H
OW DOES DRIVER EXPERIENCE INLFUENCE
PRICING BEHAVIOR IN PEER
-
TO
-
PEER MARKETS
?
E
VIDENCE FROM
B
LA
B
LA
C
AR
.
DE
MS
C
E
CONOMICS
T
RACK
: M
ARKETS AND
R
EGUALTION
M
ASTER
T
HESIS
STUDENT NAME: Herbrink Kevin STUDENT NUMBER: 11833378 SUPERVISOR: Dr. A.M. Onderstal
Statement of Originality
This document is written by Student Herbrink Kevin who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are 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.
Contents
1 Introduction 3
2 Literature Overview 5
3 The platform: BlaBlaCar.de 9
4 The Data 12
4.1 Data Gathering . . . 12
4.2 The Data . . . 14
5 Methodology 16 5.1 Trip-fixed-effect model . . . 16
5.2 Analysis of price setting . . . 18
5.2.1 The regressions . . . 18
5.3 Analysis of quantity sold . . . 19
5.3.1 How to measure quantity sold . . . 20
5.4 Analysis of profits . . . 21
6 Empirical Analysis 22 6.1 Summary Statistics . . . 22
6.2 Price . . . 23
6.2.1 Price determinants . . . 23
6.2.2 Week and Time influence . . . 30
6.3 Quantity . . . 32
6.3.1 Quantity sold Determinants . . . 32
6.3.2 Week and Time influence . . . 36
6.4 Profit . . . 37
7 Conclusion 39
8 Comparison 42
1
Introduction
Peer-to-Peer markets have become increasingly popular in the last two decades. Companies such as Uber and Airbnb have followed the footsteps of the prob-ably first online Peer-to-Peer market that has been created in 1995, eBay. These markets lead to alternatives to the usual goods and services market and the pricing in these fairly new markets is specifically interesting as mul-tiple aspects of the online market allow for more insight in pricing behavior. In these Peer-to-Peer markets, buyers and sellers are in a market where the experience of the individuals plays a fairly big role in their behavior. Partic-ularly interesting for my study will be the carsharing platform BlaBlaCar.de My research will replicate a study done in the French market by Mehdi Farajallah, Bob Hammond and Thierry Pénard in 2016. They studied what drives the pricing behaviour in peer-to-peer markets with data from the car-sharing platform “Blablacar.de”. I will in future refer to their work as (Fara-jallah et al. (2016)). They conclude that more experienced drivers as well as drivers with higher reputation asked lower prices which goes against the usual brand loyalty effect that can be observed in other markets. The usual brand loyalty effect indicates that higher reputation allows to set higher prices. According to their study, “sharing platforms such as BlaBlaCar.de are determined differently than on other types of peer-to-peer markets such as eBay.” (Farajallah et al. (2016)).
looking into the existing literature concerning peer-to-peer markets. Section 2 consists of a description of the website BlaBlaCar.de while section 3 will be concerning itself with the data gathering process as well as the data itself. Section 4 will discuss the methodology that will be used in order to assess our interpretations and regressions. Section 5 shows the empirical analysis results and the corresponding interpretations. After finishing the empirical analysis, the paper gathers final conclusions in section 6. The final section 7 will be comparison between my results and the results from the French market.
2
Literature Overview
Due to their popularity, peer-to-peer markets have seen been regareded with high importance among economists and therefore induced studies about their functioning and behavior.
These peer-to-peer markets can be split into two groups which represent the “First generation” and the “Second generation”, which in itself shows an evolution. The first generation of Peer-to-Peer markets can be described by remote interactions and contains companies representing in 4 different sections. Craigslist and eBay were the first generation of e-Commerce; Free-lance was a firm representative in Labor markets; Lending Club and Prosper existed in the Lending business and finally in the financial services the mar-ket started off with Kickstarter and CurrencyFair. Recently however, more generally speaking, starting from 2008 with the creating of the well-known accommodation sharing website Airbnb, a more personal Peer-to-Peer model appeared with additional transportation services such as Uber and Lyft (Einav, Farronato, & Levin ,2014). To summarize, this so-called “sharing economy” contains transportation, home services, deliveries, accommodation and more (Farajallah et al. (2016).
The Peer-to-Peer market usually possesses two sides that can either be guest and host, buyer and seller, or more recently also driver and passen-ger. Most of the studies done in Peer-to-Peer markets focuses on the second group, such as eBay. Nevertheless, recent studies also focus on the first
couple mentioned above; guest and host; especially concerning the accom-modation sharing company Airbnb. There exists however an inherent lack of studies for the sharing economy in the branch of transportation especially for the decentralized branch such as BlaBlaCar.de which allows drivers to post their trip offer which can then be accepted by potential riders.
Due to the fast-growing pace and increased frequency of Peer-to-Peer markets that see the face of the earth, a lot of studies have taken place in order to comprehend the markets as well as to gather more insight into how the actual market structure works. Even though all of the existing markets are different, they all share common ground, which can be summa-rized as the following: there are lower entry costs for sellers, the usage of modern technology to match buyers and sellers, pricing mechanisms, use of reputation and feedback mechanisms and in a general manner they allow individuals as well as small businesses to compete with the already existing traditional firms (Einav, Farronato, & Levin, 2015).
Particularly interesting for all these types of Peer-to-Peer markets is that all of them share some kind of reputation mechanism which allows both buyer and seller to build trust which is an essential future for these types of online marketplaces. Without trust, the markets would not be functional and efficient. These sort of reputation and experience level mechanisms that most of these markets share is in essence one of the three possibilities to as-sure a certain level of trust within the marketplace. Online markets can either use up-front inspection, reputation systems or external enforcement.
As up-front inspection and external enforcement are not the most commonly used trust mechanism, my study will focus on the reputation and feedback system. Nevertheless, even though it is the most commonly used mecha-nism, there are certain flaws within the easy to setup systems. Studies have shown that buyers that were disappointed with a service or product most of the time do not leave any feedback (Nosko & Tadelis, 2015), which will result in biased feedback results.
Additionally, depending on the actual system in place, the system may be biased due to the fear of retaliatory feedback, meaning the fear of receiving a bad feedback due to the given feedback (Bolton et al., 2013). As feedback is also most of the time positive, negative feedback usually does not lead to a huge variation within the overall positive reputation given due to the al-ready existing feedback. (Horton & Golden, 2015). These results are mainly from the online selling platform eBay, but nonetheless these problems fac-ing the system, the reputation mechanism of eBay allowed to successfully get rid of most fraudulent behavior and screen out bad actors. (Resnick et al. 2002, Dellarocas 2003, Cabral & Hortacsu, 2010)
On the other side of the spectrum, we have studies showing and analyz-ing the particular characteristics that influence the pricanalyz-ing that results from the reputation system in place. Such studies show for example that a more trustworthy personal picture on the accommodation website Airbnb results in both higher prices on the listing and a higher probability of the offer be-ing chosen (Ert, Fleischer, & Magen, 2015). Contrary to the positive effect
that they find that a personal picture has, the reputation achieved by online review does not seem to have an effect on either the listing’s price or the like-lihood of being chosen. Another study concerning the pricing of Airbnb and its reputation mechanism finds similar results. According to this study, con-sumers of Airbnb perceive certain characteristics as a sort of quality signal that leads them to accept higher prices as trust is build though that mecha-nism. Hence, they could observe that consumers of those kind of hosts were willing to pay premium prices. However, contrary to the study mentioned be-fore, they found that a profile picture was not related to higher rental prices (Wang & Nicolau, 2017). An important addition to the analysis is that they could not study the impact of racial on rental prices due to a lack of data.
Studies concerning the effects of the seller’s reputation on prices on eBay conclude that if a seller was first given negative feedback, the weekly sales dropped; but additional negative arrived more quickly once negative feed-back was already in place however their impact is less than at the begin-ning. Additionally, the lower the reputation of a seller, the more likely he is to exit the market. (Cabral & Hortacsu, 2004). Another study on the effects of a seller’s reputation on prices show that negative rating’s in fact lead to a decrease in price that the seller will receive in his auction bidding (Mickey, 2010).
3
The platform: BlaBlaCar.de
In order to comprehend the functioning of the website that will provide us with the data to analyze, I will be digging into the history of BlaBlaCar.de and its current status.
BlaBlaCar.de has its roots in France where it was created by Frédéric Mazzella, Francis Nappez und Nicolas Brusson in 2006. Nevertheless, this has not been the first version of today’s well-known car sharing platform. Initially it existed under the French name "Covoiturage.fr". Its introduction to the German car sharing market was in April 2013. Particularly inter-esting in the German market is the fact that there already existed many other carsharing websites which were quite successful like for example "Mit-fahrgelegenheit.de" and "Mitfahrzentrale.de". BlaBlaCar.de was however able to construct a leading European carsharing platform by buying the above-mentioned competitors for 100 million Dollars in 2015 after two to three years of talks.
Currently, BlaBlaCar.de is considered to be the leading car sharing plat-form by being present in over 22 countries with over 60 million users world-wide. Most of these countries are European countries but countries such as India, Russia and Mexico are also represented. Per quarter around 18 million people use the platform. Concerning the workforce of BlaBlaCar.de, they are currently represented by 400 employees in 11 international offices. The setup and utilization of the website and the principle behind it is
quite simple. The website allows drivers to post the ride that they offer with a certain price they ask for a seat. Potential riders are able to scan the website for all rides offered for a certain intercity trip they want to undergo. Riders can make themselves a image before the ride of the driver by having access to the profile of the drivers that offer the rides. Consequently, they can see the reputation a driver has, his rank within BlaBlaCar.de which will be discussed below, his preferences for music, pets and much more. This allows for an interesting interaction between the pricing of the offered trips as well as the seats that are sold due to certain aspects of the profiles which will be studied in this thesis. The different ranks and their corresponding features are shown in table (7).
The company itself underwent changes over time that play an interesting role in the analysis that will follow. In the early days of the company, cash was the main payment method which over time switched to online payment options such as PayPal. Recently however, the company allowed again for payments to be made in cash.
Additionally, and very important for the comparison with the French study is that in 2012, BlaBlaCar.de introduced a system in which prices set above the recommended price by BlaBlaCar.de will be displayed by a certain color scheme that allows potential riders to know whether and by how much a price deviates from the price that BlaBlaCar.de initially recommended for the trip. Prices would be displayed in yellow if the price was set up to 125% and it would be displayed in red if it was set between 125% and 150%. A
green color would indicate a recommended price or even a price set below the recommended price. This system is however no longer active which is a very critical point in my study as we are able to see the difference of this change. We can observe how different attributes of a driver play a role if po-tential riders do not any longer know whether a price is set above or below the recommended price. The current recommended price in Germany is set to be C5/100km and people are still able to lower or raise the recommended price by 50%.
The price finally paid by the riders that book a seat consist in addition to the recommended price also a service fee which also underwent recent changes. Recent changes introduced two so called "use packages" which can cost either C2,99 for a whole week or rather C14,99 for a total time period of six months. This replaced the old service fee which was based on the price of the trip and was do be paid each single trip by a rider.
Very recently, BlaBlaCar.de introduced rides for only female passengers called "Ladies-only" which would be an interesting study topic on how dif-ferently attributes of only female drivers and riders differ from mixed rides. However, we are now able to observe whether a seat has been booked by a fe-male or fe-male rider which will be part of my study in analyzing the potential differences of factors that affect the quantity sold by gender.
4
The Data
4.1
Data Gathering
The data collection was automated. For a single month, on a daily basis, a crawler was scraping the BlaBlaCar.de website for the German market and gathered every single trip that was offered for 40 trips between German cities.
The trips are the following: Frankfurt-Berlin, Kiel-Hamburg, Köln-München, München-Stuttgart, Berlin-Kiel, Köln-Hamburg, Düsseldorf-Hamburg, Hamburg-Berlin, Hamburg-Berlin, Münster-Hamburg, Dresden, München-Köln, Köln-Berlin, Düsseldorf-Berlin, Köln-Frankfurt, Hamburg-Frankfurt, Frankfurt-Köln, Düsseldorf, Köln, Frankfurt, Berlin-München, Berlin-Hamburg, Berlin-Nürnberg, Nürnberg-Berlin, Köln-Stuttgart, Hamburg-Köln, Hamburg-Düsseldorf, Kiel-Berlin, Hamburg-Kiel, Stuttgart-Berlin, Berlin-Stuttgart, Stuttgart-MünchenFrankfurt-München, Dresden-München, Hamburg-Münster, München-Frankfurt, Stuttgart-Köln, Frankfurt-Hamburg, Berlin-Bremen, Bremen-Berlin.
This list can be found in the Appendix with the average price and the distance for each trip. The cities chosen in the sample have a population of at least 100.000 people. These trips have been chosen in such a manner that they constitute a representative sample of possible trips in the German market with respect to the distance to travel. It is worth to mention that we do not need multiple snapshots of a specific trip (three days before departure,
one day before departure, etc.) as BlaBlaCar.de does not allow the price to change after having posted a trip offer.
Hence, the crawler ran once the day before the trip to gather all the trips available at that point in time, and it will run additionally again on the day of the trip itself multiple times according to the number of trips and departure times observed the day before in order to observe possible changes of the number of seats sold. The crawler had to run on the day of the trips themselves shortly before the departure time in order to account for either seats that have been sold after the first crawl as well as to check if a certain trip disappeared from the searches which would indicate that all seats have been sold. If this would not be done, it would be impossible to observe trip offers where all seats have been sold as they do not appear in the searches and could lead to quantity results not representing reality later on.
Accordingly, the crawler scrapes the website 1 hour before the trip in or-der to see whether the trip observed the day before disappeared, meaning that is has been fully booked, or whether more seats have been sold. This solves the problem of only gathering data from the day before the departure as changes can be happening in between the scraping and the actual depar-ture which are essential for the analysis; as we would otherwise have data not representing the actual number of seats sold but only seats sold at a certain point in time which could invalidate our interpretations.
An additional factor that will be used in my study is whether a driver has a German name or not. This will be done by identifying the origins
of the driver’s profile name with a dataset constructed by myself with all the predominant German sounding names that were gathered from multi-ple websites and datasets. The same was done for finding out the gender of the drivers by constructing a dataset of predominant names that can be associated to gender.
4.2
The Data
The data collected by the crawler will contain the departure city, arrival city, departure time, arrival time, the date, the driver’s name, his age, gender, whether a profile picture is shown, his preferences (pets, smoking, talking, music), the number of seats available, seat taken by male or female, the price of a seat, the driver’s rank, his rating, how many ratings he received, the number of Facebook friends, 2 seat warranty, immediate booking option, total number of seats offered, phone number shown or not, verified email address or not, the car model and colour, the number of trips already offered and the date of account creation.
Additionally to all the data that the crawler scrapes from the website, I will be adding additional data in order to identify whether the driver is male or female, as well as what his or her origins are. This will be done with the help of multiple websites that allow me to identify both the originof the name and the gender of a person with the help of the presented name on the trip offer observed on BlaBlaCar.de. These websites will be used as mentioned
above in order to create the necessary gender and name origin datasets. The website are the following : www.surnames.behindthename.com, www.ancestry.com, www.namepedia.org, www.genderize.io. Some existing datasets have been used as well.
However, an important fact that one needs to consider is that for the study done for the French market, from the time they gathered their data, the website of BlaBlaCar.de changed in a big fashion as one of their main functions disappeared which allowed riders to see whether or not the price was above or below the recommended price of BlaBlaCar.de. This may ex-plain the pricing behaviour of the experienced drivers for asking lower prices as people may tend to rather book a trip where it is indicated that the price is close to or exactly the recommended price. Therefore, I assume that the results may be different today as the data will be quite different as well and it may be very interesting to see how the behaviour of pricing changes due to changes of the variables that riders can observe.
The data collection has started on Thursday the 3rd of May 2018 and ended on the 26th of June 2018. The data consist of 35,507 single observa-tions of trips offered by the 21,474 different drivers. The individual drivers have been attributed by an ID which allowed their identification. The ID consists of their name, age, the year of their account creation, their name origin, their gender as well as if a profile picture is shown or not. This identi-fication process will allow us to gather summary statistics without duplicate data information as every single driver can be taken into account only once.
5
Methodology
5.1
Trip-fixed-effect model
In order to analyze what factors influence the pricing behavior and to what extent the reputation system plays a role in it, I will perform multiple regres-sions. The analysis that will follow will both explain the pricing mechanism as well as the quantity that is effectively sold by the drivers. The analysis is based on the data that has been presented beforehand which will allow us to identify the relevant factors.
In a very general matter, the regression models that will be used can be described as a fixed-effect cross-sectional data model which is similar to the one used in the French market study. The trip-fixed-effects that will be controlled for are identifiable with a pair of departure and arrival city. As BlaBlaCar.de allows drivers to set small cities as arrival or departure city, there was a need to go through all of the smaller cities in the data and associate them to the closest bigger city which gave us a total of 40 intercity trip pairs. Hence, we can use these fixed effects to get rid of trip specific characteristics such as for example the distance differences between trips.
The main econometric model can be represented as follows:
pi j= β1+ Xi+ ψj+ ²i j qi j= β2+ pi j+ Xi+ Zj+ φj+ ηi j
Where i represents the individual driver, j represents the trip andψ re-spectively ϕ accounts for the trip-fixed-effects. p represents the price as q represents the quantity of seats sold. X contains all the characteristics that we are able to observe in our data that could influence both price or quantity sold. Both the regression for the price as well as the regression for quantity of seats sold possess a constant represented byβwith subscripts 1 and 2. Z is our instrumental variable called "average-other-price instrument". Con-cerning our error termsηandεwe will assume a random error term which is normally distributed and has therefore a mean zero. The regression on price will be a simple linear regression while the regression on quantity will be a two-stage-least-squared regression with our already mentioned instrument to account for endogeneity between price and quantity.
5.2
Analysis of price setting
5.2.1 The regressions
There will be multiple variations of the above general model. Accordingly, I will initially observe how the price is affected by the different factors of the driver by also controlling for the trip-fixed-effects.
Additionally to the above regression analysis, the dataset will be sepa-rated into the different ranks that BlaBlaCar.de gives to the drivers accord-ing to their experience and other factors. Hence, we will observe how the different factors have an effect within the different ranks.
Following this analysis, I will identify how the price is affected by both the day of the week as well as the time of departure. The time of departure will be split into 3 different time periods which will be:
o -Departure before 12:00
o -Departure between 12:00 and 18:00 o -Departure after 18:00
The same day of the week and departure time analysis will be done for the actual quantity sold.
5.3
Analysis of quantity sold
Following this price analysis, our methodology continues by analyzing what factors influence the actual quantity that has been sold. As both price and quantity are endogenous, an instrumental variable will be used in order to account for the endogeneity. The instrument used will be the same one that has been used in the French study and is essentially the “average-other-price instrument" (Farajallah et al. (2016)). This means that a specific driver will be linked to other trips that he offered in order to account for certain characteristics of drivers that affects its price setting but that we are not able to observe. This argumentation has been used in the French study and holds its ground on the German market as well. Thus, the thought of the instrument to be strong can be shared.
Therefore, an average price for every single trip a driver offers will be calculated except that we do not take into account the price that he or she sets for the trip in question. One has to note however that it is possible that a driver only ever offers one trip. If this is the case, the driver will have an instrument value that is equal to zero. Concerning the exogeneity of the instrument, the argumentation will be the same way as it was in the French study as we do not change the relevant data and as such the instrument should be packed with factors concerning the driver that influences the price he sets on the specific trip in question, but not affect the demand in any other way. In a general matter, this instrument is commonly used in Industrial Organization literature and are called BLP instruments. (Berry, Levinsohn,
and Pakes (1995))
5.3.1 How to measure quantity sold
To measure the quantity sold, multiple alternatives are present. We will be using two quantity measures that have already been used in the French study which are the fraction of seat sold as well as a dummy variable that will be equal to zero if not all seats were sold and equal to 1 if all seats of the offered trip have been sold. Importantly to note is that we do not occur the problem of the French study that we need to find the maximum number of seats offered by a driver as the data has changed since then and we are able to observe the total number of seats offered even if some seats have already been booked. Accordingly, we do not need to construct this variable. We are able to improve on this point as we do not need to construct the maximum number of available seats, or also the maximum number of seats initially offered by looking at all offers of a specific driver and using the highest value as the maximum number of seats available.
The fraction sold variable will be calculated by dividing the number of sold seats by the total number of seats offered for that trip. Therefore, it takes a value between 0 and 1. It is equal to zero if no seats have been sold and equal to 1 if all seats have been sold. As mentioned above, we should be able to improve on the French study as we are able to specifically and correctly observe the denominator of the fraction.
In addition to the above-mentioned quantity sold variable, we will add the simple number of seats that have been sold as an integer value. However, more importantly, we are able to observe with our current dataset if a seat sold has been sold to a male or female individual which allows us to further our analysis by splitting our quantity sold variable into two subgroups for the gender. This was not doable with the dataset used in the French study, therefore we can analyze how the number of seats actually sold differ across gender as well as which factors have most influence for which gender.
5.4
Analysis of profits
In addition to the price and quantity sold analysis, we will also perform an analysis of the profit that a driver actually receives on a trip by multiplying the price by the actual quantity sold as well. However, we will even go a step further and run a regression on the expected profit or also called the maxi-mum profit possible on a trip. This profit can be identified and calculated by multiplying the price with the maximum number of seats which we are able to observe directly in the data. We will accordingly as we have done it for the price and quantity regressions analyze the difference accross the ranks and also study how both variable differ over the course of the week.
6
Empirical Analysis
6.1
Summary Statistics
Analyzing our data, we have found that out of all the trips that have been made within the time period that the data scraper crawled the website BlaBlaCar.de, a total of 21,474 individual drivers offered 35,507 trips in to-tal. These individual drivers have been identified according to the already mentioned procedure in section 4.2.
On average, the individual BlaBlaCar.de driver is 34 years old while his account has been created on average 3 years ago, which would be in 2015. Out of the 5 possible stars that an individual can get for his rating, on aver-age an individual has 4,3 stars out of 5. Additionally, we can conclude that for all the individuals in our dataset, the profile picture is shown in 82% of the cases while only 35% accept smoking in their car and 50% of people en-joy listening to music on the trip. On average, one has 166 Facebook friends and will share the immediate booking feature in 43% of the cases. Nearly 42% of the drivers have a non-German sounding name while 38% of the in-dividuals is female. To add more to our knowledge about the average driver, we conclude that he already did 48 trips as well as having around 35 rating feedbacks.
Concerning the trips themselves, we could observe that out of the total 35,507 trips, most of them have been from Hamburg to Kiel (both ways), from Hamburg to Berlin (both ways), from Cologne to Frankfurt (both ways)
and Stuttgart to Munich. These account themselves for 32% trips out of all the 40 routes possible. The cheapest route has been from Hamburg to Kiel with an average price of only C4.98while the most expensive route is from Berlin to Stuttgart with an average price of C31.41. The average price for all the trips in total is around C19.
In the period we gathered our data, while looking for a trip you were on average able to book one to two seats while on average three seats are offered for trips. Out of all the seats that have been offered, a total of 41% has been booked by females while the rest has been booked by male riders. As expected, most of the trips have been offered on Friday as well as on Sunday with 21% respectively 20% of all trips.
6.2
Price
6.2.1 Price determinants
In our first part of the analysis we will be concentrating on how the price is determined by the drivers, meaning how they set their prices according to their specific characteristics such as for example their experience level.
In our table 1 in the annex we are able to come to multiple conclusions. The age of the driver who sets prices seems to be affecting the price in an upward manner while the age of the account itself is negatively influ-encing prices set. So the longer the account exists, the lower the price is that a driver sets for seats that he wishes to sell. In more detail, being on
BlaBlaCar.de for an additional year leads to a price decrease of around 0.3% while being a year older leads to a 0.1% price increase. Additionally, more seats offered in total for a certain trip is raising the price. An additional seat offered leads to a 1% increase in the price set. This could be considered as being counter-intuitive as by lowering prices and offering more seats one could gain more as it could be the cheapest option for riders. Hence, this reasoning can not be applied here.
On the other hand, allowing music or being in favor of music being played on trips seems to lead to higher prices while not allowing smoking leads to lower prices. Allowance of music being played lead to a 1% increase in prices while not allowing smoking leads to a 0.4% price decrease. One could reason that as one does not allow smoking, potential riders are not willing to book a ride, so in order to attract more riders lowering the price can be a solution. The reasoning behind this could however also go into the other direction as by allowing smoking, potential riders that do not like people smoking around them leads these people to avoid those trips, thus lowering prices should even attract people that smoke but would not mind not smoking if prices were cheaper than in a smoking-allowing ride.
We can take note that the number of ratings a driver already received does not seem to influence pricing as well as the number of Facebook friends a driver has. The number of ratings received is not statistically significant and even though the number of other trips already completed is significant, the influence seems to not exist. A reason for the number of ratings a driver
already received can be explained by the possible collinearity between the number of ratings received as well as the rank level a driver has as it is integrated in it. Whether a profile picture is shown or not is insignificant as well which is reasonable to the extend that most people have a personalized profile picture which should therefore not be the price determining factor. By accepting the option to let potential riders immediately book a seat, prices will be declining by 4.1%. Thereby, we can conclude that the "Immediate Booking" option interacts in an inverse way with the price that is set.
Interestingly, we are able to observe that if a driver is female, the price charged is lower than compared to male drivers. To be exact, females have a price that is on average 0.8% lower than those men offer. This result is statistically significant at a 1% significance level. The reason for this lower price asked by woman can have multiple explanations.
This may be explained to some extent by the still controversial belief that woman tend to be more nervous while driving or are more likely to feel at ease with taking complete strangers in their car. Hence, female drivers must lower their prices to attract more riders. Proof may be the new ability given by BlaBlaCar.de to allow woman to only ride with other woman as an option to choose. Giving this option to woman and not offering the same for men seems to be giving hold to the statement that woman are not able to set the price to the level of men without sacrificing seats sold.
On the other hand, one can also argue that women need to ask for lower prices as other females feel safer with a female driver instead of male drivers.
This argument derives from the above-mentioned argument in the manner that it seems that more male individuals exist on the website; leading to a lack of potential female passengers that would rather pay more to travel among females. It follows that they lower their prices to attract male riders. If the gender distribution would be different one could maybe expect that females could ask for higher prices as other females feel safer while having a woman in the car in the sense that woman do not tend to be expected to be a criminal with an ill intend. This could elaborate that prices are higher as people tend to pay more for less risk. Nevertheless, whatever interpretation is used, we can conclude that woman in general tend to have higher prices for selling seats for trips.
Shifting our analysis to the origin of the names, we can extrude some interesting observations. First of all, we observe that in our price deter-mining regression having a non-German sounding name is the case in ap-proximately 42% of our drivers in the dataset; and leads to an increase in prices. Prices set by non-German sounding names are higher compared to German sounding names by 4% . The overall argumentation behind this may be very vague as there exist a lot of different attributes for individuals to chose higher prices. As an example, one can argue that people with a non-German sounding name may expect to sell less seats due to that fact even if they have a lower price, therefore this would lead them to increase prices in order to make more money per seat sold instead of trying to sell all seats for lower prices. Additionally, it may very well be possible that most seats sold
by a "non-German" may be booked by another individual who as well does not have a non-German sounding name. However, the current data does not allow to analyze this, but we can analyze whether a seat was booked by a male or female individual which we will be analyzing in our quantity regres-sion later on. As a note, we expect that drivers with a non-German sounding name will more likely sell less seats compared to individuals with a German sounding name.
Separately, we are now going to investigate how the experience influ-ences the price setting of drivers according to the ranking system within BlaBlaCar.de. As we will run into the dummy variable trap by including all ranks, we will be using the last rank called "Ambassador" as a base. Hence, the coefficients will be compared to that specific rank.
We observe that drivers that are new, or we should rather say do not possess a specific rank yet, have the highest prices out of all the possible ex-perience levels. Comparing a non-rank driver with an Ambassador, we can see that they have a price set that is overall 3.5% higher. We monitor a spe-cific trend as some individual increases in rank; as an individual goes from not having a rank to the first rank possible; which is that of a "Beginner", prices decline. This means that while being in the process of ranking up, individuals learn from their experience and set lower prices in accordance.
Prices declined in a very significant manner especially compared to the other milestones that the drivers experience. This step in the ranking sys-tem leads to apparently a very big change in the pricing mentality and
scheme. Drivers may have realized that by setting lower prices they are able to attract more riders by doing so. A possible reasoning could be that riders in lower ranks pay more attention to the actual price instead of the other attributes or that rather which is more plausible; that people with no rank set prices way too high by intuition as they do not know what the best price to set is as they are lacking the necessary experience. We will analyze in a later regression what attributes of a driver drives prices throughout the different levels.
The next milestone a driver experiences after being ranked as "Beginner" is the rank "Intermediate". These drivers set prices that are 1.7% higher than those set by the most experienced drivers. The upgrade in their rank from the previous one does have a decreasing effect but to a much lower power. Receiving a first rank has a far bigger change than receiving their second rank. The same reasoning as above applies, meaning that drivers learn to set lower prices, but with the addition that prices are now closer to the ones set by "Ambassadors".
Going from the rank "Intermediate" to "Expert" seems to be contradictory to the before observed trend. We observe that prices set for seats are actually higher than on the lower rank. Comparing the prices to those of our base rank leads us to conclude that prices are 1.8% higher than those of the final rank. The increase in the price is actually higher than the decrease in prices observed in the milestone before. The intuition behind this may be the fact that people assume they have gained the trust of riders by achieving a higher
rank. This rank will therefore lead them to ask for higher prices as people are expecting a good service out of the trip compared to people on lower ranks.
After this price spike for the rank "Expert", we conclude that the price declining pattern we observed beforehand by increasing ranks comes back. By achieving the last step in the reputation mechanism of BlaBlaCar.de, which is the one of an "Ambassador", prices seem to be lower than when being at any other level. This astonishing fact is very unique in the sense that we first do not observe the typical reputation mechanism where having a higher rank and reputation leads to being able to charge higher prices, as well as having a break in the decline of the pricing by ranks. The first has been found in the French study and seems to be a particularity in these types of markets. However, the latter seems to be specific to this case as the break in the trend has not been observed beforehand. It seems as if people try to use the ranks to be able to achieve the possibility of gaining trust which they think allows them to set higher prices. After trying the above explained tactic, it seems they go back to lowering their prices as they seem to have undergone negative experience by increasing their prices. Whether this scheme makes sense, works or how it is to be analyzed will be done in section 6.3.
6.2.2 Week and Time influence
In the following section we will focus on how the pricing is influenced by both the day of the weeks as well as the departure time. One can understand especially for the departure time that the price should be influenced as it may be more exhausting for drivers to drive both very early in the morning or the opposite case, very late in the evening. For both the weekdays as well as the departure time, we will be using dummy variable in order to find out how they influence pricing as well as the quantity sold.
In table (2) we are able to observe that concerning the price, there is a big influence by the departure time. Trips that depart after 18:00 have the highest prices, followed by departures before 12:00, while in the meantime trips that start their trip between 12:00 and 18:00 have a lower price. The intuition behind this may seem obvious as people expect to be able to sell more seats in early or late trips as less people drive in the early morning or late evening, as well as it is more exhausting for the driver for which riders may want to compensate for. In a general manner, concerning the weekdays, most of the coefficients are insignificant or only significant to a 10% signif-icance level which leads us to shorten our interpretation. Nevertheless, we are able to observe that prices tend to be lower on Wednesday and Tuesdays, while they seem to be higher on Thursdays. This reasoning follows that our base day is Sunday which is not included in the regression due to the dummy variable trap.
the departure time, we will take a different angle in our analysis in the following way. We will split the drivers into their corresponding rank given by BlaBlaCar.de in order to assess whether or not there is a difference in the pricing methodology associated with the different ranks.
As we can identify, some of the days of the week are statistically insignif-icant for multiple ranks, this may be due to the fact that either not enough data is available for those days or that they actually do not play much of a role in choosing one’s price for selling seats. Nevertheless, we can still ana-lyze in most cases how the departure time affects prices as they are mostly statistically significant. On that account, we can conclude that for all ranks except for the "Expert" rank where our coefficients are insignificant, prices tend to be higher if trips are offered to depart after 18:00 compared to the early departures before 12:00.
There also seems to appear a downward trend of price when the rank increases for both cases specifically by receiving their first rank as the heav-iness by which price is affected by the departure time declines from having no rank to achieving the first milestone. Hence, the departure time seems to lead drivers that are new to the website to ask for a much higher price, which changes over the course of their experience level with a big decline in prices for the unlocking of their first rank. After receiving their first rank, prices do not seems to decline in a big fashion as they did after achieving the first rank. Drivers therefore seem to learn that they set prices way too high due to early or late departures and adapt pretty quickly prices in accordance.
6.3
Quantity
6.3.1 Quantity sold Determinants
Our final step consists of analyzing how the quantity that is actually sold is affected by all the characteristics a potential rider can observe and notably taking notice on how the experience level plays a role. Additionally, as we are able to identify whether a seat has been sold to a female or male rider, we can go further into detail in our analysis by comparing how differently the two groups behave. In a first step, we will be using two quantity variables in order to assess the general effect of driver characteristics on the quantity sold. We will drop our "All Seats Sold Dummy Variable" as a dependent vari-able in our case as we do not receive sufficient significant results in order to interpret them in a meaningful manner. Hence, we will stick to the quantity sold as a simple integer value as well as the fraction sold variable that has already been explained in section 5.3.1.
In our regression table (4), we can observe how driver characteristics in-fluence the quantity actually sold. First of all, as expected a higher price leads to a lower quantity sold, which is the case for all 4 regressions. Hav-ing a higher age on the contrary seems to lead to a lower quantity sold even though the amount of influence is very low. Increasing "age" by one will lead to a small decrease of quantity sold. Compared to other coefficients, the im-mediate booking option provides a higher influence on the actual quantity sold. Having the option enabled lead to a higher quantity sold, which leads us to believe that people prefer to book trips that have this option enabled.
We can however also note that the prices that are set for these trips are gen-erally lower than others which may also play an additional role in the higher quantity sold. Not allowing people to smoke on a trip leads in the same way to an increase of the quantity that is sold. We can note that for regression (1) and (2), the number of ratings received has a very small influence on quan-tity sold while the regressions (3) and (4) seems to not indicate any influence at all.
Moreover, we have indication that having a non-German sounding name results in lower quantities sold. This hold true for all 4 regressions while the effect becomes bigger with the controlling for endogeneity between price and quantity sold. On the other hand, however, having a female sounding name does not result in significant results excepts for our regression (3) where the fraction sold is controlled for endogeneity. Here we can observe a positive influence on the quantity that has been sold.
Concerning the influence of the ranking system on the actual quantity that a driver is able to sell we have mostly significant results. We are only unable to interpret the "No Rank" coefficients as they are all insignificant. The highest amount of quantity sold can be attributed to the highest rank which is the one of an "Ambassador" followed by a decreasing manner in an expected way. This means that the higher the rank is that a driver pos-sesses, the more seats he is able to sell. We do not observe the break that we have observed in the pricing mechanism where the milestone of achieving the "Intermediate" rank lead to a spike in the otherwise decreasing trend.
Therefore, for the quantity sold we observe a steady increase in quantity sold as people rise in rank.
After having analyzed the general quantity sold without taking into ac-count the gender of the rider that booked the trip, we will be doing exactly that in the following.
As expected, increased prices lead to lower quantity sold to both male or female riders, except that for regression (1) we do not observe a significant result for price. Having the immediate booking option leads in general to a increase in the quantity sold to both genders as well as not allowing smoking on the trip. We can however note that having the immediate booking option enable has a higher influence on females than males while not allowing to smoke has a similar magnitude. The number of trips already underwent as well as the number of ratings received has no worthwhile influence on the quantity sold. However, when a driver has a non-German sounding name, we observe that for both genders, the influence is negative, meaning that less seats are sold to both males and females. Nevertheless, if we control for price endogeneity, we can conclude that for both genders the influence is nearly identical with a slightly higher influence on male riders. By not controlling for price endogeneity, we have the opposite reasoning where the influence of not having a German sounding name has a bigger influence on female riders.
In the next step we will be analyzing the ranking system in such a man-ner to identify if there is a difference on how it influences the quantity sold
to either males or females. The results are all significant except for the male quantity sold for the rank "No Rank". Concerning the female population of riders, having the highest rank leads to the highest quantity sold. The other ranks however do not seem to indicate a specific trend as by receiving the first rank; quantity sold goes down, while achieving the next milestone will again lead to an increase while the second last rank of "Expert" will again lead to a decrease in quantity sold.
Concerning the male population of riders on the other hand, we are able to observe the same as in the female population concerning the last rank. Having the last rank leads to the highest quantity sold. However, for the other ranks we are able to observe a certain trend, which consists of an increasing trend in quantity sold as one increases in rank. Hence, male riders seem to be more attracted to the higher ranks. Contrary to the female population the mechanism seems to be less influenced by the rank a driver has as quantity sold does not have a certain trend in accordance with the increased rank a driver has.
6.3.2 Week and Time influence
Shifting our focus to the quantity, we will have multiple variables in order to quantify the quantity of seats sold in order to assess whether or not there is much of an influence and compare the results to reassure the significance of the interpretations. First of all, in regression (1) we realize that most of the seats are actually sold on Sundays followed by Fridays and Mondays. This means that most users travel for the weekend; meaning they travel somewhere on Friday and come back on either Sunday or Monday.
On the other hand, our regression (2) indicates that most seats are sold on Sunday, followed by Friday, Wednesday as well as Thursday. Additionally, we conclude that early departures as well as late departure are negatively affecting the quantity of seats sold where late departure has a bigger in-fluence than early departure on lowering the quantity sold. The "All Seats Dummy" variable regression does not provide us with necessary significant results which leads us to drop its interpretation.
We then proceed to separate the quantity sold in both seats sold to fe-males as well as seats sold to fe-males. Concerning the male population, it seems that Sunday is still the most common day to travel, followed by Fri-day and MonFri-day. For the female population, we can observe that SunFri-day is as well the most commonly used day for travelling. Sunday is followed by Friday, Wednesday and Monday. By looking at our time variables, we can explain that for both quantities sold to males and females, departing after
females seems to be stricter.
6.4
Profit
In our final step we will be having a look on how the actual profit is influ-enced as well as the expected profit. Profit should be understood as the value of money drivers make or could make on their trip offers. Henceforth, in our first regression we will analyze the actual profit that is made, and the sec-ond regression will on the other hand see how the actual; or let us rather call it the maximum profit possible; meaning the total amounts of seats offered multiplied by the price asked for the trip, is influenced.
Our first regression (1) indicates that being older by a year leads to a 0.1% increase in profits while the expected profit is increased by 0.2% as seen in regression (2). On the other hand, having an account that is a year older leads to a decrease in the actual profits by 0.4% while the expected profit decreases by 0.5%. The immediate booking option has the same neg-ative effect but to a much larger magnitude, enabling the option leads to a decrease of 5.9% in actual profits but only 0.7% in expected profits. Being in favor of music only play a significant role in the expected profit where it seems to increase it by 1.4%. Personalized profile pictures are only signifi-cant for the expected profits as well and is negatively affecting the expected profit by having in general 2.4% lower expected profits.
mea-sures of profits. Having a non-German sounding name leads to an increase of 7.1% for actual profits and an increase of 7.4% for expected profit. Looking at the gender we can observe an effect going the other way around, where be-ing a female is associated with both lower actual (2.7%) as well as expected (2.4%) profits.
Concerning the ranking system and its influence on profits, we are only able to observe the expected profits with a meaningful interpretation as the actual profit does not provide us with significant results to interpret. Hence, we can note that expected profit is the highest for individuals that do not possess a rank yet and is followed by the highest rank "Ambassador". By receiving the first rank, drivers have a decline in expected profits which will however be raising again as they rank up. Therefore, we can say that for the expected profits the trend is increasing except for the very first milestone.
7
Conclusion
In our study of how the experience of the driver influences the pricing mech-anism in the carsharing platform BlaBlaCar.de as well as what the conse-quences for the quantity sold are, we can come to final conclusions. By using our trip-fixed effect model and making use of our data by splitting our re-gression in multiple ways, we were able to gather the following conclusions. First of all, we came to the final conclusion that there does not exist a tra-ditional loyalty effect. According to the loyalty effect, drivers with a higher reputation and experience are able to charge higher prices which we do not observe. We observe a general decreasing trend of prices as individuals in-crease in rank. Ranks are the reputation mechanism by which BlaBlaCar.de attributed the experience and trust level of drivers according to multiple characteristics such as a certain threshold for the number of positive rank-ings. Nevertheless, we observe a break in the decreasing trend as a driver goes from rank "Intermediate" to the rank "Expert" as people ask for higher prices. One would expect that the quantity sold will suffer accordingly but our study shows the opposite. The break does not have an influence on the increasing quantity sold as drivers increase in their rank. We can be assured that the general observation of having a price decreasing trend with respect to the increases in the ranking system is not a mere correlation as splitting the data into different ways we still observe the general trend no matter the changes we undergo. Nevertheless, as we are talking about mostly human
behavior for pricing, we will have variables that we cannot take into account as they are not observable, which could lead to some measurement errors. However, the interpretations should hold.
As we are able to split our data of quantity sold in both females and males, we can come to a more detailled conclusion than the one from a gen-eral quantity sold point of view. We were able to observe that male quantity follows the general increasing pattern of the quantity sold, but the quantity sold to females seems to suffer from the increase in prices due to the break we observed beforehand. As a driver goes from "Intermediate" to "Expert", he increases his price and this will cause the quantity that is sold to females to decrease, while quantity sold to males still increases.
The general effect on quantity seems to follow the quantity sold to males, which we are able to verify by the fact that the influence of quantity sold by ranking up has a higher influence on males compared to females and the male population of riders is higher than those of the female population. It follows that we can conclude that quantity sold to females suffers from the break in decreasing price by rank while male quantity is not affected in that way. This ability to split the quantity sold in two groups as well as the improvement of our data set by being able to observe the total number of seats initially offered allows us to improve existing literature concerning this topic.
Concerning our analysis of the gender, we conclude that females set lower prices compared to male drivers while the quantity sold by females is higher
than by their male counterpart. Focusing on the effect of having a German-sounding name or not lead us to the conclusion that having a non-German sounding name leads to higher prices set compared to drivers with German-sounding names. Accordingly, we find that as a driver with a non-German sounding name sets higher prices, the quantity that he sells is lower than that of a driver with a German-sounding name.
We were also able to add an interpretation of how the different ranks have an influence of the expected or maximum profit a driver can make ac-cording to his rank. Having no rank yet leads to the highest expected profit but they sell the least amount of seats which leads them to lower their prices and decrease their expected profits while ranking up as they will then sell more seats. Thereafter, according to our price and quantity regressions be-forehand, we can confirm that ranking up leads to lower prices and more seats sold which is replicated in this regression in a similar manner.
These differences may lead us to believe that either due to changes in the observable data people behave differently or that the German market behaves differently than the French market. An interesting point and also research question for further analysis would be how different countries be-have concerning platform such as BlaBlaCar.de. This papers proof that the study on peer-to-peer markets such as BlaBlaCar.de need even more insight as the results from the French study do not seem to hold on an international level as we are able to gather different informations out of the more recent data.
It may very well be possible that the differences are due to the changes in the available data and the website itself, but those differences may also be due to changes in peoples mentality as they no longer observe how prices are related to the recommended price. Nevertheless the cause, this paper shows that the existing literature may not be taken as fully understanding its domain, but rather as giving a small insight at a specific time in how driver experience as well as other characteristics influence pricing. The pa-per shows that people seem to behave in a different way as the French study suggests as we do not observe the typical decreasing pricing trend they ob-serve.
8
Comparison
As my study replicates part of the study already done for the French market, we will be comparing my result to theirs in order to identify differences as well as similarities.
First of all, we obtain different results concerning the setting of the price concerning the different ranks. The French market was identified by having a decrease in the setting of the price as ranks went up without any break in the trend. Our study however is marked by that specific trend as by ranking up to the rank of an "Expert", the price generally set increases. We therefore have contradicting results which my be due to differences in the behavior of the countries population, hence country specific effects we cannot
account for in our study or due to the changes in how the data as well as the pricing mechanism changed over the years. It may very well be the case that the disappearance of the color system to identify prices set above the recommended price lead drivers to behave differently.
We obtain different results concerning both gender as well as the origin of the name. In our analysis, females set lower prices than their male coun-terpart while the opposite was the case in the French study. Nevertheless, we obtain the same result concerning the quantity sold by females which is higher than for males.
We are however not able to compare the results of our quantity sold vari-ables split by gender as their data set was not able to identify the following which makes this study even more relevant as we are able to add our results to the existing literature.
The origin of the name seems to also be having different effects in our case compared to the French case as we conclude that drivers with a non-German sounding name set higher prices than people having a non-German sounding name. The opposite was observed in the French study where French sounding names lead to higher prices. Concerning the quantity sold, we do not observe the same pattern as the French study did as for our case non-German sounding names sell more seats. The opposite was the case in France.
Further, our dataset improved compared the French study and enhanced our quantity measures as well as the gender specification to whom seats
have been sold which allowed us to analyze how differently gender behaves in the booking of seats. The results we gathered gave even more insights into how decentralized platforms such as BlaBlaCar.de operate and how people who use them behave. Experience and the reputation system which is one of the main points of assuring people of the safety of their product plays as expected a big role, but we do not observe the typical loyalty effect where people ask for higher price as their reputation grows. This was the same observation as in the French study, but we have identified a break where people seem to set higher prices at a certain milestone and decrease it again afterwards, probably after learning that they cannot profit from the typical loyalty effect as it is the case in other types of markets such as eBay.
References
[1] Luca, M. (2016) Designing Online Marketplaces Trust and Reputation Mechanisms Cambridge, MA: National Bureau of Economic Research [2] Einav, L., Farronato, C., & Levin, J. (2014) The Economics of Peer-to-Peer
Internet Markets Cambridge, Massachusetts: NATIONAL BUREAU OF ECONOMIC RESEARCH
[3] Farajallah, M., Hammond, R. G., & Pénard, T. (2016) What Drives Pric-ing Behavior in Peer-to-Peer Markets? Evidence from the CarsharPric-ing Platform BlaBlaCar.de SSRN Electronic Journal
[4] Jin, G. Z., & Kato, A. (2006) Price, quality, and reputation: Evidence from an online field experiment The RAND Journal of Economics,37(4), 983-1005
[5] Zervas, G., Proserpio, D., & Byers, J. (2015) A First Look at Online Rep-utation on Airbnb, Where Every Stay is Above Average SSRN Electronic Journal.
[6] Dellavigna, S., & Malmendier, U. (2004) Contract Design and Self-Control: Theory and Evidence The Quarterly Journal of Eco-nomics,119(2), 353-402
[7] Slee, T. (2013) Some Obvious Things About Internet Reputation Systems 1-13
[8] European Commission, (2017) Exploratory study of consumer issues in online peer-to-peer platform markets
[9] Luca, Michael (2016) Designing Online Marketplaces: Trust and Repu-tation Mechanisms
[10] Han, Heejeong, et al., (2016) Implication of the Fit between Airbnb and Host Characteristics Proceedings of the 18th Annual International Con-ference on Electronic Commerce e-Commerce in Smart Connected World - ICEC 2016
[11] Wei, Zaiyan, and Mingfeng Lin. (2017) Market Mechanisms in Online Peer-to-Peer Lending Management Science, vol. 63, no. 12, 2017, pp. 4236–4257
[12] Chen, Le, et al., (2015) Peeking Beneath the Hood of Uber Proceedings of the 2015 ACM Conference on Internet Measurement Conference - IMC 15, 2015
[13] Einav, Liran, et al., (2015) Peer-to-Peer Markets Cambridge, NA-TIONAL BUREAU OF ECONOMIC RESEARCH
[14] Wang, Dan, and Juan L. Nicolau, (2017) Price Determinants of Shar-ing Economy Based Accommodation Rental: A Study of ListShar-ings from 33 Cities on Airbnb.com International Journal of Hospitality Management, vol. 62, 2017, pp. 120–131
[15] Ma, Qian, et al., (2017) Pricing for Sharing Economy with Reputation ACM SIGMETRICS Performance Evaluation Review, vol. 44, no. 3, Dec. 2017, pp. 32–32
[16] Cabral, Luis, and Ali Hortacsu, (2004) The Dynamics of Seller Reputa-tion: Theory and Evidence from EBay
[17] Mickey, Ryan, (2010) The Impact of a Sellers Ebay Reputation on Price The American Economist, vol. 55, no. 2, 2010, pp. 162–169
[18] Wilhelms, Mark-Philipp, et al., (2017) To Earn Is Not Enough: A Means-End Analysis to Uncover Peer-Providers Participation Motives in Peer-to-Peer Carsharing Technological Forecasting and Social Change, vol. 125, 2017, pp. 38–47
[19] Ert, Eyal, et al., (2015) Trust and Reputation in the Sharing Economy: The Role of Personal Photos on Airbnb SSRN Electronic Journal, 2015 [20] https://www.blablacar.de/ueber-uns [21] https://www.blablacar.co.uk/faq/question/how-are-service-fees-calculated [22] https://www.blablacar.de/member-ratings [23] https://www.blablacar.de/experience-level [24] https://goo.gl/JL9Hp9
TRIPS Observations Average Price Distance(km) Frankfurt-Berlin 733 28.63 1 552 Kiel-Hamburg 1733 5.27 2 97 Köln-München 559 29.15 3 575 München-Stuttgart 1037 11.87 4 232 Berlin-Kiel 301 17.62 5 355 Köln-Hamburg 820 21.55 6 425 Düsseldorf-Hamburg 662 20.01 7 403 Hamburg-Berlin 2154 15.31 8 288 München-Berlin 716 29.28 9 585 Münster-Hamburg 876 15.5 10 282 München-Dresden 408 23.22 11 464 München-Köln 548 29.46 12 574 Köln-Berlin 573 30.16 13 573 Düsseldorf-Berlin 661 28.10 14 573 Köln-Frankfurt 1377 10.38 15 191 Hamburg-Frankfurt 587 24.97 16 493 Frankfurt-Köln 1619 10.43 17 190 Berlin-Düsseldorf 657 28.16 18 572 Berlin-Köln 661 30.13 19 579 Berlin-Frankfurt 587 24.97 20 545 Berlin-München 728 29.45 21 584 Berlin-Hamburg 2178 15.46 22 289 Berlin-Nürnberg 1148 23.55 23 446 Nürnberg-Berlin 1177 23.54 24 449 Köln-Stuttgart 633 19.77 25 374 Hamburg-Köln 864 21.38 26 431 Hamburg-Düsseldorf 671 20.09 27 401 Kiel-Berlin 302 17.83 28 355 Hamburg-Kiel 1662 4.98 29 97,2 Stuttgart-Berlin 486 31.02 30 642 Berlin-Stuttgart 462 31.41 31 632 Stuttgart-München 1682 12.06 32 233 Frankfurt-München 834 20.92 33 392 Dresden-München 705 23.43 34 459 Hamburg-Münster 948 15.13 35 282 München-Frankfurt 772 21.02 36 392 Stuttgart-Köln 726 20.06 37 373 Frankfurt-Hamburg 674 25.44 38 498 Berlin-Bremen 278 20.86 39 394 Bremen-Berlin 254 20.74 40 407
