The Emergence of the Sharing Economy: Data Analysis
to the Impact of Uber on the Traditional Taxi Industry
June 2015
Bachelor thesis Name: Liza de Koning Student number: 10334033 Study: Economics and Business Specialisation: Sharing Economy Supervisor: Ben Loerakker Coordinator: Marcel Boumans
Abstract
In the last few years a couple of very successful firms have been established in the sector of the sharing economy. Along with the rise of the sharing economy and its tensions the central question emerged whether the sharing economy is able to provide an alternative to the traditional markets? The goal of this paper is to state a proposal to examine to what extent Uber is capturing market share of the traditional taxi industry. An extension of the difference-in-differences analysis to the use of multiple groups and time periods is proposed to examine the impact of Uber on the traditional taxi revenues.
Statement of Originality
This document is written by Student Liza de Koning 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.
Content
1 Introduction 3
1.1 Contribution to the Sharing Economy 3
1.2 Focus of Research 4
1.3 Research Design 4
2 Literature Review 5
2.1 Emergence of the Sharing Economy 5
2.2 Activities of Uber 5
2.3 Entrance of Uber in the Taxi Industry 6
2.4 Summary of Related Studies 8
2.5 Determination of a Research Strategy 9
3 Methodology 11
3.1 Difference-in-Differences Regression 11
3.2 Assumptions 13
3.3 Empirical Approach 13
3.4 Data Collection 14
3.5 Graphical Analysis of the Taxi Industry of Australia 15
3.6 Other Empirical Approaches 17
4 Conclusion 18
4.1 Analysis of Possible Results 18
4.2 Limitations 19
4.3 Further Research 19
1
Introduction
During the last decade, the sharing economy is changing entire markets around. Professional businesses have to compete with consumers, who are sharing their durable goods with others in return for monetary compensation. During the last few years a couple of very successful firms have been established in the sector of the sharing economy. Airbnb and Uber offer consumers the opportunity to rent rooms in their houses to strangers and become taxi driver for a day. Regulations for companies in the sharing economy often do not exist yet and sharing companies do not follow regulations which were originally set up for the traditional companies. These sharing companies are capturing market share worldwide by avoiding regulations as paying taxes. Traditional companies label this as unfair competition while sharing companies claim that they evolve a new market of which consumers can benefit. Along with the rise of the sharing economy and its tensions, the central question emerged whether the sharing economy is able to provide an alternative to the traditional markets?
1.1 Contribution to the Sharing Economy
Only a few economic studies have examined the impact of sharing companies on the traditional markets. Rayle et al. (2014) did research to the similarities and differences of taxi and ride-sourcing users and concluded that 39% of the ride-sourcing customers questioned were otherwise using a taxi. This proves that ride-sourcing companies have a significant impact on traditional taxi companies. Another paper written by Zervas, Proserpio and Byers (2015) did research to Airbnb and its impact on the hotel industry. Airbnb is the counterpart of Uber in sharing apartments. They found that each 10% increase in Airbnb supply results in a 0.35% decrease in monthly hotel room revenue.
To contribute to the question whether sharing companies are able to provide an alternative to the traditional markets, it is important to outline to what extent sharing companies are capturing market share of traditional companies and to what extent they serve a new market. The goal of the paper is to examine to what extent Uber affects the traditional taxi industry.
1.2 Focus of Research
There is much debate in the academic world in defining services which are offered by ride service companies like Uber, because the emergence of these services is very recent. The California Public Utilities Commission (CPUC) introduced the term Transportation Network Companies (TNCs) in September 2013 to define these companies. But definitions like ride-sharing, peer-to-peer taxi services and ride-sourcing are also very common in the academic literature. In this paper the definition ride-sourcing is used to refer to the service of ‘regular’ drivers who are offering a ride to consumers in motivation to earn an income. In general the driver and consumer do not share the same destination. This definition is introduced by Rayle, Shaheen, Chan, Dai and Cervero (2014) and is adopted to avoid confusions in the difference between ride-sharing and ride-sourcing. The services which companies like Uber and Lyft deliver are defined as ride-sourcing and services in which consumers are matched who are undertaking the same ride are defined as ride-sharing. Lyft is perceived to be the most important ride-sourcing competitor of Uber, but Uber is dominating Lyft in terms of revenue, riders, revenue per rider, and absolute growth rates (Nicholson, 2014).
This paper composes a proposal to examine the economic impact of the ride-sourcing company Uber on the traditional taxi industry. This research proposal is able to exert once Uber goes public. An IPO obligates Uber to reveal their financial details and support details of data which are needed to exert this research proposal.
1.3 Research Design
This paper starts describing the sharing economy in general and the activities of Uber and its entrance in the taxi industry. Then, related studies are described about the impact of companies which are active in the sector of the sharing economy and which analyses are used. After a description of related markets, the focus will be moved to an executable analysis to examine to what extent Uber affect the traditional taxi industry. Subsequently is decided that a difference-in-differences (DD) analysis will be used to outline the impact of the activity of Uber on the traditional taxi revenues. A DD analysis is capable of identifying a causal relationship between the entrance of Uber and eventual reduced taxi revenues. If in the end no causal relationship is identified, the research is still assumed to be a useful contribution by discovering that Uber has no impact on the taxi industry.
2
Literature Review
Despite the fact of extensive growth in the sharing economy, only a few economic studies have examined the impact of sharing companies on the traditional markets. A lot of articles and newspaper reports are to be found on the internet, but the academic literature is falling behind. Some studies which examines sharing companies and its impact are found and discussed. Before the relevant literature is discussed, the emergence of the sharing economy, the activities of Uber and its differences from the traditional taxi industry are outlined.
2.1 Emergence of the Sharing Economy
A couple of sharing companies are growing excessively and thereby influencing incumbent firms. These incumbent firms are restricted by regulations of the government and protest against these sharing companies. Regulations for companies in the sharing economy often do not exist yet. Besides that sharing companies often do not follow regulations which are originally set up for the traditional companies. This might provide unfair competition and a risk of the return of the externalities where the regulations initially were set-up for.
On the other side, technological development enabled sharing companies to use resources more efficiently and compete with regulated companies by innovation. Sharing companies use underutilized resources of consumers more efficiently. Therefore they increase competition in the markets by offering lower prices and more choice to consumers. Consumers benefit from these innovations offered by sharing companies.
These innovations may disrupt traditional markets and a policy change might be needed. Rauch and Schleicher (2015) assume that cities will forbid sharing firms to exert their activities or will spare them of regulations. If the sharing firms survive these current fights, they will exist in the long term and will be mostly free from regulations.
2.2 Activities of Uber
Uber is launched in 2010 in San Francisco. In the meanwhile, Uber became very successful by gaining revenues in San Francisco of 500 million dollar per year, while the traditional taxi industry in San Francisco is about 140 million dollar per year according to an article of Business Insider UK (2015). The app of Uber is able to detect your location by using the GPS of your
smartphone. It matches your request for a ride with the nearest driver, who will pick you up in a couple of minutes. The app is very sophisticated in delivering all kinds of information. After the match is made, you are informed about the driver’s name, his reviews, car details and a text is received when the driver arrives. Eventually the trip is charged at your credit card, so no cash is needed. Uber offers a wide variety of services to their customers by offering a choice in multiple types of cars to their consumer.
Uber claims to only connect customers to drivers and their cars. Uber does not own any vehicles and drivers are, according to Uber, independent contractors. There are some discussions about whether drivers should be employees, which resulted in a number of lawsuits. In June 2015 the first driver in California is classified as an employee instead of contractor by the California Labor Commission, but this only applies to one driver. These days lawyer Shannon Liss-Riordan is setting up a case as a representative of these Uber-drivers in California to be classified as employees (2015). By providing just the service of connecting customers to drivers, Uber is able to minimize its costs and convey these costs to their drivers. Uber earns its revenues by cutting a 10-20% from each ride.
As opposed to the traditional taxi services, Uber makes use of a dynamic pricing system. This implies that prices react at the degree of demand, which creates high prices in rush hours. Another difference in the delivered services is that customers and drivers of Uber are able to rate each other and decline a match if one is not satisfied with the match. This results for drivers in a need to satisfy their customers in the best way possible and receive positive feedback.
2.3 Entrance of Uber in the Taxi Industry
So Uber has differentiated its services from the traditional taxi services and has so far entered the taxi industry in 58 countries. A big reason Uber has grown so quickly is that Uber is not regulated in the same ways as the traditional taxi services. As mentioned by Koopman, Mitchell and Thierer (2014) regulations in the taxi industry are introduced in the beginning of the 20th century to protect consumers from externalities of the market. The entry into the taxi
industry, taxicab fares, services and quality of the rides are regulated to protect consumers and drivers from unfair competition, information asymmetries, price gouging and unequal bargaining power. Examples of these regulations are obligated insurances, training for drivers,
licenses and the obligation to paint taxicabs in matching colour patterns. Furthermore, customers are equally divided over drivers by taxi companies. Customers at train stations and airports can for example only make use of the taxi in front of the row of the waiting drivers. These regulations restrict the competition across taxi companies and individual taxi drivers and therefore the innovation in the taxi industry.
The sharing economy solves part of these information asymmetries in a very natural way. They compete with the taxi industry by minimizing costs and by differentiating their services. They expand the range of options and information available to customers. The technological development of the past decade makes it possible for Uber to overcome a lot of these externalities. Customers are provided with feedback and ratings of drivers by ex-customers and a natural selection in use of drivers eliminates drivers with negative feedback. Drivers should pass the inspection of customers. Therefore drivers are in general more motivated to satisfy the needs of the customers instead of just completing the ride. This supports an increase in quality. Uber did also use recent technological development to shorten the waiting time of customers to get a taxi. The research team of Uber pointed out an interesting fact following out research of their own data. Conor Myhrvold (2015), an internal science data analyst of Uber, mentioned in a blog that “The longer Uber has been in a city, the less willing to wait for a car everyone becomes” (para. 16). Moreover Rayle et al. (2014) conclude that ride-sourcing offers consistently shorter waiting times. In combination these findings would suggest that Uber and traditional taxi services are not able to simultaneously provide their services as long as the response time of traditional taxis does not improve.
This elaboration of regulation by Uber has two obvious effects. The first effect is that a customers’ value enhancing approach is created by the possibility of differentiating their services and natural selection by customers. This would result in higher quality. The second effect is that drivers are more liberated in delivering services, which possibly decreases safety and quality of the services. The total effect on the safety and therefore quality of the ride is ambiguous.
In the meanwhile, taxis are still limited by regulations and worried about whether their customers are moving towards Uber. In other words, they are worried about whether consumers conceive the ride services of Uber as a substitute of the traditional taxi services. The quality difference between the services of Uber and traditional taxis are ambiguous. That
is why it is tested whether Uber is perceived to be a substitute to services of traditional taxi, by measuring the impact of the entrance of Uber on the traditional taxi industry.
2.4 Summary of Related Studies
It is assumed that the car-sharing market has some interfaces with the ride-sourcing market, because both operate in the transport industry. A couple of studies look at the effects of the upcoming popularity of car-sharing. Steininger, Vogl and Zettl (1996) proved a significant decrease in aggregate private vehicle mileage and a decrease in private car-ownership due to car-sharing. Steininger et al. (1996) did research to the market segment size of car-sharing and to the individual behavioural impacts. The latter is examined by a controlled experiment in which members of car-sharing organisations keep track of a mobility diary one week before becoming a member and one week after and had taken part of a questionnaire survey. This questionnaire survey differs from other questionnaires among car-sharing organisations, because this survey was not based on memories.
Only one academic paper is found which focuses on the ride-sourcing market in specific. Rayle et al. investigated in their paper (2014) the similarities and differences of taxi and ride-sourcing users and the impact of ride-sourcing on vehicle travel miles travelled. They conclude by exercising an intercept survey that ride-sourcing customers who would have otherwise taken a taxi prefer the convenience, namely “the ease of call and payment and perceived shorter wait times” (Rayle, Shaheen, Chan, Dai, & Cervero, 2014, p. 13). Following the results of the same paper, it is concluded that ride-sourcing customers own fewer vehicles than taxi users, but this might be a result of the sampling method instead of a valid causal relationship. Out of the results also follows an important result to this study. 39% of the ride-sourcing customers questioned were otherwise using a taxi. This would point out a significant impact of ride-sourcing companies on traditional taxis.
Zervas et al. (2015) provides a sophisticated empirical research to the impact of Airbnb on the hotel industry. He estimates that each 10% increase in Airbnb supply results in a 0.35% decrease in monthly hotel room revenue by exercising a DD analysis. This study is among the first to provide sound empirical evidence of a significant change in consumer behaviour caused by a company which is active in the sharing economy.
2.5 Determination of a Research Strategy
To support the aim of this research, examining the impact of ride-sourcing companies on the traditional taxi industry, a research strategy is chosen. A survey research, experiment, grounded theory approach, case study and desk research are taken into consideration, which are also mentioned by Verschuren and Doorewaard (2007).
According to Verschuren and Doorewaard (2007) choosing the most supportive research strategy requires three choice between a broad or profound, quantitative or qualitative and empirical or desk research. Because the sharing economy is a recent phenomenon and little literature is available it is chosen to limit the research to a very specific subject which is investigated thoroughly. Because the aim of this research is to examine the impact of an economic change it is chosen to do a quantitative research. Therefore a survey research, experiment or desk research based on secondary data are capable of supporting this thorough quantitative research.
Furthermore it is chosen in which way data will be collected. Here arises a problem which could not be solved yet. Collecting data myself by setting up an experiment would be very time-consuming work, while I only possess over limited time. On the other hand, there is no secondary data available which could be used to exercise a quantitative research towards the impact of a sharing company to traditional markets. Moreover, no supportive reactions were received at sharing data by governance or private companies. As mentioned by Graham (2014), in case of Uber this might be caused by the defence strategy of Uber in not revealing details about their revenues, assets, liabilities, ownership and obligation to pay taxes. Uber is owned by a Dutch partnership which is located in Bermuda. Companies in Bermuda are not obligated to share details about their revenues, assets, liabilities, ownership, and taxation. So Uber can perfectly legally keep their business details secret until an eventual IPO, which obligates companies to reveal their business details.
After the discovery of these restrictions a shift in mind-set is made. This thesis contributes most if a research proposal is made which is possible to exert in the future. If it is assumed that every piece of desirable data is available, a research in which the effect of a ride-sourcing company can be approached contributes most value to the academic knowledge of sharing companies.
The impact of a sharing company can be examined by exercising an experiment as used by Steininger, Vogl and Zettl in 1996. According to Verschuren and Doorewaard (2007) the advantage of an experiment is that the internal validity is very strong, but the disadvantage might be the external validity. A research to the impact of a ride-sourcing company in just one city is very reliable to that city, but the results are most likely not possible to generalize. Furthermore would an experiment which includes multiple cities be very time-consuming. Another research strategy is used by Rayle et al. (2014). They use an intercept survey in which participants filled in a questionnaire. As mentioned by Rayle et al. “Like all intercept surveys, this survey is not completely representative of the ride-sourcing market” (2014, p. 7). Finally Zervas et al. provide a research (2015) based on data which was collected from Airbnb and the Texas hotel industry. A desk research is used to estimate the impact of Airbnb on the hotel industry by comparing differences in hotel room revenue before and after Airbnb enters a city (the treatment group) against a baseline of differences in hotel room revenues which are not affected by Airbnb (the control group). It is chosen to do a similar research to examine the impact of Uber on the taxi industry.
3
Methodology
It is empirically examined whether the entrance of Uber in several cities have had an impact on the revenues of the traditional taxi industry. This is done by extending the DD analysis to multiple groups and use data of 2008-2014 of several cities in which Uber is active against a baseline of comparable control cities.
3.1 Difference-in-Differences Regression
In designing the empirical strategy, an important aspect is that there is no control over the randomization of the used data and that some differences might remain between the treatment and control groups. One solution adduced by Stock and Watson (2012) is to not compare the outcomes, but the change in outcomes of the pre and post period of the treatment. By doing this, differences in the pre-treatment values are eliminated and the focus has moved to the differences in differences. The DD estimator is one of the most used analyses to examine the impact of a treatment at a variable (Abadie, 2005). A DD analysis is able to approach and measure the impact of the entrance of Uber in the traditional taxi industry against a baseline of control cities. Because Uber experienced a very rapid growth across different cities (Blodget, 2014), these spatial and temporal variations are able to exploit in a study to the impact of Uber on the traditional markets. A comparable research is done to the rise and impact of Craigslist and Airbnb to their traditional markets (Kroft & Pope, 2014) (Zervas, Proserpio, & Byers, 2015). Stock and Watson (2012) explain that the DD strategy exists over a DD estimator, which is defined in the most basic form as followed:
𝛽̂1𝑑𝑖𝑓𝑓−𝑖𝑛−𝑑𝑖𝑓𝑓𝑠 = (𝑌̅𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡,𝑎𝑓𝑡𝑒𝑟− 𝑌̅𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡,𝑏𝑒𝑓𝑜𝑟𝑒)
− (𝑌̅𝑐𝑜𝑛𝑡𝑟𝑜𝑙,𝑎𝑓𝑡𝑒𝑟− 𝑌̅𝑐𝑜𝑛𝑡𝑟𝑜𝑙,𝑏𝑒𝑓𝑜𝑟𝑒)
= ∆𝑌̅𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡, − ∆𝑌̅𝑐𝑜𝑛𝑡𝑟𝑜𝑙
The DD estimator is an unbiased and consistent estimator of the causal effect if the treatment is randomly assigned. It measures the difference over the average change in Y in the treatment group and the average change of the control group. In regression form the DD estimator is included as followed:
∆𝑌𝑖 = 𝛽0+ 𝛽1𝑋𝑖 + 𝜇𝑖 (2)
In figure 1, a graphical analysis is made of the difference-in-differences estimator. The treatment will only affect the brown bar. The time-period in which both groups are not exposed to the treatment is called the pre-trend and is given by t=1 and the post-trend is given by t=2. The orange bars correspond to the average outcome of the revenues of the control cities. The brown bars correspond to the average real outcome of the treatment cities and the yellow bars correspond to the estimated outcomes of the treatment cities if they were not treated. The difference between the real outcome of the treatment cities and the estimated outcome if no treatment was present equals to the DD estimator.
Figure 1: The Difference-in-Differences Estimator
0 5 10 15 20 25 30 35 40 45 t=1 t=2
Control group Treatment group if treatment is absent Treatment group if treatment is present
3.2 Assumptions
There are some important assumptions which have to be satisfied. The key assumption made in using a DD analysis is the ‘common trend’ assumption. This assumption states that the treatment and control group should follow the same trends if no treatment was present in the treatment group. However, during the pre-trend the two groups could follow different trends. Besides that it is possible that the treatment group would not follow the post-trend of the control group if no treatment was present (Abadie, 2005). To check whether the treatment and control group follow an equal pre-trend can be solved easily by plotting the revenues of the two groups across time and show a graphical evidence, which is also exerted by Kroft and Pope (2014). However it is more difficult to test whether the treatment group would follow the trend of the control group if the treatment was absent in the treatment group.
Another assumption is made during this study is that Uber does not enter a city because the revenues of the traditional taxi companies in the specific city have dropped. Still it is possible that Uber enters a city right after a drop in traditional taxi revenues is noticed. This phenomenon is called the ‘Ashenfelter dip’ and would lead to an upward bias of the DD estimator of the effect of entrance of Uber. However, it is assumed to be unlikely that Uber’s extension strategy is based upon a drop in taxi revenues, because an entrance after a rise in revenues is more profitable. This is based upon the microeconomic theory that competitors will enter a market if a profit is made by the incumbent firms in the specific market. Uber’s strategy in its extension to new cities is however unknown.
3.3 Empirical Approach
The basic DD equation is extended to a regression which includes multiple groups and multiple periods which is typically used. Our DD equation takes the following form:
𝑇𝑎𝑥𝑖𝐶𝑜𝑚𝑝𝑅𝑒𝑣𝑖𝑐𝑡 = 𝐴𝑐+ 𝐵𝑡+ 𝛼𝑋𝑖𝑐𝑡+ 𝛽1𝑇𝑐𝑡 + 𝜀𝑖𝑐𝑡 (3) Here:
TaxiCompRevict = annual revenues of taxi company i in city c at time t
Ac = fixed effects for the city
Bt = fixed effects for the years
Tct = a dummy variable which equals 1 if Uber has affected city c at time t and equals 0
otherwise
εict = individual specific errors
The impact of Uber on traditional taxi revenues is measured by the OLS estimate 𝛽 . A DD 1 estimator is the average change in taxi revenues for those in the treatment group, the ones subjected to competition of Uber, minus the average of the change in revenues for those in the control group.
Multiple traditional taxi companies active in the same city are included in the regression. By doing this fixed effects for the city are allowed to include in the equation. These fixed effects can indicate the effect of regulations present in the specific city. Including multiple traditional taxi companies which are active in one city is assumed to not provide a significant problem. Take for example a look at Amsterdam in which ten official taxi operators are active.
To avoid serial correlation, time series information will be ignored. The data of before Uber entered the city and after entered the city will be averaged and equation 3 will be performed as if only two time-periods exist, before and after the entrance of Uber in the treated city. As noted by Bertrand, Duflo and Mullainathan (2002), this will only work if the treatment takes place in the treated cities at the same time. This means Uber should have entered the cities which are included in the regression at the same time. Because only annual data will be used in the regression, cities will be used in which Uber entered in 2012.
The downside of this regression analysis is that at least 50 cities should be included to obtain a credible small rejection rate of around 6% as obtained by Bertrand et al. (2002). They also did perform a regression analysis on 20 cities which yield a rejection rate of nearly twice as large, 9.5%.
3.4 Data Collection
A possible explanation for the gap in academic literature is the lack of data available to scientists. This lack of data might be caused by the newness of sharing companies and the entanglement of these companies with the obligation to start paying taxes. As mentioned before, the defensive strategy implemented by Uber withholds scientists of doing research to the impact of Uber. This might come to an end as an IPO of Uber is expected to happen very
soon (Benner, 2015). An IPO will obligate Uber to reveal their financial details, which creates the possibility to exert this research proposal. A disclosure of the business details of Uber will support finding details of the entrance of Uber and its extension strategy.
These details will assist in finding a relevant time frame to research. For now it is chosen to research the time frame of 2008-2014. Uber is founded in 2009 in San Francisco, started its activities in 2010 and extended its activities to other cities in the years after that. It is chosen to examine the cities in which Uber launched in 2012, so a time frame of at least two years of growth of Uber is available. Information is required about the exact entrance of Uber in a specific city to decide which cities should be included in the research. This information is relevant to indicate whether the data is obtained from the pre-period or post-period and to decide which cities should be included.
The most important data which must be collected are the revenues of the traditional taxi companies. Information must be collected of cities in which Uber entered and a comparable city which will function as its control city. The data of these taxi companies must be collected via national or regional companies, therefore this is very time-consuming work and it has not yet been collected.
3.5 Graphical Analysis of the Taxi Industry of Australia
Data of Australian Taxi Industry Association (ATIA) have been collected and been analysed (2015). ATIA shared taxi statistics of Australia over a time frame of 2004-2014 of i.e. average fare and number of annual jobs. Unfortunately the available data is about the states of Australia instead of the cities of Australia, this provides too little data to exert the regression which is explained in paragraph 3.3. There is also some data missing of the states Queensland, South-Australia, Western-Australia and Tasmania, therefore they are excluded from this analysis. Still a graphical analysis is made of two states in which Uber is active and two in which Uber is not active to gain insight into the evolvement of the traditional taxi industry. New South Wales (NSW) and Victoria (VIC) are states in which Uber is active. Uber launched in New South Wales in November 2012 and in Victoria in January 2013. Its annual taxi revenues are projected in figure 2. The two states in which Uber is not active are Australian Capital Territory (ACT) and Northern Territory (NT) and its annual taxi revenues are projected in figure 3.
Annual taxi revenues are calculated by a multiplication of the average fare and number of annual jobs (Australian Taxi Industry Association, 2015).
Figure 2: Annual Taxi Revenues of the states in Australia in which Uber is active
Figure 3: Annual Taxi Revenues of states in Australia in which Uber is not active
No significant conclusions can be drawn from both of these graphs about the growth of the taxi industry and an eventual impact of Uber on the taxi industry. It is expected that data of cities in specific gain a better insight in growth of the taxi industry and impact of Uber. This is expected because Uber is not active in all of the cities present in New South Wales and Victoria, but only in the biggest one or two cities. If the focus is moved from the whole state to a specific city an eventual impact of Uber is more precisely approached.
0 500 1000 1500 2000 2500 3000 2008 2009 2010 2011 2012 2013 2014 Taxi Re ve n u es , in m illi on s of d ollars Years NSW VIC 0 10 20 30 40 50 60 70 2008 2009 2010 2011 2012 2013 2014 Tax i Re ve n u es , in m ill ion s o f d o llars Years ACT NT
3.6 Other Empirical Approaches
To avoid reverse causality it is chosen to not include the intensity of the treatment in the regression, which is done by Zervas et al. (2015) and Kroft and Pope (2014). These papers are subjected to the possibility that cause and effect might be mixed up in the conclusions. There also might be a two-way causal relationship present in the regression which is not noticed. In our regression it might be possible that revenues in the taxi industry are dropping because of a recession, which causes that Uber is less active in the market. If this is the case the regression is subjected to reverse causality. There are identification checks which can be used to measure the robustness of the test. Zervas et al. (2015) did an analysis to the explanation of the variation of the regression and concludes that little variation remains unexplained. However, they are forced to make the assumption that unobserved trends which are time and city-specific and can influence both revenues of Airbnb and of the hotel industry are not present. Kroft and Pope (2014) did also an analysis to justify the causal relationship of their conclusions. They did two placebo tests of which the results support their finding of a causal effect of Craigslist on rental vacancy rates. Yet it is chosen to not include the intensity of the treatment, because of the uncertainty of the causal relationship between the intensity of the treatment and the dependent variable. Moreover the identification checks are too difficult to perform for a bachelor student.
Besides that it is chosen to not include panel data in the regression, because panel data possibly involves serial correlation. Bertrand et al. (2002) propose three solutions to this problem. One solution is to limit the use of data to two time periods, this solution is used during this research proposal. The other two solutions are based on the allowance for an arbitrary covariance structure over time across states and randomization inference by using placebo tests. Both of these techniques are too difficult to perform, so it is chosen to use the first solution.
4
Conclusion
In this study a proposal for a data analysis is made to investigate the impact of sharing companies on the traditional markets. This is interesting, because the sharing economy has grown significantly and its economic impact has not widely been explored. This study focuses on the case of Uber and its effects on the taxi industry. The aim of this study is to contribute to the discussion whether and to what extent the traditional taxi industry is affected by Uber.
4.1 Analysis of Possible Results
The focus of the results is on the OLS regression of the DD estimator. The results are divided in three subgroups. The DD estimator is found to be negative, positive or not significant at all. If the DD estimator is found to be negative, the impact of the entrance of Uber is a decrease in revenues of the traditional taxi industry. Uber serves part of the original customers of the traditional taxi companies. If the DD estimator is found to be negative and large, the impact is a large decrease in revenues of the traditional taxi industry. This implies that a great part of the customers of the traditional taxis switched to the use of the services of Uber. According to some of the consumers, the services of Uber and the traditional taxis are substitutable. The services are perfect substitutes if it is proven that revenues of traditional taxi companies decreased to 0. This is only possible if all consumers are acting rational and prices of taxi rides offered by Uber are always lower that the rides offered by traditional taxis, this is very unlikely.
If the DD estimator is found to be positive, the traditional taxi industry has been proved to profit from an entry of Uber in the city. This is possible if the emergence of Uber contributes to the popularity of taxis in general. Again, this would be a very unlikely outcome and not contribute to result of a previous study of Rayle et al. (2014). If this happens to be the outcome of the research, other explanations must be considered which could contribute to a positive DD estimator.
If the DD estimator is found to be not significant, it is not proved that Uber has an impact on the revenues of the traditional taxi industry. If no causal relationship is identified the research is still assumed to be a useful contribution by discovering that Uber has no
significant impact on the taxi industry. The services of Uber and traditional taxis are not substitutes at all according to the consumer.
4.2 Limitations
This study has some limitations which are important to mention. First, in the case that the popularity of Uber has an effect to competitors in the ride-sourcing market and simultaneously affect the traditional taxi industry, the results will overestimate the effects of Uber. Lyft is a well-known but less active competitor of Uber, which could possibly profit by the popularity of Uber. So the results of this research will provide a lower bound estimate of the impact of all ride-sourcing companies. Second, the downside of the used regression is that at least 50 cities should be included in the regression to obtain a credible rejection rate of about 6% as obtained by Bertrand et al (2002).
4.3 Further Research
This research proposal can be extended to a more sophisticated regression by including the intensity of the activities of Uber and by including panel data in the regression. Some robustness checks should be exerted to validate the use of these extensions. These checks are too difficult to perform for a bachelor student, but would contribute in a positive way in the elimination of the impact of Uber on the traditional taxi industry.
Assuming that Uber has a significant impact on the traditional taxi industry, a study towards the possibility of Uber and traditional taxi companies sharing the taxi market would be very interesting. A study which is able to detect a long term equilibrium is able to contribute to the discussion whether sharing companies are able to provide an alternative to the traditional markets.
References
Abadie, A. (2005). Semiparametric difference-in-differences estimators. The Review of
Economic Studies, 72(1), 1-19.
Australian Taxi Industry Association. (2015). State & Territory Taxi Statistics. Retrieved from Australian Taxi Industry Association website: http://www.atia.com.au/taxi-statistics/ Benner, K. (2015, March 21). Uber Might Get An IPO, So Why Can't I Have One? Retrieved
from Bloomberg View website: http://www.bloombergview.com/articles/2015-03-20/uber-might-get-an-ipo-so-why-can-t-i-have-one-
Bertrand, M., Duflo, E., & Mullainathan, S. (2002). How Much Should We Trust
Differences-In-Differences Estimates? (Working Paper No. 8841). Retrieved from National Bureau of
Economic Research website: http://www.nber.org/papers/w8841
Blodget, H. (2014, June 11). I Just Heard Some Startling Things About Uber ... Retrieved from Business Insider UK website: http://www.businessinsider.com/uber-revenue-2014-6?IR=T
Blodget, H. (2015, January 19). Uber CEO Reveals Mind-Boggling New Statistic That Skeptics
Will Hate Business Insider UK. Retrieved from Business Insider UK website:
http://uk.businessinsider.com/uber-revenue-san-francisco-2015-1?r=US
Graham, D. B. (2014, July 10). Uber’s tax-avoidance strategy costs government millions.
How’s that for “sharing?”. Retrieved from 48 Hills website:
http://www.48hills.org/2014/07/10/ubers-tax-avoidance-strategy-costs-government-millions/
Koopman, C., Mitchell, M., & Thierer, A. (2014). The Sharing Economy and Consumer
Protection Regulation: The Case for Policy Change (Working Paper). Retrieved from
Mercatus Center at George Mason University website:
http://mercatus.org/publication/sharing-economy-and-consumer-protection-regulation
Kroft, K., & Pope, D. G. (2014). Does Online Search Crowd Out Traditional Search and Improve Matching Efficiency? Evidence from Craigslist. Journal of Labor Economics,
Myhrvold, C. (2015, January 13). Uber Expectations As We Grow. Retrieved from Newsroom Uber website: http://newsroom.uber.com/2015/01/uber-expectations-as-we-grow/ Nicholson, C. (2014, September 11). Study: Uber Pulls Ahead of Lyft in Riders and Revenue
With 12x Lead in U.S. Retrieved from Future Advisor website:
http://blog.futureadvisor.com/study-uber-pulls-ahead-of-lyft-in-riders-and-revenue-with-12x-lead-in-u-s/
Rauch, D. E., & Schleicher, D. (2015). Like Uber, But for Local Governmental Policy: The
Future of Local Regulation of the “Sharing Economy” (Working Paper). Retrieved from
Social Science Research Network website: http://ssrn.com/abstract=2549919 Rayle, L., Shaheen, S., Chan, N., Dai, D., & Cervero, R. (2014). App-Based, On-Demand Ride
Services: Comparing Taxi and Ridesourcing Trips and User Characteristics in San Francisco. University of California Transportation Center (UCTC).
Shontell, A. (2015, June 17). California Labor Commission rules an Uber driver is an
employee, which could clobber the $50 billion company. Retrieved from Business
Insider UK website: http://uk.businessinsider.com/california-labor-commission-rules-uber-drivers-are-employees-2015-6?r=US
Steininger, K., Vogl, C., & Zettl, R. (1996). The size of the market segment and revealed change in mobility behavior. Transport Policy, 3(4), 177-185.
Stock, H. J., & Watson, M. M. (2012). Introduction to Econometrics: Global Edition. Pearson Education Limited.
Verschuren, P. J., & Doorewaard, H. (2007). Het ontwerpen van een onderzoek. Lemma. Zervas, G., Proserpio, D., & Byers, J. W. (2015, May 7). The Rise of the Sharing Economy:
Estimating the Impact of Airbnb on the Hotel Industry (Working Paper 2013-16).
Retrieved from Social Science Research Network website: http://ssrn.com/abstract=2366898
