The effect of online hotel reviews on hotel real estate pricing : a hedonic hotel real estate valuation approach within eight European tourist destinations

The effect of online hotel reviews on hotel real estate pricing

A hedonic hotel real estate valuation approach within eight European tourist destinations

Benno Houben 10898557

Tuesday, 7 July 2015

MSc Business Economics: Real Estate Finance University of Amsterdam

First Supervisor: dr. Martijn Dröes

Statement of originality

This document is written by Student Benno Houben who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

This paper investigates the effect of hotel quality on hotel real estate pricing, with a hedonic regression analysis. Hotel quality is addressed in previous studies before, by using classification systems. These classification systems however, differ per country and do not measure hotel performance. Instead, the guest perceived quality indicator ‘Guest experience index’ is used to measure quality in this study. Also, online rating of location quality and cleanliness is used as indicator for quality. In addition, this study is one of the first cross-city studies in hedonic hotel valuation. Hotel transaction prices from eight tourist destinations throughout Europe are used as dependent variable. The transactions took place between 2010 and 2015. Cross-city effects slightly improve the explanatory value of the model.

Hotel quality is found to significantly influence hotel real estate value. The ‘Guest Experience Index’ had to be used in a separate regression analysis because of high correlation with ‘Cleanliness’. The regression model with ‘Guest experience index’ as indicator for quality as well as the regression analysis with online rating for ‘Location’ and ‘Cleanliness’ as indicator for quality, significantly influence hotel real estate prices. The regression model with ‘Location’ and ‘Cleanliness as independent variables however, explained a slightly larger amount of the variation than the regression analysis with ‘Guest Experience Index’ as indicator for quality. ‘Location’ has an effect of 23.4% on hotel prices, and ‘Cleanliness’ has an effect of 15.6% on hotel prices, for an increase of 1 on a 10-point scale.

The results of this study imply that investors of hotel real estate should take hotel quality in account when acquiring new properties, and that there might be opportunities for hotels with lower quality to increase their real estate value. As ‘Cleanliness’ correlates highly with other quality indicators ‘Room’, ‘Value for money’, ‘Service’ and ‘Condition’, they could not be included in the regression analysis. The significance of ‘Cleanliness’ and high correlation with the other variables does indicate that there are several ways for hotel managers and investors to increase hotel quality. Future research should focus on how hotel real estate investors can influence hotel quality, and particularly against what costs.

Table of contents

3. Data & methodology ... 14

3.1 Methodology ... 14

3.2 Data sources ... 15

3.3 Variables ... 15

3.4 Correlation ... 19

4. Results ... 20

5. Limitations and future research ... 25

6. Conclusion ... 26

Reference list ... 28

List of abbreviations

• ADR: Average daily rate; the average rental income per paid occupied room. • F&B: Food & beverage.

• GOP: Gross operating profit.

• Hotel room pricing: the study of pricing hotel room rate.

• Hotel real estate pricing: the study of pricing hotel real estate properties. • Hedonic pricing.

• Hotel management: the entity or person that manages the hotel. The hotel management usually pays rent to the hotel owner.

• Hotel owner: the entity or person that is the owner of the hotel. The hotel owner usually receives rental income from the hotel management.

• NOI: Net operating income.

• Occupancy rate: the number of rooms in a hotel that is rented out, compared to the total number of rooms in the hotel property.

• OLS: Ordinary Least Squares

• RevPAR: Revenue per average room, can be calculated by multiplying the number of days in a year with the ADR and the occupancy rate.

1. Introduction

There are several real estate asset types. Compared to retail, office- and residential real estate, investing in hotel real estate happens less often (van Gool, Brounen, Jager en Weiz, 2007). As hotel transactions are still a relative small portion of the worldwide real estate transaction volume, studies concerning real estate pricing focus mostly on the more common real estate asset classes. Investment appetite for hotel real estate is growing however, based on the growing volume of hotel transaction since 2009 (JLL, 2015). To get a better understanding of hotel investing, this study will focus on price determinants of hotel real estate assets.

Despite the fewer amount of studies on hotel real estate pricing, compared to more common asset classes, there are different valuation methods for hotel real estate. DeRoos and Rushmore (1993) mention the income capitalization approach (ex-ante), sales comparison approach (ex-post) and the costs approach, of which the income capitalization approach is the most common. Ex-ante models do not take several important variables into account. This is demonstrated by Corgel and deRoos (1993) who tested the ‘Average Daily Rate (ADR) rule of thumb’ (value of the hotel is a 1000 times the ADR rate), finding that the rule of thumb generally generates predictions that are not consistent with the actual market value. Ex-post pricing models on the other hand could give better estimates of the actual market value of real estate. Due to the heterogeneous nature and lower transaction volume of hotel real estate compared to other real estate markets, the volatility of ex-post hotel valuations is expected to be relatively high. Corgel and deRoos (1993) indicate that hedonic models can give a greater objectivity in ex-post hotel real estate valuation.

The few current hedonic pricing models for hotel real estate, focus primarily on the United States and explain up to 79.91% of the variation in hotel pricing (Corgel, Lin & White, 2015). These models take location aspects and hotel quality into account, but to a limited extent. Corgel (2007) for example, takes state fixed effects into account. To our current knowledge, there has not been a cross-city European based study on hotel real estate pricing. Also, current hedonic pricing models only take hotel classification into account as indicator for quality. This might not be a good indicator for hotel quality, as it does not take service quality or value for money into account, for example. A key contribution of this thesis is that online guest reviews will be taken into account as indicator for hotel quality.

The quality of a hotel is inseparably connected with the hotel real estate price. When hotels have a higher quality, this is associated with a higher Gross Operating Income (GOI) because of higher guest retention and attraction, and will eventually increase hotel performance. When the hotel performance is better, this is associated with a risk reduction for the property owner (in the case of a owner-tenant situation). It is difficult to control for quality, let alone explicitly measure quality. Previous studies that focus on measuring hotel quality give reason to believe that guest satisfaction is a good indicator for hotel quality. Claver, Tarí and Pereira (2006) for example, state that improving hotel quality will result in higher guest satisfaction. Online guest reviews measure guest satisfaction, and are hence seen as an indicator for hotel

quality. To current knowledge, online guest satisfaction has never been used before in a pricing model for hotel real estate.

It is difficult to take online guest reviews into account when making an ex-ante real estate valuation. Ex-post valuation models however, can take qualitative aspects of real estate assets into account, and enable to find cross country- or city results. Therefore, this study will use an ex-post approach to explain important variables that influence hotel real estate pricing. The research question of this thesis will be:

To answer the research question, the following sub-questions will be answered in the literature review:

1. What are current known determinants of hotel real estate value?

2. How can hotel quality be measured and will quality influence hotel real estate pricing?

Hedonic regression (ex-post) is commonly used to appraise residential real estate, but for hotel real estate this method is not widely adopted. This is most likely to be due to the scarcity of data. As a result, there are not many studies devoted to this subject. The previous studies that write about this subject will be reviewed in Chapter 2.

The logarithm of the transaction price as dependent variable and hotel individual data (including quality) as independent variables, the variable coefficients can be estimated with Ordinary Least Squares (OLS). Variables mentioned in other studies concerning hotel pricing will be taken into.

Transaction data of hotel real estate is retrieved from Real Capital Analytics, HVS Hospitality and from PwC Real Estate Advisory. Real Capital Analytics collects worldwide real estate transaction data from different property types. Their data contains over 25.000 hotel sales prices, of which 11.000 outside the US. Due to limited access to the data, about 200 of the 11.000 sales prices were gathered, of transactions between 2010 and 2015. HVS Hospitality is a hotel consultancy firm who bring out a yearly report containing European hotel transactions. PwC Real Estate Advisory collects transaction data of hotels in different countries through public and private sources.

STR Global provides hotel data such as chain affiliation, presence of swimming pool and hotel age, which can be used as independent variables in the OLS regression. The STR Global data also contains the hotel classification so that an indicator for ADR is taken into account. The guest reviews of the specific hotels are gathered from Olery, a firm that collects and analyses guest reviews through data-mining. 180 complete observations were found after merging all the data.

The results of this study indicate that quality significantly influences hotel real estate prices. The overall indicator for quality ‘Guest Experience Index’ as well as the quality indicators ‘Location’ and ‘Cleanliness’ showed significant results. This indicates that hotel managers or investors can increase the value of their property by increasing the quality of their hotel properties. ‘Guest Experience Index’ highly correlates

with ‘Cleanliness’ and could therefore not be used simultaneously in a regression analysis. ‘Cleanliness’ highly correlates with other quality indicators ‘Room’, ‘Value for money’, ‘Service’ and ‘Condition’. This indicates that these variables could all be off importance when measuring overall quality. This study uses a relatively low number of observations. A larger number of observations could help in increasing the explanatory value of the model.

The remainder of this thesis is structured as follows. Chapter 2 consist out of a literature review. The literature section will be used to explain the main theories about hotel pricing and again highlighting what this thesis contributes to the existing literature. In Chapter 3, the chosen methodology is discussed, followed by ‘Data’ where the dataset will be shown. Chapter 4 consists out of the results of the analysis followed by Chapter 5, in which the limitations and recommendations for future research will be stated. Chapter 6 consist out of the conclusion of this thesis.

2. Literature review

The literature review will focus on 3 subjects: price determinants of hotel real estate market value, price determinants of hotel income indicators, and hotel quality. The literature review will serve as input for the data & methodology section in Chapter 3, as the data that is needed is based on previous studies. Also, significant determinants of hotel real estate market value that are found in previous studies will be taken into account during the data collection.

2.1

Price determinants

In hedonic valuations, macroeconomic and regional economic indicators influence property value, next to property fundamentals and location (Hoag, 1980). There are several studies that test which of these aspects influence hotel real estate pricing, finding that hotel real estate pricing is indeed influenced economical, property fundamentals and locational aspects. Most of the studies used data containing solely hotels in the United States. One of the studies however, focuses on price determinants of hotels in the United Kingdom. None of the studies take fixed city effects and hotel quality into account.

The most recent study on hotel real estate pricing is written by Corgel, Liu and White (2015). In this study, the authors test whether hedonic valuation of hotel real estate becomes more accurate by adding city-specific Net Operating Income (NOI) of hotels. They use a sample of U.S. hotel transactions between 2005-2010 and test their hypotheses with OLS and Instrumental variables regression. Their hedonic models explain between 75% and 80% of the variation in prices. Quality in this study is addressed through the hotel classification variable retrieved from the STR global database. They could not conclude that adding income indicators in their analysis leads to a significantly higher adjusted R-squared value.

Blal and Graf (2013) use hedonic regression analysis to find whether hotels that are larger or smaller than the norm (in terms of number of rooms, thenumber of Food & Beverage (F&B ) outlets and the square meter of meeting space) influence hotel pricing. They find that when hotels are at the norm in terms of these variables, hotels are sold with a premium, contrary to when hotels are not sized at the norm, which leads to a discount when selling the property. In other words, if hotels have an excessive amount of rooms, this will lead to a price discount compared to hotels that have a more usual amount of rooms. They also find that there are differences in the extent of the discount or premium for different hotel segments. For hotel fundamentals, Blal and Graf (2013) take amenities like the presence of tennis, golf, pools, and fitness facilities into account, and for location they determine whether the location is urban, suburban or located at a highway or airport. Additionally, they found that the per capita income at the time of the sale also influences hotel pricing. Blal and Graf (2013) had by far the largest dataset of all studies (N=10722).

Real estate generally depreciates in value due to obsolescence. It is possible to renovate properties, although this causes properties to close for some time. Corgel (2007) uses a hedonic price model to find what the effect of age on hotel properties is. He finds that hotel properties decrease 1.93% in value per

year due to obsolescence, which is ought to be comparable with obsolescence in retail properties. There are differences however compared to other property types regarding obsolescence, as age turns to be positively affecting price after 3 decades. This indicates a vintage effect. Next to age, Corgel (2007) found that scale (Deluxe, Luxury, Upscale, Upper tier, Midscale, Lower tier, Economy and Budget) significant influence hotel pricing, similar to per capita income at the time of the sale. Corgel (2007) does control for state fixed effects, contrary to all other studies in this literature review.

In an earlier study of the same authors (Corgel and deRoos, 1992) the aim was to create a price index and see whether a change in tax treatment of real estate quickly affected hotel real estate, and was one of the first studies that used hedonic valuation for hotel real estate. They used the following variables, but did not report on the coefficients and significance levels of these variables:

- ARR or occupancy rate at time of sale - Number of rooms - Age - Ownership form - Chain affiliation - Casino - Conference centre - All suites

- Limited service facility - Gold

- Tennis

- Pools - F&B

- Distance to nearest airport

- Distance to nearest convention centre - Number of employees

- Change in annual rate

- Unemployment rate in the country during the month of the sale

- Effective buying income per capita in the country during the year of sale

The only study that used hotel real estate transactions outside the U.S. was conducted by Roubi and Litteljohn (2004) for hotels in the United Kingdom. Their model explains 79.4% of the variation in prices. Contrary to Blal and Graf (2012), Roubi and Litteljohn (2004) found that Number of F&B outlets and the number of employees did not significantly influence hotel pricing. Also, Roubi and Litteljohn (2004) found different significance levels for the date of the sale, in comparison with Corgel (2007). This could indicate that the U.S. hotel market behaves different than the U.K. hotel market.

Table 1summarizes the different variables that were taken into account in each study. A more extensive overview of the results from previous studies is added in the appendix (Table 6). The variables are sorted on category (NOI, Hotel fundamentals, Location, Classification, Transaction specific, Year of sale). All of these categories were taken into account in each study, except for location and NOI. Only Corgel, Liu and White (2015) use NOI in their analysis. Location is addressed in all studies except for Corgel (2007). As none of the studies use cross-city variation, adding this effect could improve hotel real estate pricing models.

Table 1: Comparison of previous hedonic hotel pricing models _{I II III IV V} NOI _✓ Hotel fundamentals _{✓ ✓ ✓ ✓ ✓} Location _{✓ ✓ ✓ ✓} Classification _{✓ ✓ ✓ ✓ ✓} Transaction specific _{✓ ✓ ✓ ✓ ✓} Year of sale _{✓ ✓ ✓ ✓ ✓} Management _{✓ ✓}

Method 1 stage OLS 2-stage OLS WLS 1 stage OLS 1 stage OLS

N= N=623 N=623 N=3810 N=10722 N=211

Sample start 2005 2005 1996 1980 1996

Sample end 2010 2010 2004 2008 2001

Location U.S. U.S. U.S. U.S. U.K.

State effects No No Yes No No

Dependent variable Ln Price Ln Price Ln Price/room Ln Price Ln Price Adj. R-squared 0.7731 0.7991 0.5347 0.725 0.794 I = Corgel et al. (2015), II = Corgel et al. (2015), III = Corgel (2007), IV = Blal and Graf (2012), V = Roubi and Litteljohn (2004).

2.2

Income determinants

As income is an important variable in ex-ante valuation models, variables that significantly influence hotel income should also be taken into account when constructing a hedonic pricing model for hotel real estate. This is confirmed by Corgel, Liu and White (2015), who found that NOI is significantly influencing hotel real estate value. Next to valuation purposes, room pricing is also important to maximise the owners’ investment objective and satisfy heterogeneous guests (Steed & Gu, 2004). When a hotel decides to decrease ADR in order to attract more guests, this may lead to a high occupancy rate but deteriorates Gross Operating Profit (GOP). Also, when a hotel charges relatively low prices in relation to their offered services, this may lead to lower guest satisfaction. Room pricing therefore is a crucial part in managing a hotel property.

There are several studies that find which variables influence income indicators of hotels (Revenue Per Average Room (RevPAR), NOI, Single/double room rate). Zhang et al. (2011) regressed RevPAR of 228 Beijing hotels with the number of rooms, star-rate, number of years since property was built or refurbished, straight line distance between hotel and its nearest scenic spot, straight line distance between hotel and its nearest transport hub. They found that all variables significantly influence RevPAR, except for distance to nearest scenic spot.

With a sample of 72 lodging firms between 3- and 5 star rating in Milan, Sainaghi (2010) found several RevPAR determinants. The author confirms that both positioning and market orientation are influencing RevPAR significantly. Sainaghi (2010) found that age of the hotel did not significantly influence RevPAR. O’Neill and Mattila (2006b) found that age did significantly influence NOI of hotels. With a sample of

1954 U.S. hotels in the years 2002 and 2003 O’Neill and Mattila (2006b) found that NOI is influenced by occupancy rate, ADR and chain scale, age of the property, brand affiliation and location. Hotel region however, was not significantly influencing NOI.

Another study by Thrane (2006) used room rate as dependent variable for 74 hotels in Oslo. He found that significant influences of room rates differ when switching between ‘single room rate’ and ‘double room rate’ as dependent variable.

All the studies that were reviewed are displayed in Table 2. A more extensive overview of the results is displayed in Table 7, which is added in the appendix. O’Neill and Mattila (2006b) show that adding income (ADR and occupancy rate) as independent variable in the regression analysis changes coefficients of most variables, but also increases the adjusted R-squared from 59.1% to 64.3%. The comparison does not lead to additional variables that could be of importance in a hedonic hotel real estate pricing model. Table 2: Comparison of studies on hotel income indicators

I II III IV V VI

Location _{✓ ✓ ✓ ✓ ✓ ✓}

Income _✓

Hotel fundamentals _{✓ ✓ ✓ ✓ ✓ ✓}

Method 1 stage OLS 1 stage OLS 1 stage OLS 1 stage OLS 1 stage OLS 1 stage OLS

N= N=228 N=72 N=1954 N=1954 N=74 N=74

Location Beijing Milan U.S. U.S. Oslo Oslo Dependent variable RevPAR RevPAR NOI NOI room rate Ln Single Ln double room rate

R-squared 0.703 0.705

Adj. R-squared 0.537 0.592 0.591 0.643 I = Zhang et al. (2011), II = Sainaghi et al., III = O’Neill and Matilla (2006b), IV = O’Neill and Matilla (2006b), V = Thrane (2006), VI = Thrane (2006).

2.3

Hotel quality

As quality of hotels is a container term for many aspects, this part of the literature study will focus on how to measure quality and which aspects influence the perception of quality. When these aspects are known, they can be added to the hedonic pricing model. Guest satisfaction, star classification, design, hotel brand, room rate, service, maintenance and staff attentiveness are indicators of quality, according to previous literature. These aspects will be reviewed in the following part.

When the occupancy rate of a hotel is not as high as demanded, hotel management can choose to decrease the room price (ADR) to attract more guests. This will affect profits drastically, so in stead, hotels should think of other factors that are influencing hotel choice. Chan and Wong (2005) suggest that facilities and the service quality influence booking behaviour. By holding a survey amongst 573 frequent individual travellers in Hong Kong, they found that location and good service are significant factors that influence hotel choice, but also previous hotel experience, convenience and company recommendation matter.

By using 3.875 actual guest satisfaction surveys at an upper midscale hotel, with a timespan of 3 years, Matilla and O’Neill (2003) found that the room rate was a significant predictor for overall guest satisfaction and guest room cleanliness, maintenance and attentiveness of staff. Occupancy rate however, did not seem to be a good predictor for guest satisfaction. As the income of a hotel is apparently influenced by guest satisfaction and other qualities that a hotel real estate contains, it is not un-doubtable that this form of quality influences the price of a hotel real estate asset. The results of Matilla and O’Neill (2003) do contradict Claver, Tarí and Pereira (2006), who state that tourists do not only focus on price. One way or the other, price seems to be connected to the perception of quality.

Claver, Tarí and Pereira (2006) state that hotel competitiveness must focus on improving quality and differentiation from competitors, to increase performance. Increasing competitiveness caused quality to become a more important factor for hospitality companies (Ingram and Daskalakis, 1999; Costa, 2004). Thus, quality could also be of importance in the value assessment of hotel real estate. Quality of hotels can be influenced internally and externally (Claver, Tarí and Pereira, 2006):

- Internal quality: increase productivity, efficiency, cost reduction, waste reduction, and results in increasing competitiveness, higher profitability through process standardization, fewer errors, effective service.

- External quality: improving corporate image, gaining competitive advantage, adapting to guests requirement, exploring the possibility to enter new markets, and results in guest satisfaction, increasing market share, new tourists.

Hotel quality is often linked to certification of hotels. There are many programmes, qualification and quality seals, which can cause confusion. Minazzi (2010) compared different classification systems. He found that there is a heterogeneous situation in a sense that hotels in different countries can have a similar quality symbol, but that the requirements for the different symbols are different per country. Electronic distribution channels of hotel ratings are becoming more common to assess hotel quality and causes brand not to be the most considered factor when booking hotels. Brands and hotel chains however, do use quality standards, service procedures and similar service levels, so that higher guest satisfaction is reached. Brand association could also be perceived as a form of quality. O’Neill and Mattila (2006c) find that brand association is considered to be value improving for hotel real estate pricing.

There are inconveniences when improving quality through qualification and brand association (Ingram and Daskalakis, 1999; Nield and Kozak, 1999):

- Extra costs obtaining quality certification - Time spent

- Greater operational bureaucracy

- Lack of attention to personnel development

This is partly confirmed by Claver, Tarí and Pereira (2006) who studied the effect of quality certification on performance. They find that positive effects on performance can be identified, but the impact on

financial performance is low. This points out that hotel certification is not a good indicator for quality when used in a hedonic pricing model for hotel real estate.

According to previous studies, quality of hotels can be captured by guest satisfaction, room rate, good service, maintenance, and staff attentiveness. Hotel brands and hotel classification systems, are not always mentioned as a good indicator for quality. Also, brand affiliation does not necessarily influence profitability, due to higher costs. Brand affiliation however, is linked to a price premium for hotel real estate. Investors might see brand affiliation as a decrease in risk, so that the price they pay is higher while income remains stable (as brand affiliation does not influence profitability of the hotel management). Guest satisfaction is expected to contain good service, maintenance, and staff attentiveness. In this thesis therefore, guest satisfaction shall be included in the hedonic pricing model. This should give a better indication of hotel quality than including hotel certification levels.

3. Data & methodology

3.1

Methodology

In previous studies that involve hedonic hotel valuations, regression analysis is generally used to find which variables significantly influence hotel prices. Previous studies used hotel transactions as dependent variable, transformed to natural logarithm and sometimes to hotel price per room. The independent variables that are used in previous studies are not exactly similar in all studies. However, the independent variables that were used in previous studies can be categorized in seven different components:

1. Hotel fundamentals

2. Location characteristics (Cross-city effects) 3. Income characteristics

4. Classification

5. Transaction Specifics 6. Year of sale

7. Management

In addition, the literature study on hotel quality suggests that hotel quality can impact hotel pricing. Therefore, the following component will be taken into account:

8. Hotel quality

The following hedonic model is estimated in this thesis:

β0 represents the intercept of the regression, β1, …, β8 represent the coefficients of the independent

variables, λ represents the time fixed effects and ε represents the error term. The influence of the coefficients will be tested on significance and can be compared to each other. In other words, the model will show which coefficients are different from 0 and to what extent the significant coefficients influence hotel prices.

In the previous studies, most of the components significantly influenced hotel prices. For this model, similar results are expected. In addition, cross-city effects are expected to have a positive influence on the adjusted R-squared value, as location aspects in the research method by Blal and Graf (2012) had relatively high coefficients and could therefore have a large explanatory value. Also, quality is expected to have a significant influence on hotel real estate value.

Ln(Hotel) Price)it=) β0) +) β1HotelFundamentalsi) +) β2LocationCharacteristicsi) +)

β3IncomeCharacteristicsit) +) β4Classificationi+) β5TransactionSpecificsi) +) β6YearOf)

Salei)+)β7Managementi)+)β8HotelQualityi)+)λt)+)εit)

3.2

Data sources

The previous studies on hotel pricing give an indication of which data is needed to conduct this research. Next to the hotel transaction prices, the eight components as discussed in the methodology section are needed to execute the regression analyses.

The hotel transaction prices are provided by PwC, Real Capital Analytics and HVS. PwC is a worldwide accounting- and consulting firm. PwC collects commercial real estate transactions, including hotels, used for consultancy and valuation purposes. Real Capital Analytics is a data and analytics firm that collects commercial real estate transactions for a database that customers can subscribe too. HVS is a consultancy firm focussing hotels. HVS publishes several reports concerning hotels on an annual basis, and also publishes hotel transactions in their reports. The hotel transactions are used as dependent variable in the regression analysis, and took place in the year 2010 until 2015.

For the independent variables, STR Global provided most data. STR Global collects data from hotels, like amenities, ADR rates, supply and demand, hotel pipeline. How the data is used and transformed will be described later in this chapter. The STR Global data includes hotel quality to some extend, but does not include client reviews. Previous studies suggest that client reviews are a better indicator of hotel quality, compared to hotel certification. Olery is a Dutch company that gathers guest reviews from worldwide hotel booking sites like Booking.com and Tripadvisor. This data is generally used by hotels to improve their online reputation. Olery provided access to their data, which will be used as indicator for quality in the regression analysis.

3.3

Variables

After collecting and merging all data, there are 180 complete observations. The descriptive statistics of the dependent and independent variables that will be used in the regression analysis are displayed in Table 3 and Table 4, respectively. The transaction price of each observation is transformed as the natural logarithmic value, so that the true changes in price can be determined. Some hotels have a relatively high value per room in comparison to other hotels. As these hotels are all marked as ‘Luxury class’, ‘Upper upscale class’ or ‘All suites’, they are not seen as outliers as all of these hotels are luxurious hotels or landmark buildings.

Table 3 shows the descriptive statistics for the selling price per room of hotel real estate (instead of the natural logarithm of the total hotel real estate selling price, that will be used as dependent variable). The average selling price of the 180 hotels was €318,981.90. The descriptive statistics for hotel real estate selling price per room in each separate city are also displayed in Table 3. The hotels in London and Paris have an average selling price per room that is higher than the average of all 180 transactions. More specifically, the hotels in Paris have an average selling price per room that is more than two times larger than the average of all hotel transactions. The hotels in Manchester and Berlin have the lowest average selling price. The differences between cities in transaction prices do not necessarily indicate that hotel real

estate is more expensive in a specific city, as it can be possible that these differences can be explained by other variables.

Table 3: Descriptive statistics for real estate price / room

Variable N Mean Std. Dev. Min Max

Descriptive statistics for all cities

Real estate price / room 180 € 318,981.90 € 334,374.81 € 15,079.37 € 2,101,449.28

Descriptive statistics for 'Real estate price/room' per city

Amsterdam 25 € 242,398.76 € 199,579.21 € 46,938.78 € 1,034,791.74 Dublin 18 € 174,866.69 € 93,037.49 € 44,333.32 € 477,758.16 London 63 € 388,826.72 € 327,577.06 € 15,079.37 € 1,257,704.58 Berlin 22 € 125,150.74 € 74,498.19 € 38,149.44 € 292,397.66 Munich 9 € 170,566.99 € 82,412.66 € 78,373.57 € 335,885.16 Manchester 9 € 131,819.24 € 93,492.36 € 51,200.14 € 350,830.94 Paris 23 € 648,802.67 € 525,056.57 € 98,252.28 € 2,101,449.28 Madrid 11 € 301,439.59 € 357,510.04 € 98,534.72 € 1,311,914.40

Table 4 displays the descriptive statistics for the independent variables, that are divided in the categories ‘Hotel fundamentals’, ‘Location’, ‘Classification’, ‘Transaction specifics’, ‘Year of sale’ and ‘Management’, which will be used as control variables. Table 4 also displays descriptive statistics for variables that measure ‘Quality’, which will be used to find whether quality is significantly influencing hotel real estate value.

For ‘Hotel fundamentals’, STR provided all data. The number of rooms differs between 24 and 1025, with an average of about 192 rooms, which indicates that hotels have diverse sizes. For all other variables within ‘Hotel fundamentals’, dummy variables were created. Because the relatively high number of missing values for ‘Age at sale’, an additional dummy variable was created indicating the missing ages. ‘Convention’ indicates hotels with 300 rooms or more, and have a minimum of 20,000 square feet of meeting space’. ‘Boutique’ is an indicator for hotels that are smaller than 200 rooms and provide authentic experiences regarding culture or history, and provide a high service level and have relative high ADR. The ‘All suites’ variable indicates that all hotel rooms have one or more bedrooms and include a separate living area.

Next to city names, the STR data provides data for location within a city (Urban area, suburban area, small metro/town area), which respectively indicates hotel location in a densely populated area, suburb or areas with a smaller population and limited services. These variables might not be appropriate indicators for location as the dataset only contains hotels within most of Europe’s largest tourist destinations. The variables will therefore not be used in the regression analyses. As the STR data also includes longitude and latitude coordinates of the hotels, it is possible to calculate distances of each hotel property to certain points of interest. Previous studies included ‘distance to airport’ and ‘distance to nearest transport hub’ as dependent variables. In this study, distance to nearest airport and distance to cities’ major transport hub are included as indicators for presence of airports or major transport hubs. The coordinates of the cities

airports and major trains stations are found on Wikipedia and are checked for correctness on www.gpscoordinates.net, a website which enables to plot coordinates on a map. The distance between the hotels and the points of interest are calculated with Haversine formula, which takes the following form:

Latitude1 and longitude1 represent the coordinates of the hotel, and latitude2 and longitude2 represent the coordinates of the point of interest. R represents the radius of the earth, which is 6373 kilometres. The mean and standard deviation in Table 4 show that hotels are on average closer to a cities major transport hub than they are to an airport. The standard deviation for distance to airport is on the other hand lower than for distance to a cities major transport hub.

For the components ‘Classification’, ‘Transaction specifics’, ‘Year of sale’ and ‘Management’, dummy variables are created. Except for ‘Year of sale’, all data was gathered from STR. ‘Year of sale’ was adopted from the transaction data. ‘Classification’ indicates the class of the hotel, ranging from ‘Luxury’ to ‘Economy’. The classification is based on the ADR-rate relative to other hotels in their geographical proximity.

Within ‘Transaction specifics’, ‘Part of portfolio-deal’ indicates whether the hotel transaction was part of a portfolio deal, which could have led to a discount or premium. Also, ‘Management change’, ‘Sale & lease-back’ and ‘Buylease-back’ respectively indicate whether the management firm was changed after the sale, if the managing company was the previous owner or whether the management company bought the hotel property. As STR reports about the current management firm, previous management firm, the companies were compared to the buyer and seller of the hotel properties to construct the variables. Also, ‘Transaction specifics’ includes a variable that indicates whether the property was a development or an existing property. Under ‘Management’, three variables are used that indicate whether the hotel is managed by an independent company, a chain management firm or whether the property is a franchising company.

For hotel quality, Olery provided different variables that indicate a score during the year of sale of the hotel property, of the variables ‘Value’, ‘Service’, ‘Location’, ‘Food’, ‘Family friendly’, ‘Facilities’, ‘Ambiance’, ‘Cleanliness’ and ‘Condition’. The score is based on online guest reviews from different online sources, and differs between 0 and 10. The average score ranges between 6.61 for facilities and 8.47 for cleanliness. The online reviews do not report on all aforementioned variables consistently. The variables ‘Family friendly’, ‘Facilities’, ‘Food’ and ‘Ambiance’ will not be used for further analyses because of the large amount of missing observations.

Distance) =) arccos(cos(90Rlatitude1)) *) cos(90Rlatitude2)) +) sin(90Rlatitude1)) *) sin(90Rlatitude2))*))cos(longitude1Rlongitude2))*)R)

Table 4: Descriptive statistics of all independent (control) variables

Variable N Mean Std. Dev. Min Max

Ho tel fu nd am en ta ls Number of rooms 180 192.9389 148.2264 24 1025 Age at sale 180 37.84444 64.3026 0 413 Dummy for missing age 180 0.1944444 0.3968764 0 1 Restaurant 180 0.8277778 0.3786267 0 1 Convention facilities 180 0.0166667 0.1283762 0 1 Spa 180 0.1333333 0.3408828 0 1 Boutique 180 0.0444444 0.2066553 0 1 All suites 180 0.0666667 0.2501396 0 1 Meeting space 180 0.55 0.4988814 0 1 Lo ca tio n Amsterdam 180 0.1388889 0.2066553 0 1 Dublin 180 0.1 0.3008368 0 1 London 180 0.35 0.4783001 0 1 Berlin 180 0.1222222 0.3284559 0 1 Munich 180 0.05 0.2185529 0 1 Manchester 180 0.05 0.2185529 0 1 Paris 180 0.1277778 0.3347734 0 1 Madrid 180 0.0611111 0.2402022 0 1

Distance to cities' closest major

transport hub 180 9.26828 9.337745 0.1663124 43.50518 Distance to closest airport 180 13.15423 5.86481 0.5502241 34.02678

Cl

as

sif

ica

tio

n Luxury class _{Upper upscale class} 180 ₁₈₀ 0.1222222 _0.2166667 0.3284559 _0.4131227 0 ₀ 1 ₁

Upscale class 180 0.2277778 0.4205683 0 1 Upper midscale class 180 0.2111111 0.4092354 0 1 Midscale class 180 0.0888889 0.2853771 0 1 Economy Class 180 0.1277778 0.3347734 0 1 Tr an sa cti on sp ecif fic Part of portfolio-deal 180 0.0444444 0.2066553 0 1 Management change 180 0.6722222 0.4707127 0 1 Sale & lease-back 180 0.1333333 0.3408828 0 1

Buyback 180 0.1444444 0.3525204 0 1 Development 180 0.0333333 0.1800062 0 1 Ye ar o f s al e Sold in 2010 _{Sold in 2011} 180 ₁₈₀ 0.0722222 _0.1277778 0.2595776 _0.3347734 0 ₀ 1 ₁ Sold in 2012 180 0.0944444 0.2932618 0 1 Sold in 2013 180 0.2111111 0.4092354 0 1 Sold in 2014 180 0.3666667 0.4832386 0 1 Sold in 2015 180 0.1277778 0.3347734 0 1 Ma na gm t. Independent management 180 0.25 0.4342205 0 1 Franchise management 210 0.2 0.40111158 0 1 Chain Management 180 0.55 0.4988814 0 1 Ho tel q ua lit y

Guest experience index 180 79.80328 10.07276 29.65 97.64

Room 180 7.907222 1.166479 2.4 9.6 Value 171 7.687719 1.137914 2.2 10 Service 180 8.172778 1.102452 2.8 10 Location 180 8.430556 0.9584145 4 10 Food 149 7.866443 1.3826296 1.3 10 Family friendly 26 6.753846 2.431334 1 10 Facilities 89 6.606742 2.291079 1.6 10 Ambiance 22 7.727273 0.7838743 6.7 9 Cleanliness 180 8.473333 1.096954 2.9 10 Condition 140 8.012857 1.272275 2 10

An additional variable that Olery provided is the ‘Guest experience index’, which is a score that is based on all aforementioned quality variables. This score is corrected for false or incorrect online reviews, so it gives a true indication of guest experience, while the variables in the previous paragraph might not give a true indication. As the ‘Guest experience index’ is based on all the other quality variables, it has to be used separately in the regression analyses. The ‘Guest experience index’ has a minimum of 29.9 and a maximum of 97.64 for the hotels in de database, during the year of the transaction. The mean of the ‘Guest experience index’ is 79.80.

Some variables that were used in previous studies are not included in the data analysis. ‘Number of employees’, ‘Landmark hotel’, ‘Water location’, ‘Parking space available’, ‘Tennis court available’, ‘CBD location’, ‘REIT buyer’ and ‘Capitalization rate’ were not available. Also, recently renovated was not included because of a large number of missing data. ‘Income per capita’ is also not included as the income rates are not available yet for 2015. As Corgel, Liu & White (2015) found that adding income variables to their regression analysis did not result in a large increase in adjusted R-squared. The dataset as described in this chapter is expected to be sufficient to answer the research question and test the research questions.

3.4

Correlation

The correlation between the independent variables is displayed in Table 8. There are some variables that show high correlation coefficients, indicating that they cannot be used simultaneously in the regression analyses. Between the variables that indicate quality of ‘Room’, ‘Value’, ‘Service’, ‘Cleanliness’ and ‘Condition’, there is a high correlation (ranging between 0.57 and 0.90). Therefore, only one of these variables can be used in the regression analyses simultaneously. The variable with the highest predictive value will be chosen during the data-analysis. The additional indicator for guest perceived quality ‘Guest experience indicator’ was composed with the aforementioned variables, and could therefore not be used simultaneously. The high correlation between ‘Room’, ‘Value’, ‘Service’, ‘Cleanliness’, ‘Condition’ and the ‘Guest experience indicator’ again confirm this. Note that the variables that indicate a specific city and the year of sale, are not included in the correlation matrix. To make sure that multicollinearity is not occurring, the regression analyses have to be checked with a variance inflation factor (VIF).

4. Results

In this chapter, the results of the regression analyses are discussed. The results are displayed in Table 5. In total, three regression analyses are performed, which are all different versions of equitation (I). All regression analyses contain the control variables ‘Location’, ‘Classification’, ‘Hotel fundamentals’, ‘Management’, ‘Transaction specifics’ and ‘Year of sale’. The three regression analyses contain different indicators for quality, and do not always contain cross-city effects. The first two regression analyses use the ‘Guest experience index’ as indicator for quality. The third regression analyses uses online guest review scores for ‘Location’ and ‘Condition’. Fixed city effects are included in the second and third regression analyses. The empty rows for the variables ‘Upper midscale class’ and ‘Franchise management’ indicate omitted variables that are used as reference category.

The first regression analysis (that does not include fixed city effects) indicates that the ‘Guest experience index’ significantly influences hotel real estate prices. Except for ‘Management’, all control variable categories include one or more variables that significantly influence hotel real estate price. The variables ‘Distance to cities closest major transport hub’ and ‘Distance to nearest airport’ are both significant in regression analysis one, indicating that an increase of one kilometre causes a 1.32% and 2.44% increase in real estate value. This result might be surprising, as this indicates that hotels nearby airports and major transport hubs are sold with a discount, instead of a reachability premium. It could be the case that hotels close to airports or transport hubs cope with noise disturbance, which might lead to a price discount. In regression analysis I, hotels with classification ‘Luxury class’, Upper upscale class’, ‘Upscale class’ and ‘Economy class’ are sold with a premium of respectively 115.6%, 75.7%, 29.2% and 47.4%, compared to hotels with classification ‘Upper midscale class’ or ‘Midscale class’, who are not significantly different from each other. In particular the price premium for hotels with ‘Economy class’ is surprisingly, as these hotels theoretically have the lowest ADR. An explanation for this price premium can be that hotels with lower ADR have higher occupancy rates, or have a better value for money. ‘Economy’ hotels can also have lower costs relative to other hotels, which leads to a relatively higher NOP.

The insignificant variables within ‘Hotel fundamentals’ are ‘Convention’, ‘Boutique’ and ‘Meeting space’, indicating that hotels with convention centres, meeting space are not sold with a premium or discount, as well as hotels that are marked as boutique hotels. The number of rooms does influence hotel real estate pricing, with an increase of 0.264% per additional room according to the first regression analysis. Also, the age of the hotel significantly improves real estate value. This can indicate that there is a vintage effect for hotels, as predicted in previous studies. It can also indicate that hotels are in business for a longer time and can therefore work on client retention and hotel reputation. When hotels have the amenities ‘Restaurant’, ‘Spa’ or ‘All suites’, this respectively leads to a 35.9%, 36.9% and 52.8% increase in hotel pricing, in regression analysis I.

For ‘Transaction specifics’, the variables ‘Portfolio deal’ and ‘Management change’ have a positive influence on hotel pricing in regression analysis I, which increases hotel real estate price with 41.2% and 33.5%, respectively. For portfolio transactions it could be possible that there are scale advantages for the investor, but the results indicates otherwise. Hotels that had a ‘Management change’ were also purchased with a premium. This can indicate that hotel investors think they can increase the ROI by changing the management company.

In regression analysis II, cross-city effects are added. Compared to regression I, the variable ‘Distance to cities closest major transport hub’ loses its significance, while the coefficient for quality indicator ‘Guest experience index’ increases from 0.0164 to 0.0190. This indicates that an increase of 1 (on a scale of 100) in ‘Guest experience index’ leads to a price premium of 1.90%, according to regression II.

Next to changes in ‘Guest satisfaction index’, regression analysis II reports lower coefficients for hotels with ‘Luxury’ and ‘Upscale’ classification (0.814 and 0.474), while the coefficients for ‘Upscale class’ and ‘Economy class’ have grown with to respectively 0.302 and 0.542. Also, the ‘Number of rooms’ coefficient has grown to 0.00281, ‘Age of sale’ decreased to 0.00417, ‘Restaurant’ to 0.331, ‘Spa’ increased to 0.414 and ‘All suites’ to 0.656 and ‘Meeting space’ to 0.273. Note that the variable ‘Meeting space’ was not significant in regression I. For ‘Transaction specifics’, coefficients for ‘Portfolio deal’ and ‘Management change’ changed to 0.472 and 0.332 respectively. ‘Constant’ has decreased from 14.03 to 13.67, which makes hotel real estate pricing in regression analysis two more dependent on the independent variables, compared to regression analysis one.

‘Guest experience index’ that was used as indicator for quality in regression I and II is a constructed variable, also including location aspects. To find in whether locational quality is important in contrast to the actual hotel quality that hotel managers can influence, different indicators for quality are used in regression analysis III. Olery provided several quality indicators that can be used instead. In the correlation matrix however, high correlation coefficients were found between many of these variables. Therefore, guest perceived quality indicators of ‘Location’, and ‘Cleanliness’ are used as indicators of quality in regression analyses III and IV. As ‘Cleanliness’ highly correlates with ‘Value’, ‘Service’ and ‘Condition’, significant results for ‘Cleanliness’ are expected to also contain perception of quality in all of these variables. In regression III, the variable ‘Cleanliness’ has a coefficient of 0.156 and is significant at the 95% confidence level. Quality indicator ‘Location’, is significant within the 99% confidence interval with a coefficient of 0.234. As both indicators for quality are significant, this indicates that as well ‘Location’ as ‘Cleanliness’ (which is highly correlated with ‘Value’, ‘Service’ and ‘Condition’) influence hotel real estate prices. Also, ‘Guest Experenience Index’ correlates highly with ‘Cleanliness’, and could therefore not be included to regression III.

When comparing regression analysis III with regression analysis II, the variables ‘Upscale class’ and ‘Economy class’ lose their significance in regression analysis three. Apparently, the variation that was explained by these variables in regression one is explained by quality indicators ‘Location’ and ‘Cleanliness.

The coefficients for classification indicators for ‘Luxury’ and ‘Upper Upscale class’ have decreased from 1.029 to 0.772 and from 0.677 to 0.443, respectively. Also, the coefficient for ‘Number of rooms’ has decreased slightly to 0.00265, ‘Age at sale’ to 0.00314, ‘Spa’ to 0.402, ‘Meeting space’ to 0.264, ‘Portfolio deal’ to 0.440 and ‘Management change’ to 0.266. The coefficients of the variables ‘Restaurant’ and ‘All suites’ however, increased to respectively 0.377 and 0.669. In addition, the variable ‘Buyback’ is significant in regression analysis III, and has a negative coefficient. This indicates that when a management company buys the hotel property they are currently renting, the property is acquired with a discount.

The adjusted R-squared ratio increases in each consecutive regression analysis. Regression analysis I has an adjusted squared ratio of 0.626. After adding cross-city effects in regression analysis II, the adjusted R-squared ratio increases to 0.655. In regression analysis III, a different indicator for quality is used as dependent variable. This results in a increase of the adjusted R-squared ratio to 0.688. Of all the three regression models, regression analysis III explains the highest percentage of variation. Therefore, regression analysis III is seen as the most optimal model in this thesis, and will be used to base the conclusion on.

Generally, the results in regression analysis III are in accordance with the results from previous studies. Hotels in classification ‘Upscale class’, ‘Upper midscale class’, ‘Midscale class’ and ‘Economy class’ are not significantly different from each other. As a higher classification is connected to higher ADR, this could indicate that the NOP between the insignificant classifications does not differ much, despite difference in ADR. Hotels in lower classifications might have fewer costs than hotels in a higher classification, leading to a higher NOP. This might be the reason that hotels with classification ‘Upscale class’, ‘Upper midscale class’, ‘Midscale class’ and ‘Economy class’ are not significantly different from each other.

All hotel fundamentals are significant, except for ‘Convention’ and ‘Boutique’. As most hotel fundamentals can provide additional income for hotels, the results are generally as expected. ‘Convention’ and ‘Boutique’ might be associated with higher costs, and are therefore not significantly influencing real estate pricing. The type of management is insignificant, indicating that the quality of the management is most likely to be more important than the type of management. Within ‘Transaction specifics’, the coefficients for the variables ‘Portfolio deal’ and ‘Management change’ are significant and positive, indicating that hotels that are sold within a portfolio sale, and hotels whose management is changed after the transaction, are sold with a premium. This result was not expected for ‘Portfolio deal’, but for ‘Management change’ it is likely that investors pay a premium if they think they can generate more income after changing the hotel management. ‘Buyback’ is associated with a discount on hotel pricing. This might be caused be a better bargaining positions of the management company who buys the property.

The significant result for ‘Guest experience index’ in regression analysis two indicates that the quality of a hotel as perceived by the guests is significantly influencing hotel real estate pricing. The significant result for quality implies that hotel quality is taken into account when investors acquire hotel real estate. It is not certain whether investors actually look at customer perceived value, as it could also be the case that the

‘Guest experience index’ is a good indicator for the overall hotel quality. ‘Guest Experience Index’ also measures quality of the location to some extent. Quality of the location is not something that hotel manager can influence easily. Therefore, quality indicators ‘Location’ and ‘Cleanliness’ are used instead of ‘Guest Experience Index’ to find whether the quality that is accountable to the hotel management is also significantly influencing hotel pricing. ‘Cleanliness’ is significantly influencing hotel real estate prices and is correlated with ‘Room’, ‘Value’, ‘Service’ and ‘Condition’. These indicators for quality are influenceable by the hotel management. Hotel real estate pricing therefore, can be influenced by hotel quality.

To test for multicollinearity, a VIF test is displayed in Table 9. For each variable within each regression analysis, a value is shown. If the values do not exceed 10, the model is not likely to suffer multicollinearity. In regression analyses I, II and III, the VIF values do not exceed 10. It is unlikely that the regression analyses have multicollinearity.

Table 5: Regression results

Regression nr. I II III Title Re gr es si on wi th ou t c ro ss -ci ty ef fect s an d G E I as in dic ato r f or qua lit y Re gr es si on wi th cro ss -ci ty ef fect s an d G E I as in dic ato r f or qua lit y Re gr es si on wi th cro ss -ci ty ef fect s an d ‘L oc at io n’ an d ‘C le an lin es s; as in dic ato rs f or qua lit y

Dependent variable log(Price) log(Price) log(Price)

Qu al ity Location 0.234*** (-0.0479) Cleanliness 0.156** (-0.0619) Gues Experience Index 0.0164** 0.0190***

(-0.00644) (-0.006)

Loc

at

ion

Distance to cities' major transport hub 0.0132** -0.00377 0.000776

(-0.00631) (-0.0122) (-0.0119)

Distance to closest airport 0.0244*** 0.0359** 0.0354***

(-0.00817) (-0.0138) (-0.0135) Cl as si fic at io n Luxury class 1.156*** 1.029*** 0.772*** (-0.235) (-0.239) (-0.219)

Upper upscale class 0.757*** 0.677*** 0.443**

(-0.18) (-0.195) (-0.185)

Upscale class 0.292* 0.302* 0.209

(-0.163) (-0.181) (-0.165)

Upper midscale class

Midscale class 0.243 0.355 0.332 (-0.233) (-0.269) (-0.24) Economy class 0.474* 0.542** 0.417 (-0.256) (-0.272) (-0.256) Ho te l f un da me nt al s Number of rooms 0.00264*** 0.00281*** 0.00265*** (-0.000539) (-0.000645) (-0.000599) Age at sale 0.00475*** 0.00417*** 0.00314*** (-0.000969) (-0.00083) (-0.000819)

Dummy missing age 0.125 0.0775 -0.0277

(-0.212) (-0.203) (-0.186)

Restaurant 0.359** 0.331* 0.377**

(-0.758) (-0.73) (-0.705) Spa 0.369** 0.414** 0.402** (-0.168) (-0.166) (-0.161) Boutique 0.358 0.296 0.338 (-0.222) (-0.232) (-0.216) All Suites 0.528** 0.656*** 0.669*** (-0.211) (-0.207) (-0.206) Meeting space 0.171 0.273** 0.264** (-0.143) (-0.137) (-0.133) Ma na ge m en t Independent management -0.231 -0.209 -0.138 (-0.182) (-0.207) (-0.202) Franchise management Chain management 0.0391 -0.0489 -0.0186 (-0.159) (-0.184) (-0.172) Tr an sa ct io n sp ec ifi cs Development 0.161 0.339 0.143 (-0.229) (-0.216) (-0.214) Portfolio deal 0.412*** 0.472*** 0.440*** (-0.13) (-0.124) (-0.118) Management change 0.335* 0.332** 0.266* (-0.171) (-0.159) (-0.142)

Sale & lease-back -0.197 -0.14 -0.235

(-0.158) (-0.17) (-0.166) Buyback -0.404 -0.422 -0.496** (-0.29) (-0.263) (-0.214) Constant 14.03*** 13.67*** 12.24*** (-0.66) (-0.634) (-0.72) Adjusted R-squared 0.626 0.655 0.688

Fixed city effects No Yes Yes

Year of sale Yes Yes Yes

N 180 180 180

5. Limitations and future research

A first limilation is that the number of observations in this study is rather low, which might have resulted in fewer significant variables. Repeating this study with a larger number of observations helps in increasing the explanatory power of the model. Also, as cross-city effects did not improve the explanatory value of the model by much, but it did change the estimated effect of hotel quality.

As the data sources did not always provide month of sale for the hotel real estate transactions, it was not possible to track monthly changes in online guest reviews and hotel real estate pricing. Using monthly or quarterly data might be useful to find more significant relations between for example a management change or owner change, and hotel quality as perceived by the guests. Having monthly or quarterly data instead of yearly data might also help in increasing the significance of the ‘Guest experience index’.

After finding that online guest reviews significantly influence hotel pricing, new research should focus on these online reviews and how they are composed. If the online reviews for example are influenced primarily by location, the hotel manager cannot change much to increase guest perceived quality. If the online reviews on the other hand are influenced by staff attentiveness or cleanliness, the hotel manager can influence quality more easily.

Another aspect that should be included in future research is the capital expenditures that are associated with increasing quality. If for example quality is largely influenced by cleanliness and condition of the building, it is interesting to know whether the costs that are associated with properly maintaining hotel real estate weigh up against the benefits.

6. Conclusion

There are several studies that find which variables influence hotel real estate pricing. Not all of the studies take location characteristics into account. To our knowledge, none of the studies took hotel quality into account. In this thesis, different locational characteristics such as distance to airport, and guest perceived quality of location are used to find what the impact of location is on hotel real estate pricing. Also, perceived quality, based on online guest reviews is taken into account. Hotel quality is often linked to classification systems, which would be easy to measure. Classification systems however, differ a lot per country and are not comparable with each other. Also, classification systems do not measure performance. Therefore, guest perceived quality is used as indicator for quality. This study aims on finding whether cross-city effects and hotel quality help to explain hotel real estate prices, in addition to the variables that were found to be of significant importance in previous studies.

The previous studies used variables in the categories ‘Hotel fundamentals’, ‘Location characteristics’, ‘Income characteristics’, ‘Classification’, ‘Transaction specifics’, ‘Management’ and ‘Year of sale’ for their analyses. The ‘Location characteristics’ that were used mainly consist out if ‘State fixed effects’ and distances to transport hubs and airports. This study included cross-city effects for eight large tourist destinations in Europe.

Next to location, this thesis also addresses hotel quality. There are several issues when including hotel quality in the data analyses. Quality is influenced by many different aspects, like image, service, attentiveness, facilities and price, which is not easy to measure. Online guest reviews are ought to capture most of the factors that influence hotel quality, according to previous studies. Therefore, these online reviews are used as measure for quality.

210 Hotel real estate transactions from 2010 to 2015 are regressed on cross-city effects, hotel quality and all other control variables that were found important in previous studies. The transaction data is gathered from PwC, HVS and RCA, and data for hotel characteristics is provided by STR. Data of online guest reviews is provided by Olery.

Cross-city effects, citrus paribus on hedonic and quality aspects of hotels, do not have a very large explanatory value when explaining hotel real estate value. Hotel quality, and in particular quality as perceived by guests, is significantly influencing hotel real estate pricing. An increase of 1 in Olery’s ‘Guest experience index’ (ranked 0-100), hotel real estate value is increasing with 1.90%. The ‘Guest experience index’ also contains locational quality, which is not influenceable by hotel managers. When using online ratings for ‘Location’ and ‘Cleanliness’ instead of the ‘Guest experience index’, both variables significantly influence hotel real estate prices. This implies that not only location is of importance when investing in hotels, as ‘Cleanliness’ is also important. Because ‘Cleanliness’ is highly correlated with other online guest reviews for ‘Room’, ‘Value’, ‘Service’ and ‘Condition’, this indicates that hotel managers can increase hotel quality by improving these aspects. ‘Location’ and ‘Cleanliness’ have an effect of 23.4% and 15.6% on

hotel prices, for every increase of 1 on a 10-point scale. When a hotel would success in improve their quality indicator ‘Cleanliness’ from 6.1 to 8.1 for example, the hotel would increase 31.2% in value, holding all other variables constant. This is a large increase and should therefore be considered by hotel investors.

The results indicate that online hotel reviews are a good measure of hotel quality and that hotel quality was previously taken into account by hotel investors. It also indicates that investors can increase future property value by increasing hotel quality. It is therefore useful for investors to pay close attention to the online guest reviews when acquiring a hotel property. It is also useful for investors to keep track of hotel quality when managing hotel real estate assets. Next to the financial advantages, including online guest reviews might also help determining the type of investment (core, value-add), as it can be management intensive to improve the rating of online guest reviews. Generally, improving hotel quality or guest perceived quality is associated with costly measurements. By creating a cost-benefit overview of increasing quality, hotel investors can determine whether it is feasible to improve quality with respect to hotel real estate value.

Reference list

Blal, I. and Graf, N. (2013). The discount effect of non-normative physical characteristics on the price of

lodging properties. International journal of hospitality management, 34, 413-422. DOI:

10.1016/j.ijhm.2012.12.003

Chan, E. and Wong, S. (2005). Hotel selection: when price is not the issue. Journal of vacation marketing,

12(2), 142-159. DOI: 10.1177/1356766706062154

Claver, E., Tarí, J. and Pereira, J. (2006). Does quality impact on hotel performance?. International journal of

contemporary hospitality management, 18(4), 350-358. DOI: 10.1108/09596110610665357

Corgel, J. (2007). Technological change as reflected in hotel property prices. Journal of real estate finance and

economics, 34, 257-279. DOI: 10.1007/s11146-007-9011-4

Corgel, J. and deRoos, J. (1992). Pure price changes of lodging properties. Cornell hotel and restaurant

administration quarterly, 33(2), 70-77.

Corgel, J. and deRoos, J. (1993). The ADR Rule-of-thumb as predictor of lodging property values. Retrieved on April 5 2015, from http://scholarship.sha.cornell.edu/articles/649/

Corgel, J., Liu, C. and White, R. (2015). Determinants of hotel property prices. Journal of real estate finance

and economics, 1(25), pages unknown. DOI: 10.1007/s11146-015-9494-3

Costa, J. (2004). The Portfugese tourism sector: key challenges for human resources management.

International journal of contemporary hospitality management, 16(7), 402-407.

DeRoos, J., and Rushmore, S. (1993). Hotel Valuation Techniques. Retreived on April 4 2015, from http://www.hvs.com/Bookstore/HotelValuationTechniques.pdf

Gool, P. van, Brounen, D., Jager, P. and Weisz, R. (2007). Onroerend goed als belegging (4th _edition).

Groningen: Noordhof uitgevers

Hoag, J. (1980). Toward indices of real estate values and returns. Journal of Finance, 35(2), 569–580.

Ingram, H. and Daskalakis, H. (1999). Measuring quality gaps in hotels: the case of Crete. International

journal of contemporary hospitality management, 11(1), 24-30.

JLL. (2015). Hotel investment outlook 2015. Retrieved from:

http://www.jll.com/Research/JLL%20Hotel%20Investment%20Outlook%202015.pdf?27a8225c-9969-4c4b-b8ed-0c1f73858e0f

Jud, G. and Winkler, D. (1995). The capitalization rate of commercial properties and market returns.

Journal of Real Estate Research, 10(5), 509–518.

Matilla, A. and O’Neill, J. (2003). Relationships between hotel room pricing, occupancy and guest satisfaction. Journal of hospitality & tourism research, 27(3), 328-341. DOI: 10.1177/1096348003252361