hotels remain a very popular option because of the convenience of being located near major highways, which explains why economy sub segment revenue is influenced by domestic trips by US residents.

The inbound trips by non-US residents also had a significant effect on aggregate hotel revenue which can be explained by the fact that USA receives the largest share of international tourism receipts in the world (AH&LA, 2015). In 2016, 75.9 million international tourists visited the USA (Statista, 2018); while the industry-wide revenue from international inbound tourists was not readily available, AH&LA reported that in 2014, international travel contributed to twenty per cent of the US hotel revenue (AH&LA, 2015).

Furthermore, inbound trips by non-US residents had significant positive influence on revenues of all sub-segments except for economy sub-segment. This finding aligns with previous studies that international tourism is considered as luxury (Lim, 1997); non-US residents who can afford trips to the USA should be affluent in their home countries (Crouch, 1995). The top six tourist generating countries to the USA are Canada, Mexico, UK, Japan, China and Germany (National Travel and Tourism Office, 2015). These six countries accounted for 78.2 per cent of the total international tourist arrivals in the USA in 2015 (National Travel and Tourism Office, 2015). Travelers from most of the above-mentioned countries generally have great buying power due to their strong economy. For example, 1.8 million Chinese travelers visited USA in year 2014 and contributed $21.1bn to the US economy (Willett, 2015). Stats also show that overseas travelers spend around $4,360 during their visit to the USA (USA Travel Association, 2016), which is substantially more than what most domestic leisure and business travelers spend on a trip. However, among 74.8 million internationals that visited the USA in 2014, only 26.5 million stayed in a hotel (AH&LA, 2015). Thus, many international tourists especially the more budget conscious ones prefer to stay with their friends and families during their visits. In addition, the tremendous growth of Airbnb in recent years (from 1.5 million in year 2014 to 2.5 million rooms in year 2015) could be one of the reasons why international inbound trips do not have a positive effect on the low-end hotel sub-segment (AIRDNA, 2016). For example, Zervas et al.’s (2017) study on the effect of Airbnb on the revenues of hotels in Texas found that economy sub-segment hotel revenues were most vulnerable to increased competition from Airbnb rentals. Additionally, many young and mostly solo budget conscious international tourists, such as backpackers, often choose affordable shared accommodation options such as hostels and Airbnb’s over hotels.

The number of hotels rooms did not add a significant negative explanatory power to aggregate hotel industry revenue, which contradicts previous finding that the demand for the aggregate hotel industry is price inelastic (Canina and Carvell, 2005), and that the increase in the number of hotel rooms in the industry would result in reduced aggregate hotel revenue. The insignificant relationship suggests that US hotel industry has a relatively balanced supply and demand. The growth of hotel supply can be absorbed easily by the market demand. The 2018 pipeline data from STR reported that the number of hotel rooms under construction in the USA has declined or remained flat in recent months (HotelNewsNow, 2018). Furthermore, STR’s segmentation analysis in 2017 found that the demand is strong and is not equal to supply growth (Cushman and Wakefield, 2017). However, sub-segment analyses revealed that the number of luxury hotel rooms and the number of upper upscale hotel rooms both had significant negative impact on their respective sub-segment revenues. The result was consistent with previous research finding that demand is price inelastic for luxury hotel sub-segment and upper upscale sub-segment (Canina and Carvell, 2005; Hiemstra and Ismail, 1993; Tran, 2011; Tran, 2015). Increased hotel supply in luxury and upper upscale sub-segments resulted in higher level of market competition, leading to the lower ADR in both sub-segments. As demand is price inelastic in these two segments, the increase in demand is lower than the changes in price, resulting in reduced revenue for luxury and upper upscale hotel sub-segments. However, the relationship between number of sub-segment hotel rooms and sub-segment hotel revenues was insignificant for all other hotel sub-segments. This finding might also be explained by previous research finding that the price elasticity of demand varied across hotel sub-segments with lower class hotels having higher price elasticities than higher class hotels (Canina and Carvell, 2005). The changes in lower class hotel room rates caused by supply fluctuation will result in a similar level of change in demand from consumers’ side. As a result, the hotel revenue will not have significant fluctuation for these hotel sub-segments.

In summary, analyses on sub-segments indicated that the aggregate US hotel industry revenue model cannot be applied for all six hotel sub-segments. This is perhaps the most interesting finding of this study. Substantial variations exist in the number and type of predictors that can be used in different hotel sub-segments. The following table provides a summary of these segment specific predictors.

Table VII underscores an important point for practitioners, policymakers, researchers and other stakeholders. When it comes to estimating revenue, what works in one sub-segment might not work for another. As a result, stakeholders need to consider segment specific predictors when developing these models. Even the aggregate model might not be fully applicable to different sub-segments of the hotel industry, as evident from our findings.

6. Implications

To our knowledge, this is the first study that analyzed the predictive capacity of macroeconomic variables on aggregate hotel industry and sub-segment revenue. The ARDL bounds approach has not been applied in the past to estimate hotel industry revenue. Numerous differences were found in estimation models between sub-segments. A segment specific revenue forecasting strategy is therefore recommended. One approach fits all might not be the right strategy for tracking hotel industry revenue.

The findings provide implications for hotel industry practitioners, policy makers, industry associations, and researchers. Hotel operators, especially those with multiple properties can make marketing and operating decisions regarding pricing, inventory allocation, and strategic management based on the revenue estimation models specific to their segments. For upper midscale and midscale hotels, the operators should track consumers’ confidence, PCE and international inbound tourism for making important strategic decisions on pricing, mergers and acquisitions, and investing. On a macro level, the model can be used by practitioners to predict changes in hotel revenues based on estimations of the macroeconomic variables used in this study. For luxury and upper upscale hotel sub-segments specifically, the practitioners need to pay close attention to the number of hotel rooms in their sub-segment to adjust pricing strategies to obtain better revenue. In addition, upper upscale, upscale and economy hotel operators also need to take domestic travelers’ needs into consideration.

Practitioners, investors, policymakers and researchers, in hindsight, can track aggregate hotel industry revenue by keeping an eye on domestic trips, international inbound trips, CCI and PCE. In summary, a segment-specific strategy is suggested for improving estimation accuracy.

7. Limitations

The availability of the right type of data was one of the limitations of this study. For example, hotel revenue data consist of a representative sample of the US lodging industry. More specifically, the data received from STR only comprise of hotels that voluntarily provide data to STR. Domestic trips was measured by the number of domestic flights only, which did not take into account domestic travelers traveling by other modes of transportations. In addition, there is room for adding more relevant explanatory variables to the model.

8. Conclusion

In summary, this study successfully developed a parsimonious model of hotel revenue and identified four major macroeconomic predictors of US hotel revenue specific to different hotel sub-segments. However, the range of variability the model exhibited indicates that there is room to add more predictors to the model. Therefore, future studies should consider more predictor variables such as exchange rate and cost of transportation. Furthermore, it might be worthwhile to test the utility of the model to hotel sub-segments grouped by their locations such as urban, suburban, airport and interstate settings.

Note

1.

NOHR for luxury and upper upscale sub-segments are (I (1)), that is to say, NOHR for all other hotel sub-segments are I(0).

Figure 1.

US hotel industry and six sub-segments revenue trend

[Figure omitted. See PDF]

Figure 2.

Plot of CUSUM and CUSUMSQ (stability test)

[Figure omitted. See PDF]

Table I.

STR Hotel classification

Hotel segment	ADR* range
Luxury	$150-220
Upper-upscale	$120-150
Upscale	$100-120
Upper-midscale	$85-100
Midscale	$70-85
Economy	$55-70

Note:

*Annual 2017 ADR Range

Table II.

Unit root test results using ADF and PP test

	ADF	PP
Variable	τc	τc+t	τc	τc+t
Aggregate revenue	−0.636 [13] (0.859)	−3.953** [13]	−2.950** [7]	−6.481*** [6]
Luxury revenue	−0.869 [13] (0.797)	−4.146*** [13]	−2.116 [53] (0.239)	−7.708*** [6]
Upper-upscale revenue	−0.827 [13] (0.809)	−3.703** [13]	−3.293** [17]	−7.350*** [8]
Upscale revenue	−0.261 [13] (0.927)	−3.968** [13]	−1.634 [12] (0.464)	−6.836*** [6]
Upper-midscale revenue	−0.557 [13] (0.876)	−4.152*** [13]	−2.411 [7] (0.140)	−6.318*** [6]
Midscale revenue	−0.960 [13] (0.767)	−3.345* [13]	−5.495*** [3]	−6.707*** [5]
Economy revenue	−2.461 [13] (0.126)	−3.085 [13] (0.112)	−6.276*** [2]	−6.668*** [3]
DT	−1.471 [13] (0.547)	−2.635 [13] (0.265)	−6.916*** [7]	−9.546*** [6]
CCI	−2.0740 (0.256)	−2.1640 (0.507)	−2.045  [1] (0.268)	−2.1640 (0.507)
IA	−0.615 [13] (0.864)	−2.112 [13] (0.536)	−3.641*** [8]	−6.744*** [8]
PCE	−3.315** 0	−1.8170 (0.694)	−2.828* [8]	−1.802 [8] (0.701)
NOHR _aggregate industry	−0.919 [12] (0.781)	−5.226*** [12]	−3.27** [9]	−3.440** [9]
NOHR _ luxury scale	−1.405 [12] (0.580)	−2.595 [12] (0.283)	−1.327 [8] (0.617)	−1.674 [9] (0.760)
NOHR _ upper upscale	−1.887 [12] (0.338)	−2.447 [12] (0.355)	−1.818 [11] (0.372)	−1.544 [11] (0.812)
NOHR _ upscale	−0.974 [12] (0.763)	−4.718*** [12]	−2.610* [10]	−2.449 [10] (0.353)
NOHR _ upper midscale	−1.124 [12] (0.707)	−5.743*** [12]	−3.708*** [8]	−2.926 [8] (0.156)
NOHR _ midscale	−3.333** [12]	−4.559*** [12]	−4.506*** [7]	−3.827** [7]
NOHR _ economy	−2.547 [12] (0.106)	−4.483*** [12]	−0.312 [9] (0.920)	−4.476*** [8]
Δ Aggregate Revenue	−2.675* [12]	NA	NA	NA
Δ Luxury Revenue	−3.228** [12]	NA	−40.775*** [51]	NA
Δ Upper-upscale Revenue	−3.141** [12]	NA	NA	NA
Δ Upscale Revenue	−2.792* [12]	NA	−20.908*** [18]	NA
Δ Upper-midscale Revenue	−2.580* [12]	NA	−17.004*** [13]	NA
Δ Midscale Revenue	−2.644* [12]	NA	NA	NA
Δ Economy Revenue	−2.411 [12] (0.140)	−2.429 [12] (0.364)	NA	NA
Δ DT	−4.752*** [12]	−4.743*** [12]	NA	NA
Δ CCI	−16.812*** [0]	−16.793*** 0	−17.001*** [5]	−16.987*** [5]
Δ IA	−4.257*** [12]	−4.269*** [12]	NA	NA
Δ PCE	NA	−17.413***0	NA	−17.676*** [8]
Δ NOHR_aggregate industry	−1.282 [11] (0.638)	NA	NA	NA
Δ NOHR_ luxury scale	−2.412 [11] (0.140)	−2.601 [11] (0.280)	−12.928*** [9]	−12.967*** [9]
Δ NOHR _ upper upscale	−2.405 [11] (0.141)	−2.807 [11] (0.196)	−22.061*** [11]	−23.081*** [11]
Δ NOHR _ upscale	−1.292 [11] (0.634)	NA	NA	−14.871*** [10]
Δ NOHR _ upper midscale	−1.066 [11] (0.729)	NA	NA	−12.130*** [9]
Δ NOHR _ midscale	NA	NA	NA	NA
Δ NOHR _ economy	−2.083 [11] (0.252)	NA	−14.678*** [8]	NA
1% level***	−3.457	−3.995	−3.455	−3.994
5% level**	−2.873	−3.428	−2.872	−3.427
10% level*	−2.573	−3.137	−2.573	−3.137

Notes:

All variables are in logs in the series and are results of stationery tests with intercept and with trend and intercept; [.]s for the ADF test are the appropriate lag lengths selected by SIC (Schwarz Info Criterion), and for PP test are the optimal bandwidths. △ denotes the first difference of the variable; *p < 0.1; **p < 0.05; ***p <0.01;

*p value is in the parenthesis

DT; – Domestic trips; CCI – Consumer Confidence Index; IA – International arrivals; PCE – Personal Consumption Expenditure; NOH – number of hotels

Table III.

ARDL Bounds test for Co-integration

	lnAggregate revenue	lnLux	lnUpperup	lnUpscale	lnUppermid	lnMidscale	lnEconomy
F-Bound test	14.644***	10.485***	4.739***	12.448***	25.723***	10.005***	10.543***

Table IV.

Optimal lag length for each hotel Sub-Segment revenue

	Aggregate industry	Luxury	Upper-upscale	Upscale	Upper-midscale	Midscale	Economy
Optimal lag length	3 or 8	8	3 or 6	4 or 8	6 or 8	8	7 or 8

Table V.

Estimated Long-Run coefficients for the industry and Sub-Segment revenue models

Coefficients Variables	Dependent variable and ARDL Model
	lnAggregate revenue	lnLux	lnUpperup	lnUpscale	lnUppermid	lnMidscale	lnEconomy
	(7,10,9,10,0,10)	(8,7,7,0,8,6)	(6,11,9,0,7,9)	(8,10,10,10,5,9)	(4,10,10,10,0,10)	(7,12,11,10,12,11)	(6,11,12,11,0,12)
In DT	0.203** (0.101)	−0.112 (0.123)	0.305*** (0.116)	0.154** (0.074)	0.072 (0.122)	0.148 (0.119)	0.390*** (0.135)
In CCI	0.176*** (0.012)	0.049** (0.024)	0.026 (0.013)	0.130*** (0.011)	0.146*** (0.015)	0.190*** (0.016)	0.258*** (0.019)
In IA	0.175*** (0.033)	0.444*** (0.041)	0.280*** (0.045)	0.279*** (0.026)	0.167 *** (0.038)	0.074* (0.038)	−0.003 (0.052)
In PCE	0.746*** (0.178)	2.008*** (0.184)	1.644*** (0.199)	0.924*** (0.133)	1.050*** (0.207)	0.767*** (0.214)	0.075 (0.263)
In NOHR	−0.004 (0.157)	−0.565*** (0.205)	−2.868*** (0.435)	−0.132 (0.105)	0.190 (0.115)	0.030 (0.133)	0.209 (0.237)

Notes:

*p < 0.1; **p < 0.05; ***p < 0.01;

*Note: standard error is in the parenthesis;

all regressions are with unrestricted constant and unrestricted trend.

F-Bound Test reported for Co-integration;

DT – Domestic trips; CCI – Consumer Confidence Index; IA – International arrivals; PCE-Personal Consumption Expenditure; NOH – number of hotels

Table VI.

Error correction representation for the selected ARDL models

	Dependent variable and ARDL Model
Coefficients Variables	<