CUSTOMER SATISFACTION:
DYNAMIC EFFECTS OF
CONCESSION YEARS
Niklas Möttö
S2909367
Supervisor: Dr. Maarten, J, Gijsenberg
Second reader: Prof. Laurens, M, Sloot
Thesis mentor: Dr. Marc Stemerding
In cooperation with: Goudappel Coffeng
JUNE 15, 2020
UNIVERSITY OF GRONINGENAbstract
In their efforts to increase sustainability, governmental agencies have attempted to increase the amount of people going from using private cars to using public transportation. Therefore, it is important to understand what makes people more satisfied with public transportation. In this paper, a random effect panel data model was created to investigate the relationship between changes in customer satisfaction and concession years. This relationship was moderated by a performance-based bonus. The research was conducted with longitudinal data from public transport concessions in the Netherlands. The results show a significant positive effect of first year on the change in customer satisfaction. Furthermore, performance-based bonus proved to have a significant positive effect on its own and as an interaction with the first year. The data was collected from OV-Klantenbarometer archives from 2006 to 2018.
Table of Contents
1 Introduction ... 1
2 Theory ... 4
2.1 Practical background ... 4
2.2 Existing research and hypothesis ... 5
2.2.1 Customer satisfaction ... 5
2.2.2 A monopolistic market ... 6
2.2.3 Limited budget ... 7
2.2.4 Delayed customer satisfaction ... 8
2.2.5 Pricing cannot be used ... 9
2.2.6 Performance-based contract in a moderating role ... 10
2.2.7 Control Variables ... 11
2.3 Hypotheses ... 12
3 Data ... 12
4 Methodology ... 13
4.1 Cross-sectional panel data model, fixed, random, and mixed effects ... 13
4.2 Model specification... 15
4.3 Measurement and definition of the variables ... 17
4.3.1 Customer satisfaction ... 17
4.3.2 Main independent variable ... 18
6.2 Analysis of results ... 28
6.3 Limitations and future research ... 30
6.4 Managerial Implications ... 32
7 Conclusion ... 33
References ... 34
1
1 Introduction
Shifting as many people as possible towards using public transport is a common goal for many
governments. To encourage such behavior, it is important to make the usage of public transport as
enjoyable as possible. If the government wants to maximize enjoyment, it must understand what
makes a good service. Once they know the answer, they will be able to purchase a service which
delivers the highest quality with the lowest price (Mouwen, 2015).
Besides issues related to sustainability, public transportation is often operated by private
companies. Shareholders and managers want to know how their companies are performing.
Therefore, there has been a steady demand for reliable ways of measuring performance. As an
answer to this demand, several different key performance indicators have been identified by
scholars. One of the most important identified KPI’s is customer satisfaction (Bhattacharya,
2013; Mouwen, 2015). Due to the importance of customer satisfaction, it has become vital to be
able to manage it and to know what make increases and decreases it. Therefore, I will study the
dynamic effect of specific years of a concession on satisfaction in public transport.
To understand satisfaction, it is important to understand what drives satisfaction. Previous studies
have been able to identify several different significant influencers on satisfaction (Mouwen, 2013;
Gijsenberg, van Heerde & Verhoef, 2015; Mouwen, 2015; Verhoef, Heijnsbroek, & Bosma,
2017). Additionally, more studies indicate that the heterogeneity of consumers results in a
varying mix of explanatory powers which influence customer satisfaction (Eboli & Mazulla,
2
satisfaction, it is time to investigate the effect of one fundamental factor, time. How will customer
satisfaction evolve throughout the first years of a concession?
When it comes to literature about concessions in public transport, a popular topic of the last two
decades has been proving the usefulness of competitive tendering versus non-tendering
procedures. As the superiority of competitive tendering has been established and most current
public transport concessions are resulting from a competitive tendering procedure (Mouwen &
Rietveld, 2013; Hensher & Wallis, 2005; Van Velde, Schipholt & Veeneman, 2008), it is time to
take a step further and see how the customer satisfaction is influenced throughout the concession.
The task of providing a high service quality in a considerable volume and simultaneously filling
the strict requirements of the governmental client is not easy, and this research will inspect this
challenge more deeply.
In continuation to the topic of competitive tendering, a possibly important aspect to look at is the
performance-based bonus. In an increasing amount of tendering procedures, other factors than
just price are gaining an increasing amount of attention due to the emergence of customer
oriented service contracts and performance based contracts (Mouwen & Rietveld, 2013; Hensher
& Wallis 2005; Hensher & Houghton, 2004). Such factors include qualitative aspects, such as
requirements regarding the equipment and personnel. Equipment related requirements are
commonly accessibility, capacity, and environmental requirements. The moderating effect of
such additions into concession contracts will be studied in this paper.
Once there is a decent overview of the source of satisfaction, it is necessary to make it
measurable. Some previous papers have been able to cut the overall satisfaction into several
subparts, which have later been used as a basis for customer satisfaction surveys (Eboli &
3
satisfaction in transportation (Anderson, Klein-Pearo & Widener, 2008). This paper will serve as
an extension to the research in satisfaction in public transport that has been made in the
mentioned papers.
I will conduct an empirical research using archival data from 2006 to 2019. The area of focus is
on bus, tram, metro or regional train concessions in the Netherlands This study will focus on
Dutch public transport, more specifically on many different concessions happening in different
Dutch regions. The OV-Klantenbarometer, will be the main source of data. This study will
contain all transport modalities available in the survey data, including buses, trains, trams, ferries
and metros.
With this study, I will contribute towards academic fields of transportation and marketing. As it
has been pointed out before, the mixture of transportation and marketing has been lacking in
terms of academic research (Gijsenberg & Verhoef, 2019). This paper is a first step to fill this
existing gap. Additionally, like Gijsenberg and Verhoef, also this paper will add to the lacking
longitudinal sustainability related longitudinal research with time series data (Iyer & Reczek,
4
2 Theory
2.1 Practical background
Public transport in the Netherlands is operated mostly by private companies which compete in
tendering procedures to obtain the rights to operate in a certain region for a given amount of
years. The tendering is conducted by the twelve different provinces and in two cases by regional
organizations - the Amsterdam transport region and the metropolitan region of Rotterdam and
The Hague.
As competitive tendering helped to reduce cost of subsidizing by 20% to 30%, another problem
emerged in terms of fears of decreasing quality due to the main target being minimization of
costs. This led to the growth in popularity of performance-based contracts (Hensher & Houghton,
2004; Hensher & Wallis, 2005). A performance-based contract is expected to encourage
companies to do better, which increases customer satisfaction. In the Netherlands, this clause in
the contract is known as a bonus/malus, where malus means a sort of a penalty for not reaching a
certain minimal level of service. Different measures are being used for determining the bonus, for
example, certain customer satisfaction elements or timeliness. As an example, in the
Groningen-Drenthe concession, the perceived friendliness of drivers as an individual factor is worth up to
10,000 euros per year (OV Bureau Groningen Drenthe, 2017). The province/municipality might
also use other methods, such as mystery shoppers, to determine whether bonus requirements are
being met. Malus requirements are usually quite loose or non-existent, for example, customer
satisfaction level of less than 5 is a quite common line for malus, but this is in practice
unrealistically low, as the lowest scores are usually around 7, except for price which is regularly a
5
being assigned to solely supervise the fulfillment of bonus requirements, and companies reward
their employees for achieving the bonus. This indicates a considerable importance of the
bonus/malus system.
The business model is partially consumer (henceforth: B2C), partially
business-to-business (henceforth: B2B), and partially business-to-business-to-government (henceforth: B2G). Ticket sales
go to the operator. This creates an interesting setting for research focused on service. As
tendering procedures are highly competitive, the winning operator will end up doing it for such a
price which will not be profitable without a sufficient amount of ticket sales. Additionally, the
operator does not get to set the price, which is set by the public transport officials. This requires a
different kind of mindset than the traditional setting where ticket sales go to the client and the
operator merely focuses on minimizing cost. This setting requires finding a balance where cost
minimizing is still a highly important factor but there is potential in attracting as many passengers
as possible. As it has been discussed above, the concession operator gets to have a monopolistic
position in the market for the coming few years, releasing pressure from offering the highest
quality. However, as it has been discussed by Bhattacharya (2013), even in a oligopolistic
situation, it is important to put in effort to increase customer satisfaction as a happy customer is
more profitable than a customer who simply has no other choice.
2.2 Existing research and hypothesis
2.2.1 Customer satisfaction
As Eboli and Mazzulla (2011), conclude in their paper, it is ultimately the customers who give
6
important factor when estimating the performance of the company. The relevance of satisfaction
as an important KPI has been established in terms of relationship to sales (Fornell, Rust, &
Dekimpe, 2010), determining customer behaviour (Mouwen, 2015), and as a frequently used
metric to satisfy the requirements of the governmental client (Verhoef, Heijnsbroek & Bosma,
2017). Even though the importance of satisfaction might be sometimes overstated, creating a
satisfaction trap (Reichheld, 1996), there is enough evidence for the importance of satisfaction.
Previous literature suggests an increase in customer satisfaction when a new concession starts
(Mouwen & Rietveld, 2013). However, that research had its main focus on the comparison
between competitive and non-competitive contracting procedures, neglecting the effect of other
important matters, such as a performance-based bonus or delayed effect. Interestingly, the same
study suggests that a new operator has a negative effect on satisfaction. This is particularly
interesting because a later research by Mouwen (2015), shows that new equipment has a
significant positive influence on customer satisfaction in some of the major segments in public
transportation (elderly and people from dense urban areas). Having a new operator would mean
even more new equipment so the positive effect should be even higher. Unfortunately, in the
dataset which was used for this research, there were not enough concessions with a new operator
to have any significant results.
2.2.2 A monopolistic market
There are some rather black and white competitive elements in public concessions. As an
example, when an operator operates in a certain region, it is not threatened by any competition,
7
a transport provided can be regarded to be in a monopolistic position with its focus on
minimizing cost and fulfilling the contractual requirements. Customers have little possibility to
change companies, with the only competing option being a private car. This means that many
people do not have an alternative. Based on this, the natural marketing strategy for such operators
is defensive value recovery (Dube & Maute, 1998). Value recovery is focused on actions such as
service guarantees and complaint management. In some regions, there is low incentive to attract
new travellers as the capacities are already pushed to the limits and the real financial incentive to
keep the customer satisfaction on a high level is the performance based bonus (Hensher &
Houghton, 2004; Hensher & Wallis, 2005). However, the idea of oligopolistic companies being
able to hide behind high switching costs in stationary markets has been challenged in several
studies (Biglaiser, Cremer & Dobos, 2013; Jones, Mothersbaugh & Beatty, 2000; Shapiro &
Varian, 1998). Even if the customer was stripped of alternatives, the company would have an
incentive to keep the travelers satisfied due to, for example, lessened costs due to fewer
complaints (Bhattacharya, Morgan, & Rego, 2016).
2.2.3 Limited budget
Due to competitive tendering procedures, the operation itself is mostly executed with a limited
budget. In terms of customer satisfaction, this is an important aspect as budget restraints could
influence the way how customer satisfaction is managed. When an operator is dealing with a
limited budget, there are some studies which indicate opposing directions of action in order to
increase profitability. Gijsenberg and Verhoef (2019) were able to discover a significant positive
effect of advertising and promotions on travelled distance. Additionally, Rust, Moorman, and
8
cost reduction or dual emphasis tend to be more successful. However, Steenkamp and Fang
(2011), argue in their research on business cycles that focus on internal process when dealing
with financial limitations is more effective than focus on advertisement. This financial limitation
is not the same as the one coming from working with a maximized efficiency, but it has some of
the same principles. Given that the study conducted by Gijsenberg and Verhoef is specifically
focused on public transport, and it is 8 years more recent, it does have a higher value in relation to
this research. It is still important to acknowledge existing schools of thought. Additionally, a
study on budget hotels shows that value for money and core products continues to play a critical
role on satisfaction even in a business with high efficiency (Rahimi & Kozak, 2017). Based on
this, whatever the approach is, the operator must first secure the fundamental service.
2.2.4 Delayed customer satisfaction
Due to a possibly “undeserved” spike in customer satisfaction in the beginning of a concession, it is important to measure the delayed effects on customer satisfaction for the following couple of
years. Previous research shows that the longer the relationship between a customer and a
company, the less variance there will be in the satisfaction of that certain customer (Bolton,
1998).. This should indicate reduced fluctuation in satisfaction after the first year of the
concession, but not the direction of the effect. Additionally, a study from Gijsenberg, van Heerde
and Verhoef (2015), indicates a significant effect caused by past satisfaction. Other studies have
been able to show how active customer relationship management has significantly improved
customer satisfaction over time (Mithas, Krishnan & Fornell, 2005; Hassan, Nawaz, Lashari, &
Zafar, 2015). In the field of public transportation, Gijsenberg and Verhoef (2019), noticed a
9
capacity issues (Anderson, Fornell, & Lehmann, 1994; Rego, Morgan, & Fornell, 2013).
Combining the findings from Bolton (1998), and Gijsenberg & Verhoef (2019), should give a
reason to expect at least a decrease in the growth of satisfaction after the first concession year.
By observing the summarized customer satisfaction results of the OV-Klantenbarometer from
previous years, a steady increase in overall satisfaction score is visible, with only price remaining
rather constant. A possible reason for this could be found in discussions held with some people
who work in the industry. What they have suggested is that companies put higher emphasis on
those elements which score low in customer surveys while high scoring elements remain stable,
which slowly increases their overall score. This type of a practice is not uncommon in the
professional environment across industries. Such a method is called management by exception
(Bittel, 1964). As the year in the reference shows, management by exception has been around for
more than 50 years already and is still regarded as a popular and effective way of quality
management. According to management by exception, management puts higher emphasis on
things that are varying most significantly from the planned course of action. Management by
exception would be in line with the assumption that customer satisfaction becomes more stable
after the initial shock from the first year.
2.2.5 Pricing cannot be used
An important marketing action which must be addressed is pricing. According to existing
research, pricing is an important tool when increasing customer satisfaction (Bhattacharya, 2013;
10
monopoly pricing. In the Netherlands, public transport pricing is conducted by
Governmental authorities. This makes sense as the monopolistic pricing of necessary public
goods has been proven to be highly problematic (Havighurst & Richman, 2011). Independent
pricing would also create tremendous friction in the national public transport network as a whole
as intercity travelling would become more unpredictable.
2.2.6 Performance-based contract in a moderating role
Even though a public transport customer is faced with a great switching cost, it does not liberate
the operator from focusing on quality of the service. As it was pointed out by Bhattacharya
(2013), a customer who remains loyal for positive reasons is more profitable than a customer who
stays loyal for negative reasons, meaning that satisfaction is an important factor for the business
even in a monopolistic situation. As for the alternatives for the customer, it is a decision between
a few transport modalities. People who do not own a private car, might be left with no alternative
choice. Therefore, the role of the government as the supervisor is significantly important. To
reduce the risk of operators cutting costs at the expense of the overall travel experience,
governmental agencies have started to take a larger role in co-creating higher quality with the
operators (Verhoef, Heijnsbroek & Bosma, 2017). The most common form of governmental
interference in the service quality is a performance based contract (Hensher & Wallis 2005;
Hensher & Houghton, 2004) As performance based contracts have become increasingly common
they have also become rather specific, for example, in the Groningen-Drenthe concession, the
perceived friendliness of drivers as an individual factor is worth up to 10,000 euros per year (OV
11
to improve the quality of the service, I expect that it will have a positive moderating effect
between the concession years and customer satisfaction.
2.2.7 Control Variables
Travel mode is one of the control variables of this study. The public transport modalities in this
research contain traditional core services of public transportation. These core services are bus,
train, subway, tram, and ferry. As bus is the most frequent mode of transport in the dataset, it will
be the one compared to the rest. Modality can have a significant effect on the results as it has
been confirmed before that in a multi-channel environment, customer perceptions can change
significantly even though the core service remains largely the same (Falk, Schepers,
Hammerschmidt, Bauer, 2007).
Another control variable is motivation to travel (commuters and students vs others). The reasons
to pick commuters and students as one group versus others is that they are mostly not paying for
their own tickets and they use public transport the most. As it was concluded before, there is a
negative relationship between travelled distance and satisfaction (Gijsenberg & Verhoef, 2019).
This might indicate a possible significant effect in the results of this study. As different elements
of the public transport market remain relatively stable throughout the years and the data is
12
2.3 Hypotheses
Based on the discussion above, following hypotheses were constructed.
H1a: First year of the concession positively influences the change in customer satisfaction. H1b: The growth of customer satisfaction after the first year is an empirical question. H2: Performance-based bonus positively moderates the relationship between concession years and customer satisfaction.
Figure 1. Conceptual model.
3 Data
Most of the data was collected from the OV-Klantenbarometer archives held by Goudappel
Coffeng. Contract related data, such as performance-based bonus was collected by reading
concession contracts and asking from region representatives by email.
Efforts to quantify the level of satisfaction towards public transport in the Netherlands have been
taken in many forms, one of them being the OV-Klantenbarometer. OV-Klantenbarometer is a
national customer survey which measures several different aspects of customer satisfaction, such
13
annually since 2006 in the same set-up. This survey will serve as an important source of
information for this research.
The surveys were distributed in varying public transport vehicles on an annual basis until 2017,
after which it became a quarterly procedure. People who received the survey were asked to rate
their experience on a scale from 1 to 10. These questions are related to a wide array of topics, all
believed to be an important factor in overall satisfaction. For this research, the most important
question is the one asking about overall satisfaction, since this gives the overview of how the
traveler felt about the trip.
The survey consists of 33 questions which are used to measure 25 different parameters. The
measurable parameters are provided in the appendix. The survey provides plenty of data on
individual level. However, the data will be aggregated on a yearly level. Analysis on individual
level is simply unnecessarily much for the sake of this research especially considering other
variables such as performance-based bonus, which does not seem logical for the individual level.
4 Methodology
4.1 Cross-sectional panel data model, fixed, random, and mixed effects
A cross-sectional panel data model is a model where the dataset consists of data from different
periods in different sectors. In this case, different years and different concessions. Panel data
models are approached in one of three ways, fixed effects, random effects, or mixed effects.
The main difference between fixed effects and random effects models is the source of variation.
14
variation within a certain sector, in this case, a certain concession. It will not take variation
between different sectors into account, making it stricter than the random effects model and less
generalizable (Hedges & Vevea, 1998). Therefore, a random effects model is seen as more lenient
and more efficient model, which is why it is usually preferred if it can be used. Some researchers
take an even stronger stance, saying that in most research scenarios, random effects provides
everything fixed effects provide and more (Bell, Fairbrother & Jones, 2019). Whether it can be
concluded that a dataset contains random effects, there are some assumptions. One assumption is
that when a dataset represents a larger population, it should be considered to contain random
effects, which in this case would suggest using random effects. To confirm the correct decision, a
Hausman test is conducted to determine whether random effects model can be chosen over a
fixed effects model (Baltagi, Presson & Pirotte, 2003).
Besides random and fixed effects, a mixed effects model will also be tested. A mixed effects
model is a mix between fixed effects and random effects (Bates, Mächler, Bolker, Walker, 2014).
If a mixed effects model is used, there needs to be a decision made about which variables are
generating random effects and which are generating fixed effects. This could make sense as
concession years could be a representor of a larger population, whereas more concession specific
variables, such as modality, could be a fixed effect variable.
Apart from the Hausman test, there does not seem to be a widely accepted consensus about the
selection criterion between these three methods. Therefore, the best thing to do is to create all
three models and see which method performs the best. Previous literature on the topic of
choosing between fixed and random effects shows that the are no great differences and the main
argument for the decision should be what kind of inferences the researcher wants to make
15
4.2 Model specification
As the variables of the final model are known, a model can be estimated. Whether it will be a
fixed, random, or mixed effects model, it will be fully pooled so that the parameters of the model
can be used to estimate all concessions. A complementary model with overall satisfaction as the
dependent variable was also produced for a more complete overview of the market.
∆𝑆𝑎𝑡𝑖𝑗= 𝛽0+ 𝛽1𝐹𝑖𝑟𝑠𝑡+ 𝛽2𝑆𝑒𝑐𝑜𝑛𝑑+ 𝛽3𝑇ℎ𝑖𝑟𝑑+ 𝛽4𝑓𝐹𝑖𝑟𝑠𝑡∗𝐵𝑜𝑛+ 𝛽5𝑆𝑒𝑐𝑜𝑛𝑑∗𝐵𝑜𝑛+ 𝛽6𝑇ℎ𝑖𝑟𝑑∗𝐵𝑜𝑛+ 𝛽7𝐵𝑜𝑛+ 𝛽8𝐵𝑢𝑠+ 𝛽9𝐶𝑜𝑚+ 𝜀
Where:
∆𝑆𝑎𝑡𝑖𝑗 = Change in satisfaction between t and t-1 during a certain year in a certain concession 𝛽0 = Intercept
𝛽1𝐹𝑖𝑟𝑠𝑡 = First year of the concession 𝛽2𝑆𝑒𝑐𝑜𝑛𝑑 = Second year of the concession 𝛽3𝑇ℎ𝑖𝑟𝑑 = Third year of the concession
𝛽4𝑓𝐹𝑖𝑟𝑠𝑡∗𝐵𝑜𝑛 = Interaction between first year and performance bonus 𝛽5𝑆𝑒𝑐𝑜𝑛𝑑∗𝐵𝑜𝑛 = Interaction between second year and performance bonus 𝛽6𝑇ℎ𝑖𝑟𝑑∗𝐵𝑜𝑛 = Interaction between third year and performance bonus 𝛽7𝐵𝑜𝑛 = Performance bonus
𝛽8𝐵𝑢𝑠 = Bus
16
While the model above is the basic version describing the relationship, there are some aspects to
take into consideration depending on whether it is a fixed or random effects model. If it is a
random effects model, the following equation shows what happens within the statistical software
while creating the estimations. The following discussion is largely inspired by the work from
Bell, Fairbrother, and Jones (2019).
The time varying variable receives two parameters, one for within effects, 𝛽1𝑊, and one for
between effects, 𝛽2𝐵. The separation between the within and between effects can be left out
when the assumption of them being equal holds. The time invariant variable receives a parameter
for between effects (𝛽3). In the case of this research, this model would look longer due to a
higher number of variables, but the idea remains the same. The random effects are represented as
𝑣𝑖0 and 𝑣𝑖1. The former is attached to the intercept while the latter is attached to the within slope.
This is done to account for heterogeneity in the within effect of 𝑥𝑖𝑡 across concessions.
Random effects formula
𝑦
𝑖𝑡= 𝜇+ 𝛽1𝑊(𝑥𝑖𝑡− 𝑥̅𝑖) + 𝛽2𝐵𝑥̅𝑖 + 𝛽3𝑧𝑖+𝑣𝑖0 + 𝑣𝑖1 (𝑥𝑖𝑡− 𝑥̅𝑖)+ 𝜀𝑖𝑡0
As for fixed effects, again giving credit to Bell, Fairbrother, and Jones (2019), the between effects
would not be considered as a fixed effects model is only interested in the within effects. Like the
random effects model, also the fixed effects model makes use of mean-centering in the within
effects, making the estimations based on deviations of the mean over time, which makes it less
useful compared to random effects as it cannot describe relationships with independent variables
which do not change over time. Therefore, using fixed effects leads to a loss of important
17
Fixed effects formula
(𝑦
𝑖𝑡−𝑦̅
𝑖) = 𝛽1(𝑥
𝑖𝑡− 𝑥̅𝑖) + (𝜖
𝑖𝑡)
4.3 Measurement and definition of the variables
4.3.1 Customer satisfaction
Yearly level aggregation of the self-reported overall satisfaction scores. The average score of all
respondents in a certain concession area in a certain year. Customer satisfaction ratings are taken
from the OV-Klantenbarometer. The OV-Klantenbarometer report contains several different
aspects of customer satisfaction, one of them being the overall satisfaction of the traveler. In this
research, the overall satisfaction will be the one that represents the variable. In some concession
contracts it can be seen from the performance-based contract measures that it is other factors than
the pure overall satisfaction which is seen to be more important. However, overall satisfaction is
the only one which can be reasonably argued above all else, without the need of evaluating the
parameter of choice against all other parameters.
The variable will be computed to represent the difference between the current and the previous
year, in other words, creating a first difference variable. The first difference variable is used
because it represents better whether there is an increase in satisfaction in a certain year.
Hypotheses 1a and 1b are focused on the growth of customer satisfaction. These questions could
not be answered with overall satisfaction as the dependent variable as it does not take into
account the growth or decline in relation to the previous year. Therefore, first difference gives a
18
As the dependent variable will be the increase in customer satisfaction in a specific year of the
concession, I will not be able to measure 2006 or the first years of some concessions which
entered the system during the 2006-2018 period. The loss of one possible year of data is
unfortunate but necessary.
4.3.2 Main independent variable
The independent variable will be the sequential year of a concession. This is a series of dummy
variables which takes the value of 1 for a specific year of a concession. This will allow me to
have an estimate for several individual years of the concession. In this research, I will focus on
the first three years.
4.3.3 Moderator
The moderator, performance-based contract, will be a dummy variable. The dummy will
represent the existence of a performance-based clause in the contract. This data will be collected
from differing sources, for example, going through concession contracts and emailing regional
19 4.3.4 Control Variables
Several different control variables have been identified. As there are different transport modalities
present, it is important to control for that. Therefore, I will include a dummy variable to indicate
whether it is a bus or not. The reason to single out only bus is that it is the most common
modality in the dataset, especially because the national train network was only introduced rather
recently to the OV-Klantenbarometer.
Motivation for travelling will represent the percentage of commuters and students in the dataset.
Therefore, the value will be between 0 and 1.
4.4 Statistical validity
Besides the models, the dataset will be tested for many issues regarding statistical validity, such
as heteroscedasticity, multicollinearity, and non-normality. Problems arising from these issues
might require some changes to what has been written above.
4.5 Preliminary statistics
Table 1 shows the descriptive statistics and correlations between variables. None of the
correlations are too high so multicollinearity is not an issue. The dataset contains 1051
observations from a total of 106 different concessions. This table contains two version of the
dependent variable. First is satisfaction first difference and the second is satisfaction.
The dependent variable, satisfaction first difference, was tested for normality with a Shapiro-Wilk
20
variable can be assumed to have a normal distribution. Therefore, it will not be necessary to
adjust for nonnormality.
A simple model without interactions was created to run a Breusch-Pagan test for
heteroscedasticity, which was significant with a test statistic of BP=27.45 and a p-value of < 0.00.
This means that there are no issues with heteroscedasticity in the dataset. The variance of the
dependent variable remains stable throughout the dataset.
Table 1. Descriptive statistics.
Figures 2 and 3 show how satisfaction evolves throughout the years. In figure 2, the plot shows
that overall average satisfaction increases with time. In figure 3, the average satisfaction in
relation to the last year of the previous concession is plotted. Figure 3 shows the effect of first
Mean Std. Deviation Satisfaction, First Difference Satisfaction First Year Second Year Third
Year Bus Commuters
Performance Bonus Satisfaction, First Difference 0.05 0.19 1 Satisfaction 7.43 0.31 0.320** 1 First Year 0.1 0.30 0.124** -0.041 1 Second Year 0.11 0.31 0.012 -0.013 -0.11** 1 Third Year 0.1 0.30 0.002 -0.017 -0.11** -.115** 1 Bus 0.6 0.49 -0.046 0.107** 0.045 0.044 .075* 1 Commuters 0.38 0.11 -0.023 -0.024 0.017 0.036 -0.042 0.019 1 Performance Bonus 0.42 0.49 0.068 .174** .102** .083** 0.054 .180** 0.048 1
** Correlation is significant at the 0.01 level (2-tailed).
21
year having the strongest increase in satisfaction while it keeps increasing after the beginning.
These figures are important to justify the usage of a first difference variable as the dependent
variable to control for the trend of increasing satisfaction.
22
Figure 3.Satisfaction during concession years in relation to last year of previous concession
5 Results
5.1 Method Selection
The first step was to decide upon the most suitable model. The three options are fixed effects,
random effects and mixed effects models. The mixed effects model was not suitable for this
dataset, due to singularity issues arising from the complexity of the model. Therefore, the
statistical software was not able to estimate the confidence intervals and reported some of the
23
Both fixed effects and random effects could be created. Theoretical reasoning suggested that
random effects would be more suitable as this dataset is an aggregated compilation from a larger
population. Furthermore, a Hausman test was conducted to have statistical evidence for the
decision between the two. The Hausman test resulted in a p-value of 0.78, which means that
random effects could be used. The decision to use random effects is also supported by a higher
adjusted r-squared, which the statistical software reported as a negative value in a fixed effects
model.
5.2 Testing hypotheses
To test the hypotheses, six different random effect models were created. Models 1, 2, and 3 are
the main models as they represent the first difference satisfaction. These models are found in
table 2. Models 4, 5, and 6 are complementary models with overall satisfaction being the
independent variable. The latter models perform better in terms of a higher r squared, meaning
that changes in the dependent variable are more explained by the independent variables than in
table 2. However, models 4, 5, and 6 do not exactly measure what needs to be measured, which is
the change in satisfaction in relation to previous year, giving it a supporting role. First and fourth
models are without interaction effects. Second and fifth are with an interaction between first year
and performance bonus. Third and sixth models with an interaction between first, second, and
24 5.2.1 Models 1, 2, and 3
Table 2. Output of random effects models, Satisfaction First Difference as dependent variable.
model 1 model 2 model 3
Intercept 0.061** 0.064** 0.064**
First Year 0.081*** 0.011 0.013
Second Year 0.016 0.016 -0.001
Third Year 0.013 0.013 0.042
Interaction First Year and Bonus 0.113* 0.109*
Interaction Second Year and Bonus 0.031
Interaction Third Year and Bonus -0.061
Bus -0.021 -0.021 -0.021 Commuters -0.053 -0.044 -0.05 Performance Bonus 0.022* 0.012 0.016 R-Squared 0.023 0.030 0.034 Adjusted R-Squared 0.015 0.021 0.023 Chi Squared 18.428 24.268 27.307 P-value 0.005 0.001 0.001 Signif. codes: 0.01 *** 0.05 ** 0.1 *
First year of the concession has a significant positive effect on the dependent variable, indicating
that customer satisfaction does increase during the first year of the concession. When the
interaction effect is introduced, the first year variable loses its significance and only has a
significant influence when it interacts with performance based bonus. The same thing happens to
the performance bonus variable. However, the combined effect of first year and performance
bonus is considerably higher than any other parameter, explaining most of the effect when the
25
Second and third year do not have a significant effect on the change in satisfaction. The
parameters are considerably closer to 0 than first year, which gives a slight indication of a
slowing increase in satisfaction, but this cannot be concluded since the significance is on a low
level. Furthermore, the p-value for second year increases from 0.4 to 0.9 when the interaction
effect is introduced. Both control variables, bus and commuters, are insignificant. Both have a
negative sign, which in the case of commuters is against the reasoning of why it was introduced
to the model.
Even though all the models are overall significant, their explanatory power is low. The adjusted
R-squared implies that the independent variables only explain few percentages of the dependent
variable. This means that interpretations should be taken with some precaution, but they will
26 5.2.2 Models 4, 5, and 6
Table 3. Random effects with Overall satisfaction as dependent variable.
Model 4 Model 5 Model 6
Intercept 7.11*** 7.12*** 7.12***
First Year -0.08*** -0.08** -0.08**
Second Year -0.05** -0.05** -0.07**
Third Year -0.04* -0.04* -0.05
Interaction First Year and Bonus -0.00 -0.00
Interaction Second Year and Bonus 0.04
Interaction Third Year and Bonus -0.04
Bus 0.09** 0.09** 0.09** Commuters 0.33*** 0.33*** 0.33*** Performance Bonus 0.27*** 0.27*** 0.27*** R-Squared 0.848 0.848 0.874 Adjusted R-Squared 0.847 0.847 0.845 Chi Squared 4902.79 4882.03 4844.02 P-value <0.00 <0.00 <0.00 Significance codes: 0.01***, 0.05**, 0.1*
As mentioned earlier, due to the weak performance of models 1, 2, and 3, models 4, 5, and 6 were
created. In table 3, first year dummy has a significant negative effect on overall satisfaction. This
does indicate a dip in satisfaction when entering a new concession. However, the following years
show that the negative parameter climbs closer towards 0, indicating an instant growth in
satisfaction after the dip. Furthermore, the difference between first and second year is greater than
the difference between second and third year, indicating that the growth does in fact start slowing
27
The combined inspection of table 2 and table 3 shows that the importance of performance bonus
is significant. In both tables, performance bonus has a significant positive influence, making it the
only variable besides the intercept which has the same sign in both tables. Table 3 was not able to
support the interaction effect which was seen in table 2. This could be because the year dummies
became negative in table 3, giving them the opposite effect than performance bonus. Therefore, it
seems that the satisfaction level as such does not depend on a performance bonus but the speed of
the increase after the dip in satisfaction does strongly depend on the bonus being present in the
contract.
Bus and Commuters variables became significant in table 3 with a positive sign. Combining this
with the knowledge that they have no significant contribution towards the change in satisfaction,
the most likely reason is that concessions which already had a high satisfaction level were
operated by buses and had a high level of commuters and students on board.
5.3 Robustness check
The VIF scores resulted in all values being below 3. The VIF test was conducted for model 3 as it
included all variables. This means that there are no issues with multicollinearity. The same
models were ran using a fixed effects method. This resulted in similar signs of individual
variables, except for “Year Two”. This variable has a negative sign in random effects models when interaction is present but stays positive in three first models in fixed effects. For models 4,
5, and 6, fixed effects gave similar signs but with a lower significance. Furthermore, the
28
6 Discussion
6.1 OverviewPollution and congestion are some of the major issues facing most countries in the coming future.
Therefore, it is vitally important to accelerate the switch from private vehicles to public
transportation. For this to be possible, a deeper understanding is required of what makes people
prefer public transportation. Previous findings show that satisfaction of public transportation
users has been an important indicator for guiding people to leave their private cars. In this paper,
random effects panel data models were created to further investigate the dynamic effects of
concession years on satisfaction. The resulting model show that customer satisfaction increases
most rapidly in the very beginning of a new concession. Furthermore, when the concession makes
use of a performance-based bonus, the slope of the increase in satisfaction becomes even steeper.
In the following is a more detailed discussion about the results and implications.
6.2 Analysis of results
Hypotheses 1a is confirmed while there is some strong indication about the truthfulness of 1b.
Performance-based bonus seems to help boost satisfaction, especially in the beginning. This
seems to be in line with previous research (Hensher & Houghton 2004; Hensher & Wallis,
2005). This might not come as news to transport officials since performance bonuses have been
increasing in popularity in public transport concessions in previous years. This could explain the
relationship between performance bonus and overall satisfaction as later years have a higher rate
of satisfaction, but it would not explain the positive significant influence towards the first
29
First year does seem to boost customer satisfaction the most. However, entering a new concession
does seem to cause a slight dip before it starts increasing. This is partially in line with research
conducted by Mouwen (2015), who did not report a dip in customer satisfaction when entering a
new concession but also concluded that the beginning of a concession does increase satisfaction.
This additional boost to satisfaction could come from updated equipment, but it does not explain
why satisfaction continues growing after the beginning. Despite constant growth, there is a visible
decrease in the intensity of the growth after the first year. This is in line with previous research
from Bolton (1998), showing how the variance in customer satisfaction decreases the longer the
customer relationship lasts. In the context of public transport, it could be argued that the
beginning of a new concession is not a beginning of a customer relationship as the same person
has already been using the same service before the concession started. However, as the new
concession comes with updated equipment and sometimes with an updated scheduling scheme, it
can be seen as a partial rebirth for the customer relationship. This would especially be the case if
the operator is not the same as in the previous concession.
As the two major differences for customers at the start of a new concession is new equipment
(sometimes new operator), and sometimes a new scheduling scheme, the underlying reasons
could be searched from there. First, customers get used to these changes, which is why their
satisfaction will not keep on rising at the same pace. Same can be also said about employees who
might also experience a period of more energetic approach towards customer service as they are
operating with new equipment. Previous research shows a significant link between employee
wellbeing and customer satisfaction (Bulgarella, 2005). Working with updated equipment could
be a part of the reason. A practical explanation to what might influence the continued increase in
30
punctuality and tidiness. These functions can always be improved while they are universally
common reasons for complaints in the public transport industry.
Commuters seem to have a weak effect on growth of satisfaction while it simultaneously seems to
have a strong positive effect on overall satisfaction. Same goes for buses. This would indicate that
concessions where satisfaction was already on a high level before the start of measurements had a
high degree of commuters and/or operated by buses. Therefore, it would be wrong to say that
commuters are more dissatisfied than others or buses are a less preferred modality. The positive
relationship between commuters and satisfaction provides some interesting insights. First,
commuters and students travel in general at least 10 times per week, accumulating a substantial
number of travelled kilometers. This could be a conflicting result to what Gijsenberg and Verhoef
(2019), noticed about the negative effect of travelled distance to customer satisfaction. However,
a unique element about commuters and students is that in most cases they travel with a discount
or their trips are paid by the employer. Combining this to the fact that price is, by a substantial
margin, the element which consistently has the lowest satisfaction score, bringing down the
overall satisfaction, people who do not need to pay the full price may report a higher overall
satisfaction.
6.3 Limitations and future research
This research only contains concession from the Netherlands. Therefore, these results might not
be fully generalizable globally. This might be especially the case in countries where the public
31
Performance based bonuses were measured as a dummy variable based on the sole existence of
the bonus. As there are different types of performance bonuses which might differ in the strength
of effect, some important insights might be lost.
OV-Klantenbarometer surveys were handed out in differing external conditions. Such variance in
external conditions can be seen in exceptionally busy days or a day when there was a unusual
problem, for example, air conditioning of a bus was broken. In such cases, much of the data does
not represent the overall situation in that concession. The size of the sample should control for
that, but full control of external conditions is not possible.
This research has raised some important questions which should be further investigated. The first
question is the following: What drives the continuous growth of customer satisfaction? Transport
providers in the Netherlands seem to have found an answer to a question which is regularly in
minds of marketeers in companies across most industries. They have managed to increase
customer satisfaction annually for over a decade. Much of the reasoning in this paper is focused
on the first years of the concession, which was rightfully expected to have the strongest effect,
but these explanations are not valid to explain how customer satisfaction keeps on growing after
the very beginning of the concession. Along with this topic, future research could investigate the
relationship between employee satisfaction and customer satisfaction in public transport and how
employee satisfaction could be most efficiently increased.
What are the most important aspects of a performance bonus in the context of public transport?
This information is important for transport officials who oversee providing quality transportation
efficiently. It is important to understand which factors can be influenced by the transportation
providers and which cannot. As an example, in the case of Dutch concessions, some performance
OV-32
Klantenbarometer, while some bonuses are about punctuality. Some of these
OV-Klantenbarometer elements are out of control for transport providers, such as chance to find a
seat. If the transport provider receives a penalty because of such a factor, it can have a
demotivating effect and can be generally perceived as unfair. Therefore, this is an important
question for future research.
6.4 Managerial Implications
Since performance-based bonuses are significantly important, managers should understand them
well. By having more extensive data on performance-based bonuses from previous concessions,
transportation companies could have an overview on which bonuses are more likely attainable
and which conditions require a higher pay-out. Furthermore, as it was mentioned earlier, transport
providers who operate in the Netherlands seem to have a process in which they annually check
the weakest performing elements of their score from the OV-Klantenbarometer survey and focus
on those elements more in the coming year. This could be an important part of the reason why
satisfaction keeps marginally rising in general. Managers in other markets could perhaps
investigate the possibility of adding elements of management by exception into their own
processes.
For policy makers, it seems that a performance-based bonus acts as a powerful tool to increase
customer satisfaction. Therefore, it is worth giving the bonus clause more attention during a
tendering procedure than some policy makers might have done until now. Some policy makers
could perhaps even consider paying higher bonuses to see if it boosts satisfaction even more.
33
reaching the upper limits in satisfaction in about a decade. This will call for modifications in
bonus conditions as increases in satisfaction become increasingly difficult and penalizing minor
decreases would become rather unfair.
7 Conclusion
In conclusion, in a continuously urbanizing environment, public transportation is vitally
important to create a sustainable future where pollution and congestions are minimized. To
maximize the benefits of public transport, governments need to find means to maximize the usage
of public transport while the need for private vehicles lessens. For this to happen, people must be
satisfied with the service.
This study contributed towards an increased understanding of which factors drive customer
satisfaction and how it can be increased. The results show that satisfaction increases more rapidly
during the beginning of the concession, while it continues increasing throughout the concession.
Furthermore, performance-based bonus has a significant positive effect on satisfaction, which
34
References
Anderson, E. W., Fornell, C., & Lehmann, D. R. (1994). Customer satisfaction, market share, and profitability: Findings from Sweden. Journal of marketing, 58(3), 53-66.
Anderson, S., Klein-Pearo, L., & Widener, S. (2008). Drivers of Service Satisfaction: Linking Customer Satisfaction to the Service Concept and Customer Characteristics. Journal of Service Research, 10: 365-381.
Andreassen, T. W. (1995). (Dis) satisfaction with public services: the case of public transportation. Journal of Services Marketing. 9(5), 30–41.
Baltagi, B. H., Bresson, G., & Pirotte, A. (2003). Fixed effects, random effects or Hausman– Taylor?: A pretest estimator. Economics letters, 79(3), 361-369.
Baltes, M.R., 2003. The importance customers place on specific service elements of bus rapid transit. J. Public Transport. 6(4), 1–19.
Bates, D., Mächler, M., Bolker, B, M., Walker, S, C., 2014. Fitting linear mixed-effects models using lme4. Journal of Statistical Software.
Bell, A., Fairbrother, M., & Jones, K. (2019). Fixed and random effects models: making an informed choice. Quality & Quantity, 53(2), 1051-1074.
Bhattacharya, 2013 Switching Costs and Sustained Competitive Advantage. International Journal of Business and management invention. 2(9): 101-111.
Bittel LR (1964). Management by exception: systematising and simplifying the managerial job. McGraw-Hill, New York.
Bolton, 1998. A Dynamic Model of the Duration of the Customer’s Relationship with a Continuous Service Provider: The Role of Satisfaction.
Brons, M., Rietveld, P., 2009. Improving the quality of the door-to-door rail journey: a customer-oriented approach. Built Environ. 35 (1), 30–43.
35
C. Shapiro and H. R. Varian, (1998). Information rules –a strategic guide to the network economy. Boston: Harvard Business School Press.
Dell’Olio, L., Ibeas, A., Cecin, P., 2010. Modelling user perception of bus transit quality. Transp. Policy 17, 388–397.
Diana, M., 2012. Measuring the satisfaction of multimodal travelers for local transit services in different urban contexts. Transp. Res. Part A 46, 1–11.
Dube, L., Manfred, M. F., (1998). Defensive strategies for managing satisfaction and loyalty in the service industry. Psychology & Marketing, 15(8), 775-791.
Eboli & Mazzulla, 2011. A methodology for evaluating transit service quality based on subjective and objective measures from the passenger’s point of view.
Eboli, L., Mazulla, G., 2009. A new customer satisfaction index for evaluating transit services quality. J. Public Transport. 12 (3), 21–38.
F. F. Reichheld, The loyalty effect(Boston: Harvard Business School Press, 1996).
Falk, T; Schepers, J; Hammerschmidt, M; Bauer, H H. (2007) Identifying cross-channel dissynergies for multichannel service providers. Journal of Service Research, 10(2), 143-160.
Fornell, Claes, Roland T. Rust, and Marnik G. Dekimpe (2010), “The Effect of Customer Satisfaction on Consumer Spending Growth,” Journal of Marketing Research, 47 (1), 28–35. G. Biglaiser, J. Cremer, & G. Dobos, (2013). The Value of Switching Costs, Journal of Economic Theory, 148(3): 935-952.
Gijsenberg, M. J., Van Heerde, H. J., & Verhoef, P. C. (2015). Losses loom longer than gains:
Modeling the impact of service crises on perceived service quality over time. Journal of
Marketing Research, 52(5), 642-656.
Gijsenberg, M. J., & Verhoef, P. C. (2019). Moving Forward: The Role of Marketing in Fostering Public Transport Usage. Journal of Public Policy & Marketing, 38(3), 354-371.
36
Havighurst, C, C., Richman, B, D., 2011. The Provider Monopoly Problem in Healthcare. Oregon Law Review. 89, 847-884.
Hedges, L. V., & Vevea, J. L. (1998). Fixed-and random-effects models in meta-analysis. Psychological methods, 3(4), 486.
Hensher & Wallis, 2005. Competitive tendering as a contracting mechanism for subsidising transportation: The bus experience.
Hensher, David A., and Erne Houghton (2004), “Performance-Based Quality Contracts for the Bus Sector: Delivering Social and Commercial Value for Money,” Transportation Research Part B: Methodological, 38 (2), 123–46.
Iyer, E. S., & Reczek, R. W. (2017). The intersection of sustainability, marketing, and public policy: introduction to the special section on sustainability. Journal of Public Policy & Marketing, 36(2), 246-254.
M. A. Jones, D. L. Mothersbaugh, and S. E. Beatty, (2002). Why customers stay: measuring the underlying dimensions of service switching costs and managing their differential strategic outcomes, Journal of Business Research, 55, 441–450.
Mattila, A. S., & O'Neill, J. W. (2003). Relationships between hotel room pricing, occupancy, and guest satisfaction: A longitudinal case of a midscale hotel in the United States. Journal of Hospitality & Tourism Research, 27(3), 328-341.
Mithas, S., Krishnan, M. S., & Fornell, C. (2005). Why do customer relationship management applications affect customer satisfaction?. Journal of Marketing, 69(4), 201-209.
Mouwen, 2013. Does competitive tendering improve customer satisfaction with public
Mouwen, Arnoud (2015), “Drivers of Customer Satisfaction with Public Transport Services,” Transportation Research Part A: Policy and Practice, 78 (8), 1–20.
Nerlove, M. (2005). Essays in panel data econometrics. Cambridge University Press.
Netherlands: Central Planning or Functional Specifications? Transportation Research Part A: Policy and Practice42(9): 1152-1169.
37
Rahimi, R., & Kozak, M. (2017). Impact of customer relationship management on customer satisfaction: The case of a budget hotel chain. Journal of Travel & Tourism Marketing, 34(1), 40-51.
Rego, L. L., Morgan, N. A., & Fornell, C. (2013). Reexamining the market share–customer satisfaction relationship. Journal of Marketing, 77(5), 1-20.
Rust, R. T., Moorman, C., Dickson, P. R. (2002). Getting return on quality: revenue expansion, cost reduction, or both? Journal of Marketing, 66(4), 7-24.
Saito, H., Nakayama, D., & Matsuyama, H. (2009). Comparison of landslide susceptibility based on a decision-tree model and actual landslide occurrence: the Akaishi Mountains, Japan.
Geomorphology, 109(3-4), 108-121.
Steenkamp, J. B. E., & Fang, E. (2011). The impact of economic contractions on the effectiveness of R&D and advertising: evidence from US companies spanning three decades. Marketing
Science, 30(4), 628-645.
Tyrinopoulos, Y., Antoniou, C., 2008. Public transport user satisfaction: variability and policy implications. Transp. Policy 15, 260–272.
Van de Velde, D., Veeneman, W., & Schipholt, L. L. (2008). Competitive tendering in The
Netherlands: Central planning vs. functional specifications. Transportation Research Part A:
Policy and Practice, 42(9), 1152-1162.
Verhoef, P. C., Heijnsbroek, M., & Bosma, J. (2017). Developing a service improvement system
38
Appendix
Ov-Klantenbarometer elements
- Chance of finding a seat
- station at beginning end of the trip
- punctuality
- ease of getting in
- friendliness of personnel
- driving style of the chauffeur
- quickness of the trip
- cleanliness
- noice
- atmosphere
- design of the vehicle
- information at the stop/station
- travel information in the vehicle
- ease of purchasing a ticket or loading saldo
- price
- other passengers
- level of stress/relaxation
- frequency
- transfer time
- information about delays
- ease of use of OV-chip card
- safety overall
39 - safety at the station
- overall score of the trip
Model 1
estimate Std. Error z-value P-value (>|z|)
Intercept 0.061926 0.025673 2.4121 0,015861* First Year 0.081448 0.023485 3.4681 0,0005241*** Second Year 0.015631 0.021309 0.7336 0.4632208 Third Year 0.012897 0.021172 0.6092 0.5424146 Bus -0.020722 0.014656 -1.4139 0.1573926 Commuters -0.052607 0.058559 -0.8984 0.3689929 PerBon 0.022894 0.013567 1.6874 0.0915187
Total Sum of Squares 28.344
Residual Sum of
Squares: 27.692
R-Squared 0.023024 Adj. R-Squared: 0.015528
Chi Squared 18.4287 DF 6
40
Model 2
estimate Std. Error z-value P-value (>|z|)
Intercept 0.064138 0.025613 2.5042 0,01227*
First Year 0.011014 0.037608 0.2929 0.76962
Second Year 0.016545 0.021248 0.7787 0.43617
Third Year 0.013308 0.021109 0.6304 0.5284
Interaction First Year And
Bonus 0.113787 0.047546 2.3932 0,0167*
Bus -0.021758 0.014618 -1.4884 0.13664
Commuters -0.044455 0.058482 -0.7601 0.44717
Performance Bonus 0.012808 0.014168 0.904 0.36599
Total Sum of Squares: 28.344
Residual Sum of
Squares: 27.49
R-Squared: 0.030136 Adj. R-Squared: 0.021443
Chisq: 24.2676 DF 7
41
Model 3
estimate Std. Error z-value P-value (>|z|)
Intercept 0.064368 0.025793 2.4956 0,01257*
First Year 0.013008 0.037781 0.3443 0.73062
Second Year -0.001181 0.031289 -0.0377 0.96989
Third Year 0.042674 0.029229 1.46 0.14429
Interaction First Year And
Bonus 0.109511 0.048263 2.2691 0,02327*
Interaction Second Year and
Bonus 0.031355 0.042596 0.7361 0.46166
Interaction Third Year and
Bonus -0.061056 0.042289 -1.4438 0.1488
Bus -0.02154 0.01461 -1.4743 0.14039
Commuters -0.050368 0.058562 -0.8601 0.38975
Performance Bonus 0.016888 0.016449 1.0267 0.30456
Total Sum of Squares: 28.344
Residual Sum of
Squares: 27.384 R-Squared: 0.033867 Adj. R-Squared: 0.022705
Chisq: 27.3073 DF 9