Statistical Analysis for Business
Using statistical analysis to make informed decisions is an essential organizational practice that helps reduce bias and systematic error, promoting better risk management and increased prospects for market growth. Statistics for business can be applied at various levels, depending on the depth required and the desired outcomes based on the questions posed (Black, 2023). One way to apply statistics to decision-making is through predictive models, which help estimate the expected values of organizational parameters. Every manager wants to bring certainty and clarity to the foreseeable future.
In this assignment, predictive models are proposed for Airbnb, a large and well-known travel aggregator of accommodation bookings. The basic concept of Airbnb is that anyone can provide a part of their home or entire property for short-term rental to customers and receive money in return — the company receives a commission from this transaction, which is its business model (Lee & Kim, 2019). The company’s market success requires special attention and can be studied through financial analysis.
Company Information
History
Airbnb has a comparatively short history, but in a short time, it has gained high popularity among travelers. According to official information, Airbnb began its journey in 2007 from the personal homes of the founders, who decided to create a project to make money and connect people (Airbnb, 2023). By June 2011, the company had expanded globally by opening its first office outside the United States, specifically in Germany. By this point, Airbnb had already established its processing platforms, rebranded itself, and had a minimal U.S. customer base.
Throughout its history, the company has worked extensively with nonprofit and volunteer organizations. During Hurricane Sandy, Airbnb organized a disaster relief tool for those in need. In 2010, the company entered into a multi-year partnership with the International Olympic Committee to create opportunities for many tourists traveling to the sporting event (Dolnicar, 2021).
In response to the deleterious effects of the COVID-19 pandemic, the company has taken several measures to help preserve earnings for the Hosts and retain audiences. These measures include a protocol for improved room cleaning, creating an online experience platform for Hosts who have lost their earnings (Airbnb, 2023). Currently, the company is actively developing the technological capabilities of the service and offering enhanced services for both Hosts and Guests to expand market share meaningfully.
Organization
The discussion of Airbnb’s organizational aspects deserves special attention. According to Yahoo, the company employs over 6.8 thousand full-time workers (YF, 2023). As of December 2020, the company is publicly traded, so the necessary financial information can be found in public sources (Airbnb, 2023). Specifically, Airbnb’s top administrative staff is represented by ten directors, each responsible for a different area of organizational development. As of September 2023, the market capitalization of the brand is estimated at $84.71 billion, and its profitability ratio for commercial operations is 25.31% (YF, 2023).
Finance
In terms of dynamic capital structure, as shown in Figure 1, revenue profit has increased post-pandemic over the last two years, which can be attributed to both effective financial management and increased demand for tourism services following the pandemic (Ranasinghe et al., 2020). It can also be observed that the company’s operating expenses did not exhibit a consistent trend during the four years of observation, instead fluctuating periodically, which may indicate Airbnb’s adaptation to the changing market landscape and its measures to retain existing customers and attract new ones. Notably, the company’s total assets showed growth throughout the period, which is a favorable pattern that increases the company’s total capital and illustrates expansion as part of its organizational strategy. At the same time, the total debt remained practically unchanged, which, together with other data, indicates competent financial management and successful dynamics.
Competitors
The company has many competitors, which demonstrates the sustainability of brand growth. From its early days, it is essential to understand that Airbnb offers travelers an alternative experience, an escape from traditional hotel service, and a deeper integration into the culture of local communities (Andreu et al., 2020). From this perspective, hotel businesses are not direct competitors because they serve different segments. However, in reality, the decision of where to stay is often made by tourists based on budget, geographical, and personal considerations. Hence, price and offer comparisons between hotels and Airbnb take place.
Booking.com is one of the primary alternatives to Airbnb, serving as a similar aggregator that initially focused on hotel bookings but has since expanded to include guesthouses, campgrounds, glamping sites, bungalows, and other experiential options (Santos et al., 2020). Another iconic competitor is TripAdvisor, a platform that collects traveler reviews and experiences, with a real estate booking option (Zervas et al., 2021). The company has numerous guest reviews for each location, allowing customers to choose a place in more detail; however, there is no guarantee that the published reviews are genuine.
The number of similar competitors is constantly increasing, particularly during the post-pandemic period of heightened travel demand. Still, Airbnb has managed to monopolize up to 20% of the market and occupy a sustainable niche that combines high brand loyalty, the allure of cultural integrations, and an enjoyable experience for tourists (Cuofano, 2023). Thus, in terms of competitive analysis, Airbnb has a reasonably solid market position, and, combined with the financial analysis outcomes, the brand is making active efforts to maintain its market share.
Application of Statistical Analysis
The proposed statistical analysis aims to profoundly investigate the relationships influencing Airbnb’s success in the Melbourne market. The results will enable the construction of working predictive models to help predict useful metrics for the brand based on management performance. Airbnb is thus a key beneficiary of this analysis, as the results will have increased practical value in improving operational management.
On the other hand, improving brand performance in the Melbourne real estate market is expected to enhance travelers’ experiences, as the company will be able to implement improvement strategies based on the data, thereby improving the guest experience. Hosts, as company partners, will also benefit from these strategies, as they can enhance the tourism appeal of the local travel destination and consequently improve their own revenue. Thus, it can be expected that the Melbourne real estate market will also benefit from this analysis as it develops and grows qualitatively.
Problem Identification
Finding options to improve operational management for Airbnb to occupy a more sustainable niche, attract more customers, create strong partnerships with local hosts, and increase commission revenue is an applied challenge for Airbnb. The problem addressed in this paper is to develop predictive regression models that explore the relationships between variables in the Airbnb rental market in Melbourne.
Specifically, the first predictive model proposes investigating the impact of characteristics such as the number of bedrooms, rating, being a Superhost, and profile picture on pricing, which will more accurately identify the factors directly affecting the final booking price. The second predictive model proposes utilizing the attributes of location, room block, superhost feature, profile picture, and pricing to examine their impact on the final rating.
With global travel demand rising and Airbnb’s continued desire to expand embedded in the company’s historical backdrop, it makes sense to explore new working methods. Currently, the brand has approximately 8,000 properties in Melbourne, with an average daily rate of 141 Australian dollars (approximately USD 90.54) (Houst, 2023). The authors also report that only slightly more than half of the homes can generate a monthly profit of AUD 3459 (USD 2221.09), which creates missed opportunities for other local properties that are not generating moderate to significant returns for various reasons.
There is also a problem in determining pricing and the factors that contribute to it. Figure 2 and Figure 3 show examples of six renditions for the same query (Melbourne, October 1-7, 1 guest). It can be seen that, although the price difference is not very significant, the accommodations are pretty different. This justifies the need for a more detailed understanding of pricing and property ranking, which will provide a solid basis for predicting a place’s price range and expected popularity.
Regression analysis will effectively address the problem under study because it seeks to identify relationships between continuous variables. This type of analysis is used when there is a need to determine the causality of relationships between variables, hence determining how one variable affects another (Al Shehhi & Karathanasopoulos, 2020). As the research problem is framed around determining the influencing factors of pricing and rating, multiple regression models will prove to be appropriate tools for analyzing the type and influence of the relationship and predicting the significance of the outcome when the parameter changes.
Literature Review
The problem of predictive models in business analytics is not uncommon in academic discourse. Authors have widely studied it, as the outcomes of such studies are of high practical value. The authors have repeatedly demonstrated that predictive modeling has significant beneficial effects on businesses, enabling informed organizational decisions and increasing profits, particularly in the tourism industry, which has experienced substantial growth in recent decades (Kohli et al., 2020; Li et al., 2021; Ghalehkhondabi et al., 2019).
Al Shehhi and Karathanasopoulos (2020) report that this industry is expected to grow qualitatively in the foreseeable future, with tourism’s current contribution to the global economy exceeding $7 billion. Given such a rapid increase and, therefore, growing competition, it is natural for managers to consider strategic options to expand and capture a larger market share. Thus, the tourism industry is experiencing significant market growth and is projected to continue increasing.
Appealing to predictive models is one way to facilitate informed organizational decision-making. Advanced statistical tools and big data have been demonstrated to facilitate a deeper understanding of market structure, the development of more nuanced strategies, and the creation of positive business opportunities (Kohli et al., 2020). The authors additionally report that forecasting is used to predict future values of the target variable using mathematical methods, which means this approach qualitatively reduces data bias and inhibits the influence of human error.
Real estate is an important area where forecasting is characterized by increased practical utility, as predictive models demonstrate good applicability to price data (Thamarai & Malarvizhi, 2020). Chen (2022) agrees with this, demonstrating that predictive regression models yield results that, considering fluctuations and errors, align well with actual price trends. In the field of forecasting, however, researchers pay more attention to more serious methods of analysis, whether it is the use of machine learning or artificial intelligence. Still, regardless of the process, each of the tools in some way turns to regression analysis to study the causal relationships between variables (Al Shehhi & Karathanasopoulos, 2020).
On the other hand, such methods are still emerging and require substantial investment in resources, including finance and computing power. Consequently, the most commonly used analysis tactics remain more conservative models, such as multiple regressions on datasets and time series (Li et al., 2021). The findings suggest that regression models are widely employed in academia to address forecasting problems, particularly in the tourism sector.
Critical discussion of regression analysis has also received much attention, as the results of such tests are aimed at solving applied business problems. Kohli et al. (2020) demonstrate that the regression model, although easy to use, offers improved accuracy compared to other modeling methods and enables more accurate predictions of future sales dynamics, a finding also supported by Thamarai and Malarvizhi (2020) and Al Shehhi and Karathanasopoulos (2020).
However, the method also has several limitations: the fundamental values of target variables can rarely be predicted with high accuracy, so it is often worthwhile to introduce some errors into the interpretation (Chen, 2022). Furthermore, a larger dataset is necessary for more accurate analysis, as the limitations of their application may lead to systematic errors and biases (Li et al., 2021). Ghalehkhondabi et al. (2019) also note that the seasonality of tourism demand and unpredictable external factors can significantly impact the accuracy of the analysis. Therefore, effective managers should always consider the analysis results in context.
The literature review also demonstrated a knowledge deficit in investigating price and rating variables in the context of predictive models for Australia. Published work has focused on market research in India and the US, but no relevant evidence was found for Melbourne (Thamarai, M., & Malarvizhi, 2020; Chen, 2022). This creates a gap in valuable knowledge, the solution to which is part of this academic study. The results of building predictive models for Airbnb will enable the verification of previous authors’ findings, as well as make a valuable contribution to understanding the tourism market structure in Melbourne.
Data Collection
An obvious advantage of Airbnb is the publication of official data that can be used for analysis. Data is published for each region where the platform is present, along with the corresponding time frame (IA, 2023). For Melbourne, listing data was posted for four periods: September 4, June 6, March 13, and December 10. All data were exported to an Excel spreadsheet and placed in four corresponding tabs. Most of the included variables were not of research value to the paper and were removed.
Thus, the sets represented distributions of the following variables: Superhost (dichotomous), Profile Pic (dichotomous), Number of Bedrooms (continuous), Price (continuous), and Rating (continuous). Recodings were performed for the Super Host and Profile Pic variables, where 0 indicates the absence of the parameter, and 1 indicates its presence. Additionally, outliers were identified using the IQR method, and all outlier values for Price and Number of Bedrooms were removed. After removing all rows with blank and redundant values, the final sample sizes for each dataset were determined and are shown in Table 1.
Table 1 — Data processing results of the four sets of data
Analysis and Discussion
After processing, all data sets were ready for regression analysis. The primary interest was to examine the descriptive patterns for the variables used. A table illustrating the collected results is shown below. It can be seen that the data measured at different time intervals were not significantly different from each other, indicating the relative constancy of the trends over time.
A total of ten regression models were created, two for each dataset, where either host rating (Model I) or reservation cost per day (Model II) was used as the dependent variable. All results of the analyses are shown in Appendices A-E. First, it makes sense to discuss the accuracy of the constructed models using the coefficient of determination (Table 3).
The reliability of the constructed models is relatively low, with a maximum explained variance of 30.5% for the second model in the first data set. A potential reason for the low coefficient of determination could be the high variance in the recall distributions, as this variable has not been cleaned. The rationale for not cleaning the variable is that the IQR method suggested leaving only scores of 4 and 5 as indicators for hosts, which would render the analysis meaningless. Residual plots shown in the attached MS Excel report indicate that the regression models are of low reliability when accounting for the nature of the variables. However, interestingly, each of the ten models was statistically significant, according to F-analysis (Appendices A-E), which means they can be used for prediction.
Table 3 — Coefficients of determination (R2) for all constructed models
Each regression model allowed us to formulate equations that capture the relationships between the variables. For all the first models, it can be seen that superhost status, price, and photo availability contribute to higher customer ratings as they grow. The number of bedrooms had a differential effect across periods. For all second models, it can be seen that each variable differentially affected the price per night depending on the time of observation. Superhost status, the number of bedrooms, and rating positively influenced the prices of most models.
Recommendations
This research work aimed to conduct a statistical analysis to identify causal relationships between predictors (potentially influencing factors) and daily prices or seller ratings. The analysis showed that having super host status and a profile picture positively affect rating and price (in most models). Increasing the number of bedrooms in most models had a negative effect on the rating and a positive impact on the daily price. In all cases, price had little to no effect on the rating, and the rating led to an increase in price.
In other words, the predictors exhibited differential effects on the target variables across models and different observation periods, although the trends described above were generally consistent. Based on the results, some conclusions can be drawn that will allow customers and hosts on the Airbnb platform to fine-tune their booking experience:
- For customers, to lower the booking price, customers can opt out of selecting super host status, reduce the number of bedrooms, and choose options with lower reviews. Having a photo of the host, in general, also increases the price.
- For clients, to find a host with a higher rating, you can choose hosts with super host status, with a photo, and with fewer bedrooms. However, the price will not affect the rating in any way.
- For hosts, to increase their rating on the platform, gaining super host status, posting a profile picture, and reducing the number of bedrooms were shown to be significant predictors. Setting a low or high price did not affect the host rating.
- For hosts, to sell their bookings at a higher price, hosts can work on obtaining super host status, increase the number of bedrooms in the room, and strive to get higher customer reviews.
References
Airbnb. (2023). About us. Airbnb.
Al Shehhi, M., & Karathanasopoulos, A. (2020). Forecasting hotel room prices in selected GCC cities using deep learning. Journal of Hospitality and Tourism Management, 42, 40-50.
Andreu, L., Bigne, E., Amaro, S., & Palomo, J. (2020). Airbnb research: An analysis in tourism and hospitality journals. International Journal of Culture, Tourism and Hospitality Research, 14(1), 2-20.
Black, K. (2023). Business statistics: For contemporary decision making. John Wiley & Sons.
Chen, Z. (2022). Prediction and analysis of house prices in Boston based on regression model [PDF document].
Cuofano, G. (2023). Airbnb competitors. Four Week MBA.
Dolnicar, S. (2021). Airbnb and events–before, during and after COVID-19. The University of Queensland.
Ghalehkhondabi, I., Ardjmand, E., Young, W. A., & Weckman, G. R. (2019). A review of demand forecasting models and methodological developments within tourism and passenger transportation industry. Journal of Tourism Futures, 5(1), 75-93.
Houst. (2023). Investing in Melbourne real estate: A guide to Airbnb laws and best rentals. The Insider @ Houst.
IA. (2023). Get the data. Inside Airbnb.
Kohli, S., Godwin, G. T., & Urolagin, S. (2020). Sales prediction using linear and KNN regression. Advances in Machine Learning and Computational Intelligence, Algorithms for Intelligent Systems, 321-329.
Lee, K. H., & Kim, D. (2019). A peer-to-peer (P2P) platform business model: The case of Airbnb. Service Business, 13(4), 647-669.
Li, X., Law, R., Xie, G., & Wang, S. (2021). Review of tourism forecasting research with internet data. Tourism Management, 83, 1-11.
Ranasinghe, R., Damunupola, A., Wijesundara, S., Karunarathna, C., Nawarathna, D., Gamage, S., & Idroos, A. A. (2020). Tourism after corona: Impacts of COVID 19 pandemic and way forward for tourism, hotel and mice industry in Sri Lanka [PDF document].
Santos, G., Mota, V. F., Benevenuto, F., & Silva, T. H. (2020). Neutrality may matter: Sentiment analysis in reviews of Airbnb, Booking, and Couchsurfing in Brazil and USA. Social Network Analysis and Mining, 10, 1-13.
Thamarai, M., & Malarvizhi, S. P. (2020). House price prediction modeling using machine learning. International Journal of Information Engineering & Electronic Business, 12(2), 15-20.
YF. (2023). Airbnb, Inc. (ABNB). Yahoo Finance.
Zervas, G., Proserpio, D., & Byers, J. W. (2021). A first look at online reputation on Airbnb, where every stay is above average. Marketing Letters, 32, 1-16.
Appendices
Appendix A
Appendix B
Appendix C
Appendix D
Appendix E
