Russian karting market

Размещено на http://www.allbest.ru/

Introduction

Karting is the sport that is considered to be “the youngest” part of motorsport that gives young drivers an opportunity to learn the basics of racing, to improve their physical abilities and, most importantly, to start their career in motorsport that may lead them to Formula 1 (F1) world championship - most prestigious race championship in the world. Almost all F1 drivers live like celebrities and receive astronomic salaries. However, this is not how it works at the start of each driver's career, as the karting market in generally works in the different way. All drivers irrespective of their age act as clients, while race teams act as suppliers of racing services. These services include the transportation of cars to competitions, engineering and mechanical work with cars during competitions, coaching work, administrative services and many others. Drivers (or their parents in case of young drivers) usually pay very large amounts for these services.

Speaking about Russian karting market, we can notice that it is quite young and not well-developed. It emerged from the Soviet Union karting market that had been fully controlled by special government organizations (such as DOSAAF). Main players in this market are race teams ruled by the most successful karting drivers, coaches and engineers of the Soviet time. All these people are high qualified specialists in their area, but usually they are not good specialists in management. Methods which they use in the team management are mostly old-fashioned and seemed to be not efficient in the modern time. For example, the most popular advertising method is the word of mouth advertising which, for sure, can be also efficient in some situations. However, the heavy reliance on this method only makes to think about the probable efficiency of modern marketing methods.

Therefore, my hypothesis states that the usage of such methods will be efficient and will help race teams to increase their revenues. In my work I am going not only to test this hypothesis, but also to determine the most efficient marketing instruments which are available in the Russian karting market.

To test this hypothesis, I will use 2 methods. First, I will use the Ordinary Least Squares (OLS) model, present its main implications and state the main problems of this model that make its usage not appropriate for this case. Next, I will use the Structured Equation Modelling (SEM) approach which will be much more suitable due to the possibility to account for complex interrelationships between variables and for latent (unobservable) variables. I will present main findings of this method and will compare them to OLS results.

The main goal of my work is to develop the efficient management method that can be easily used by each race team in the Russian karting market. This method should help to choose the most appropriate marketing instruments and to split the team's budget among these instruments in the optimal way. Therefore, in such way it should attract more new drivers to the team and increase its revenues. I am going to test this method not only in theory, but also in practice by incorporating its finding into the business model of my race team - MSKS K. P. Kart Racing Team (MSKS K.P.). However, as the full implementation of this method takes a lot of time (more than 1 year, for sure), its testing in real business will majorly stay outside of this paper. Here I will provide only preliminary results of changes in the non-financial performance of MSKS K.P. as compared before and after the introduction of this method.

In this paper I will first provide the Russian karting market analysis, present its main features and problems, explain what do I mean by race team management efficiency and why sales and marketing processes are of such a great importance for the race team management. In the second chapter I will turn to the development of the model to evaluate efficiency of different methods. First, I will describe the main variables which probably influence sales. Second, I will use OLS model to determine the significance of these variables and then state some serious problems of this model. Third, I will describe the method of Structural Equation Modelling (SEM) in general using an appropriate scientific literature. In the next chapter I will come to the process of SEM testing by introducing my data sources and data collection methods. Next, I will develop SEM model which can be used to study the functioning of the race team and to evaluate the efficiency of different marketing instruments In the final chapter, I will present the main results of SEM, compare them to the results of OLS and explain their possible usage in the race team management. That is, according to the main goal of my work, I will develop the race team management method. Next, I will introduce some preliminary results of this method implementation for the MSKS K.P. race team. I will also name some possible actions which can be done in the further work to improve the reliability of current results and to extend the current analysis. Finally, I will make a conclusion about the completed work.

1. Russian karting market analysis

In this chapter I will provide the market analysis which is required to understand the mechanics of this market and to develop a suitable model to evaluate management efficiency. I will start from some background information about the market which will help to better understand all business processes of race teams. Second, I will present the most essential market features and characteristics for my further analysis. These characteristics are: asymmetric information, similar cost structure, high mobility of labour, excess supply of racing equipment and lack of drivers. Next, assuming that stated above features are true, I will define the management efficiency as the efficiency in sales and marketing processes. Starting from this moment I will use this definition of efficiency in my work. Finally, I will turn to the next chapter where I will present methods to evaluate management efficiency.

Background Information about the Market

This section is aimed to give you a better understanding of the Russian karting market mechanics. First, I will introduce the main agents in this market and their behaviour. Next, I will describe the interaction of 2 types of agents in which we are the most interested - race teams and drivers.

The demand side in this market is represented by drivers. The market is quite small in the number of drivers (about 1 000 drivers by rough estimate), but rather big in the money value (about 5 bln rubles per year by rough estimate) as the participation in competitions is very expensive. Karting drivers can be of different ages as there exist different karting categories. Children can participate in competitions from the age of 6, while in some elder categories the participation is restricted only to adults older than 30 years. Drivers derive utility from the participation in karting competitions. They do not receive any money prizes or other types of material awards except the race cups. Drivers (or their parents in case of young drivers) pay for the participation in these competitions. Participation costs usually include: administrative fees, transportation costs, living costs, mechanical and engineering service costs, expandable materials and some others. There also exist some fixed costs (car, equipment, tools and others) which are incurred either by drivers or by race teams, but usually they are incurred by drivers. In my work I am going to assume that these cost are incurred by drivers only. In case of high results in karting competitions drivers can attract a financing from sponsors (which are usually represented by private companies) to participate in formula racing series such as Formula 4, for example. However, in Russia this happens quite rarely as the sponsorship practice is not well-developed. Therefore, the main motivation for participation in competitions is just the utility from competitions themselves. management russian karting market

The supply side is represented by race teams. The market is quite competitive and consists of approximately 150 small teams. Race teams are usually ruled by former karting drivers, engineers or mechanics which are not adequately qualified in management. These teams usually consist of 3-10 persons and provide mainly mechanical and engineering services to drivers. That is, they prepare cars for competitions and work with different adjustments to make the car fast. Each team usually have just one engineer which acts also as a team manager. The job of engineer is quite specific and cannot be done by other specialists. Mechanics act as engineer's assistants and make not very karting-specific job. Therefore, their labor can be easily substituted by that of car industry mechanics. Each team has some fixed assets such as the car park, special tools, specially equipped lorry and, sometimes, racing karts. In my model the assumption is that the race teams do not hold any inventories of race cars. Race teams receive membership fees as the main source of their income. They can also earn revenues from dealership activities.

There also exist intermediaries in this market. First of them are race organizers. They act as event-agencies and are responsible for all aspects of the competition organization. The main race organizer is the government organization called Russian Automobile Federation (RAF). It indirectly controls all competitions by stating the rules and directly controls the most prestigious competitions of the federal level. Other race organizers are represented by private companies. Among other intermediaries are karting tracks which provide the area for competitions and equipment suppliers which provide personal equipment, cars and other things for the participation in competitions.

The main interaction takes place between drivers and race teams. Drivers search for the team which can offer them the best conditions for the participation in competitions. Drivers are looking for the affordable price level, appropriate experience and suitable equipment. Traditionally, race teams use the word of mouth marketing to inform drivers about their services. However, nowadays more and more teams are switching to the modern methods of marketing such as social media marketing (SMM), targeted advertising and others. When the deal between team and driver is made, the cooperation can be of the form of a long-term contract for a season, a contract for a one separate race or just the sale of some number of racing days without any contract. The most popular form is the long-term contract as it allows drivers to receive a good value for money offer and allows teams to receive a steady flow of income.

Information Asymmetry

The quality of race teams' services is clearly unobservable to clients due to following 2 reasons. First, the information is hard to find. As team managers are majorly using word of mouth marketing, it may take great effort to find information about some particular team. Second, this information can be unreliable as it is based on the subjective opinion of some particular driver. Moreover, it is hard to distinguish between the performance of a driver and the performance of a team. That is, if a driver has a good result and is fully satisfied with his team, it does not necessary mean that the team is high quality. It means that either driver or team or both are high quality. Even a driver himself in this case can hardly make this distinction, because it is usually hard to objectively evaluate yourself and, moreover, he may have no expertise in the area of karting technical adjustments. These 2 reasons lead to the adverse selection problem. This problem here means that a team which advertises most extensively is the one which is going to get more drivers. On the other hand, drivers are not ready to pay very high fees for the unobservable quality and are ready to pay only average fees. Therefore, a team should concentrate on marketing activities rather than on quality improvement to attract new drivers. This is the main implication of this section for my model. However, the quality improvement can be still important for current drivers and for the establishment of long-term relations with them. I am going to study its importance also later in my model.

Similar Cost Structure

The costs of karting competitions are extremely high. Motorsport (including karting) is considered the most expensive sport in the world. Main expenses are related to the racing technique. It includes the purchase of racing cars, regular maintenance costs, spare parts and expandable materials and personnel salaries. These costs are fully incurred by drivers. Speaking about race teams, their main costs are: car park rent payments, lorry maintenance costs and marketing costs. These costs comprise a very small part of total costs and are very similar across different teams. Moreover, driver contracts are usually structured in such a way that these costs are also shifted on the client. Therefore, the cost management affects only the team's ability to make a good value for money offer to drivers. However, as these costs comprise just a small part of total costs for clients, the team's cost management is not capable of changing the price greatly. Its implication for my model is the fact that I am going to abstract from the cost management in this work, but it can be still important and become the basis for the further researches.

High Mobility of Labour

The major part of race teams' employees consists of mechanics. This specialization does not require many special skills. All necessary skills can be taught just in 1-2 days. Moreover, mechanics working in normal repair services can easily substitute mechanics working in karting. It means that there exists a large pool of potential candidates for the position of mechanic in a race team. Moreover, these candidates should be more interested in the work in race team, as salaries in karting are on average greater than salaries in repair services. The structure of the mechanic's contract is determined by the structure of the driver's contract. If driver has signed a long-term contract for a full season, mechanic would have also such contract as each mechanic is assigned to a single driver. Therefore, in my model I will assume that it is possible for a team to hire a new mechanic anytime he is needed, and it is possible to fire him whenever he is not needed anymore. The last part about the firing is also quite realistic as in this industry mechanics usually work without any official contracts.

Excess Supply of Racing Equipment

This section analyses the position of intermediaries which supply racing technique and equipment (further referred as equipment) to drivers. Almost all equipment is delivered from the Italy, as Italy has the strongest karting industry and the best equipment. Italian producers always have conditions limiting the quantity of equipment supplied. In countries where the karting is well-developed, it is not a problem for suppliers to buy a required quantity of equipment from Italian producers. However, for small Russian market it is a problem. Russian suppliers are required to buy equipment which sometimes they are not capable to sale. For sure, they negotiate cooperation condition with Italian producers to decrease the required limit, but it does not solve the problem completely. Moreover, each year Italian producers introduce new types of equipment which are usually better than old ones. Russian suppliers are forced to buy these new arrivals as otherwise they will have no buyers. That is, their inventories are usually overloaded and they are ready to sell necessary equipment to a race team anytime. Therefore, in my model I assume that a driver can immediately buy all required equipment as soon as he signs a contract with a team.

Lack of Drivers

Russian karting market can be said to be in a downturn due to the bad economic situation. As I have written above, almost all equipment is bought from the Italy and is denominated in euro. Therefore, after a serious depreciation of ruble happened a year ago drivers' expenses almost doubled. It has significantly decreased the number of drivers willing to take part in karting competitions. This fact had a great influence on the market situation. Previously consumer-driven market became even more consumer-driven. The important implication for my model is related to the relationships with race organizers. To take part in competitions all drivers are required to receive special licenses, and there exist some requirements to receive these licenses which drivers should met. Nowadays these requirements are seriously eased as organizers have not enough drivers to run their championships. It makes participation in competitions easier and faster for drivers. Therefore, in my model I will assume that there are no barriers at all, and it is possible for drivers to start racing as soon as they sign a contract with a team. This assumption is not critical for the analysis and is just made for simplicity.

Efficient Sales & Marketing processes are required

Summing up what is written above, we can develop the requirements for the efficient race team management method. First, it should concentrate on marketing activities to overcome the problem of information shortage. Second, it should not concentrate on the cost management as its potential result will be not very significant. Third, drivers are able to start racing as soon as they sign a contract, and the method should not concentrate on the issues related to the equipment, labour or licenses. Therefore, an efficient method for the race team management should improve the quality of team's sales and marketing processes to attract new drivers to the team. For this reason, in the next part of my work I will concentrate only on these processes and will study the factors affecting them. That is, my management method will include only the usage of the most efficient marketing instruments to increase sales.Chapter 2: Choice of the model to evaluate efficiency

In this chapter I will develop the model to evaluate efficiency of the management method. I will start from the description of numerous variables which probably affect race teams' sales and will explain the reasons for this potential influence. Some of these variables are related to marketing instruments, while others are related to product quality characteristics. Next, I will turn to the model which will help to estimate the influence of these variables. At first, I will use the OLS model which is not very suitable for my case, but is the easiest one to estimate, and it will be also useful to compare its results with the results of SEM. Of course, I will describe the problems of OLS in details to justify the usage of SEM. Next, I will describe the Structural Equation Modelling method using scientific literature. As this method is quite new and not very famous, I will explain all its mechanics and steps in detail. Finally, I will turn to the next chapter, where I will present results of empirical tests.

Variables Affecting Sales

Here I will introduce all factors which are presented in my model. The dependent variable which I am going to predict using SEM model is the number of sales. The independent variables are: Price, WebSMM, Content, Networking, Targeted, Identity, Banners, ResultsW, International and Crashes. I will explain the meaning of these variables further. All these variables' values are taken for the one last race season period.

Sales variable is the number of races sold by the race team during the race season. Race season is the period from March to November when all races and tests take place. The number of sales usually ranges from 5 to 50. This variable sums races bought by different drivers within the team. That is, it does not matter whether one driver has bought 5 races or 5 drivers bought 1 race each. Such way of calculating sales allows to include not only drivers with long-term contracts, but also drivers who have just bought a one race during the season. I assume here that there are no limits for the number of sales as the supply of mechanics' labour, supply of equipment and the ability to receive a racing license will be always available due to the current market situation.

Price is the first independent variable which probably affects sales. This variable is measured in the thousands of rubles per one race. For sure, the price for a single race will be lower in a case of a long-term contract. However, we are interested in the relative price value rather than in the absolute one. As we have no grounds to assume that the difference between a one-off race and a single race in the long-term contract will be significantly different across team, we can assume that this Price variable will be a good indicator of the team's price level. We can expect this variable to negatively influence the number of sales, as drivers should prefer to pay lower prices than high prices.

Next variable is WebSMM. This variable is calculated as the sum of number of followers in the most popular Russian social networks (Vkontakte and Facebook) and the number of website visitors per month. If a team does not use social networks and website, it would have received a zero for this variable estimate. If a team uses Vkontakte and website, but do not uses Facebook, this variable would calculate the number of followers in Vkontakte and the website traffic leaving the value of Facebook followers of zero. These particular social networks have been chosen as the most popular ones. Race teams barely use other social networks. From the collected data we can see that just few teams (about 5-10) use Instagram and YouTube networks, while other social networks are not used at all. Therefore, there is no sense to include the number of followers in these networks in the model. However, in the future this situation can change so that teams will start to use these networks intensively, and it will be interesting to include these variable in the model also that time. I also understand that there is some intersection of social network followers with website visitors. However, as there are no grounds to assume that this intersection will be significantly different across race teams, we can use this variable without accounting for interception as we are again interested in the relative rather than absolute values. We can expect WebSMM variable to positively influence the number of sales as the large number of followers or website visitors implies the large company (or brand) awareness. That is, the greater the number of followers and visitors, the greater number of people know about this team. We also assume here that people visiting team's website and social pages are interested in becoming drivers themselves or in making drivers their children. That is, there is some probability that each person visiting team's pages will become the team's driver. This assumption seems quite realistic as karting is not very popular sport, and there are not many fans which visit team's pages just for fan purposes.

Content variable is the dummy variable which takes value of 1 if a race team creates unique content within its social network pages or website, and a value of 0 if a race team does not create this content. This content can be represented by specialized articles about karting, photo and video reviews of races, press releases and other similar things. It is important for this content to be unique (that is, to be created by this team) as a content reposted from some other source cannot attract any significant attention of followers as these followers will associate this content with originating source rather than with the team. However, not unique content can also have some positive influence on sales, and it will be good to study its influence in case of the proven significant influence of the unique content. We can expect the Content variable to have a positive influence on sales as the unique content should attract an attention to the team increasing the number of followers and visitors and growing the interest to the team. Of course, this content should be of a good quality to make this work. That is, images should be in a high definition, texts should be written without grammatical mistakes and etc. That is why, during the data collection a bad quality unique content will be treated as an absence of unique content.

Networking is also a dummy variable which takes value of 1 if a race team actively communicate online with its followers and value of 0 otherwise. This communication can be represented by the public discussions in the social media pages, promoting followers to communicate their thoughts in comments, answering these comments, making public questionnaires and votings and etc. It is important for this conversations to be regular as not regular conversations are unlikely to have significant effect. We can expect this variable to have a positive influence on sales as increasing communications with potential clients most probably leads to their increased involvement, which in turn leads to the increased probability of signing a driver's contract.

Next variable is Targeted. It is also a dummy variable which takes value of 1 if a team uses targeted advertising via Google AdWords or Yandex Direct and value of 0 otherwise. Of course, it will be better to take into account not only the fact of targeted advertising usage, but also the quality and volume of this advertising. The problem is that it is impossible for an outsider to collect this information. Therefore, the dummy variable will be used. However, even just a fact of targeted advertising presence or absence can significantly influence sales. I am going to prove this point in my model, where I expect this variable to positively affect sales. At first, as each type of advertising does, targeted advertising increase the brand awareness. Second, it increases the website traffic or the number of followers. Moreover, this traffic can be expected to be more valuable (that is, with greater probability of becoming a driver) because of the nature of this type of advertising as only interested people will click the link and turn to the team's website.

Identity is also a dummy variable which takes value of 1 if a team has its own corporate identity and value of 0 otherwise. In my notation, corporate identity means a combination of logotype, corporate colors and a sustainable team name which had not been changed for a reasonable period of time. In other words, we can think of this variable as of the existence of the team brand. This thing can seem quite obvious for each time to have, but in reality not many teams try to create any type of brand. Many of them just take a name like “The Team of Ivan Ivanov” and do not care about any logotypes, colors and etc. Therefore, we can expect teams which have corporate identity to have more drivers and more sales than teams which do not have it. That is, an expected influence of Identity on sales is positive.

Banners is a dummy variable related to the presence of the offline advertising during karting competitions. It takes value of 1 if a team uses this type of advertising and value of 0 otherwise. This type of advertising includes the usage of different tent banners, branded flags, customized team apparel and etc. Usage of any type of these instruments would be regarded as the usage of offline advertising in my model. In future work it will be possible to differentiate among different types of offline marketing instruments, but for now it will be enough to test the significance of this type of marketing in general. We can expect this variable to have a positive influence on sales for the same reasons as for other types of advertising.

ResultsW is the variable which takes into account the results of competitions in which the team has taken part. All competitions can be divided into 2 categories: federal and regional. Of course, federal competitions are more important and difficult to win. Therefore, this variable should give a greater weight to results in federal competitions. Next thing to consider is what to count as a good result. In each race usually take part from 10 to 50 driver, and 45^th position is not very different from 49^th position, for example. That is, not every result is assumed to influence the team's performance. However, the top 3 positions (the podium) in every race is always good for drivers. Only these places are always awarded with race cups and are spotlighted in the press. It is reasonable to assume that exactly these results influence the performance of teams in the eyes of public. Next thing is to construct the appropriate index to account for the number of good results (podiums), total number of races and the type of these races. The easiest one will be the index showing the ratio of podiums to the total number of races. It will also have a good interpretation. If a team has this index of 50%, it means that a team have finished 50% of races on the podium. The problem is that this index does not account for the type of race (federal or regional), and this distinction is significant. This problem can be solved by giving different weights to the podiums in different races. The exact weight for each type of race is arguable and can be subject to its own research, but for now I will use the methodology of the Russian Automobile Federation (RAF). They have special requirements allowing to receive titles of the Candidate Master of Sports, Master of Sports and etc. These requirements are adjusted for the type of race. In general, the coefficient for federal races is 2. That is, federal race is 2 times more significant than the regional race. I am going to use this coefficient also in my model. Therefore, the final index (variable ResultsW) is constructed as the ratio of weight-adjusted number of podiums to the total number of races). This variable is expected to have a positive influence on sales as clients should have treat the team's results as the signal of the team's quality.

One more variable in my model is International, which is a dummy variable. It takes value of 1 if a team takes part in international competitions and value of 0 otherwise. As the Russian karting market is not well-developed, it is very important for Russian drivers and teams to gain an international experience. This experience tends to significantly increase their skills as they learn from the best experts. The learning should not be necessary be explicit. Drivers can learn driving skills just chasing their competitors in the race, engineers can learn some technical secrets just watching their international colleagues in the paddock and so on. It is also an empirical evidence that all successive Russian drivers are always taking part not only in Russian, but also in international competitions. Therefore, this variable is expected to have positive influence on sales as the team taking part in international competitions will be treated as high quality team by the public. We assume here that it will be not profitable for low quality teams to take part in international competitions to pretend high quality teams. In this case the signal will be efficient. It is a reasonable assumption, as going to international competitions requires very high costs which usually cannot be afforded by low quality teams. Moreover, license costs will be also much greater for low quality teams than for high quality teams as the RAF does not like low quality teams to go to international competitions. It is important to notice that a dummy variable is used here instead of some performance index as in previous section for a one reason. Russian drivers have never achieved any good results in international competitions, unfortunately. That is, there are no significant international results to compare across the teams.

My final variable is Crashes. It determines the frequency of crashes happening during the races. This variable is expected to significantly negatively influence the number of sales, because the number of crashes is the thing that greatly depends on the work of an engineer and mechanic (that is, on the work of a race team). Most crashes usually happen due to some technical faults that usually can be prevented by the team. Therefore, this variable can be seen as a good indicator of the team's technical work. The problem here is that some technical crashes could be cause be the driver himself, rather by his team. However, these situations are very rare and can be ignored in the model. This variable is calculated as the ratio of the total number of crashes to the total number of races. It also has a useful interpretation. If its value is equal to 50%, it means that in 50% of races the driver had technically crashed.

Problems of the OLS Model

First method which comes to the mind is the simple linear regression model. It is the easiest one to construct and understand. I will not describe in details its methodology as everybody should probably be familiar with it. Moreover, this method is not very important for my work and serves just for the comparison of results with the SEM method. I will move straight to the model specification for my study.

It will be a model with 1 dependent variable (Sales) and 10 independent variables (Price, WebSMM, Content, Networking, Targeted, Identity, Banners, ResultsW, International, Crashes).

,

Where i denotes some particular race team, and betas are linear regression coefficients. This model can predict the number of the team's sales based on the information about these 10 observable variables. However, this model is significantly flawed due to the following reasons:

1. Explanatory variables have complex interrelations among each other. That is, they are correlated, and it leads to the problem of multicollinearity. I am going to prove this point mathematically in the next chapter, but here I will just provide a rationale for these interrelations. Consider the variable WebSMM. To remind you, it shows the number of the team's social network followers and website monthly visitors. We can reasonably assume that this number depends on the presence of unique content in these resources, on the usage of targeted advertising, on the existence of corporate identity and on the usage of active online communications with potential clients. That is, the variable WebSMM can be determined by variables Content, Targeted, Identity and Network. One more example is the variable ResultsW. It is also reasonable to assume that the team's results are influenced by the number of technical crashes (variable Crashes) and the participation in international competitions (variable International). Therefore, these variables should be also correlated. This problem does not make the OLS model invalid, but it probably makes it inefficient (Dougherty, 2001).

2. Next, this model includes many explanatory variables. At first, it makes the model more difficult to interpret. Second, even the model still stays valid, it can become inefficient if some of these variables had been irrelevantly included here (Dougherty, 2001).

3. Finally, there may exist some unobservable factors (latent constructs) influencing the number of sales. However, these factors cannot be included into standard OLS analysis.

Of course, OLS model can be improved to get rid of these problems (except the inability to include latent constructs), but I will use SEM method as a more convenient one and allowing to include latent constructs.

Description of the Structural Equation Modelling

In this section I will make the literature overview to provide a detailed description of the Structural Equation Modelling (SEM) method. First, I will provide its definition and rationale for its usage. Next, I will explain each step of this process, but in this section I am not going to explain practical estimation issues. This topic will be covered in the next chapter related to the empirical testing. The section will be finished with the analysis of advantages and disadvantages of SEM method.

“Structural equation modeling (SEM) refers to a diverse set of mathematical models, computer algorithms, and statistical methods that fit networks of constructs to data” (Kaplan, 2007, p. 79-88). These model include multiple steps such as explanatory factor analysis, confirmatory factor analysis, path analysis, partial least squares path analysis, LISREL and latent growth modeling (Kline, 2011). However, they do not necessary use all these steps. SEM method is a flexible one and can be adjusted to better estimate the model. The most widely used option of SEM method is the combination of factor analysis and regression (or path) analysis (Hox and Bechger, 1998). SEM method is commonly used in social and behavioural sciences as it allows to study the influence of some unobservable variables (latent constructs). It also provides a visualization for the interrelations among variables in the form of a path diagram which is easy to understand and interpret.

SEM analysis is always started by drawing a path diagram. This diagram was invented by Sewall Wright (Wright, 1921) and uses the following notation. Observable variables are denoted by rectangular box, while unobservable variables are denoted by ellipses. Single headed arrows are used to denote causal relationships in the regression. The variable in the pointing end of arrow is caused by the variable in the other end. Double headed arrows are used to denote covariance or correlation between variables.

After drawing this diagram, it is customary to turn to the factor analysis. There are 2 possible options to start: explanatory factor analysis (EFA) or confirmatory factor analysis (CFA). EFA is used if we have no prior hypothesis about the influence of some latent constructs on our data, while CFA is used if we have such hypothesis and want to test its validity. That is, in EFA the model is arbitrary and it load all variables on all factors. Here, some transformation method should be used to improve the interpretation of results. The most widely used one is the Varimax rotation. First step of EFA is to find the appropriate number of factors to use. There are many way to perform this step such as Kaiser's (1960) eigenvalue-greater-than-one rule (also known as K1 or Kaiser criterion), Cattell's (1966) scree plot, Revelle and Rocklin (1979) very simple structure, Optimal Coordinate and Acceleration Factor, Velicer's (1976) Minimum Average Partial test (MAP), Parallel analysis, Ruscio and Roche's (2012) Comparison Data and etc. I will not explain each of these methods, but rather explain the one which I am going to use further in my model - Cattell's (1966) scree plot. The procedure of this method is as follows: “compute the eigenvalues for the correlation matrix and plot the values from largest to smallest; examine the graph to determine the last substantial drop in the magnitude of eigenvalues; the number of plotted points before the last drop is the number of factors to include in the model” (Cattell, 1966, p. 245-276). This method seems to be quite subjective, as it is not always clear what can be counted as the substantial drop (Courtney, 2013). However, in my case this drop is clearly seen, and this method can be properly used. Next step is to estimate the factor loading and variance for the appropriate number of factors determined earlier. It can be done using the Maximum Likelihood (ML) or Principal axis factoring (PAF) approach depending on the normality of data. ML should be used for normally-distributed data, while PAF should be used otherwise (Leandre et al., 1999). Finally, we will receive factor loadings on all variable for some number of factors. These loading can be interpreted as the strength of factor influences on each variable. At this moment researches can take variables with greatest loading for each factor to find the relation between them and to name this factor. It can be done to make the further work with factors more convenient and interpretable. The problem of this approach is that it still has a subjective element in it, as it requires researchers to search for some relation between variables.

CFA may be used as the alternative first step or as the second step in SEM. The main goal of this method is to test whether a hypothesized model fits data well. That is, the prior hypothesis (which can be represented by the path diagram) is required. In SEM this analysis has 2 important purposes. “First, it aims to obtain estimates of parameters (which are factor loadings), variances and covariances of the factor, residual error variances of the observed variables. The second purpose is to assess the fit of the model, i.e. to assess whether the model itself provides a good fit to the data” (Hox and Bechger, 1998, p. 3). In CFA we should fix one factor loading for each factor to 1 to give this factor an interpretable scale. The reason for it is the fact that “factor loadings are a function of the variance of the latent factor, and the variance of the latent factor is a function of the loadings” (Hox and Bechger, 1998, p. 3). That is, we cannot estimate simultaneously unique values for them. There are many ways to evaluate the model fit in CFA. They can be grouped into the absolute fit indices and relative fit indices. “Absolute fit indices determine how well the a priori model fits, or reproduces the data”, while “relative fit indices compare the chi-square for the hypothesized model to one from a “null”, or “baseline” model (McDonald and Ho, 2002, p. 64-82). Absolute fit indices include chi-squared test, root mean square error of approximation, root mean square residual and standardized root mean square residual, goodness of fit index and adjusted goodness of fit index and some others. Relative fit indices include normed fit index (NFI), non-normed fit index (NNFI also known as Tucker-Lewis index or TLI) and comparative fit index (CFI). I will describe only those of them which I am going to use. First, I will use a chi-squared test as the most convenient one. This test indicates the difference between observed and expected covariance matrices. The null hypothesis is that the model is a good fit. That is, greater the p-value, greater the significance of the model. Second, I will use relative indices NNFI and CFI. I will not use NFI as it tends to be negatively biased in most cases, while NNFI resolves some of the issues of negative bias (Bentler, 1990). This index “analyzes the discrepancy between the chi-squared value of the hypothesized model and the chi-squared value of the null model” (Bentler and Bonett, 1980, p. 588-606) and ranges from 0 to 1. The value above 0.95 indicates that the model is a good fit (Hu and Bentler, 1999). CFI “analyzes the model fit by examining the discrepancy between the data and the hypothesized model, while adjusting for the issues of sample size inherent in the chi-squared test of model fit, and the NFI” (Bentler, 19990, p. 238-46). It also ranges from 0 to 1, and the value around and above 0.9 indicates a good model fit (Hu and Bentler, 1999).

In the next final step, the measurement model (or path analysis) is constructed. This model is a multivariate regression which assumes data to have a multivariate normal distribution. That is, this model can be represented by a system of equations. Once these equations are written using the constructed earlier latent factors, “SEM software uses complex algorithms that maximize the fit of the model, taking all model restrictions (fixed parameters, equality constraints) into account” (Hox and Bechger, 1998, p. 8). ML method is usually used for the estimation. The results are then assessed by some statistical tests to check their significance. Chi-square test, NNFI and CFI indices again can be used to assess the overall quality of the model. The procedure will be the same as in the previous section, so I will not discuss it again here. T-tests can be used to assess the significance of separate variables within the model. Finally, the model can be modified using special modification indices such as Lagrange Multipliers and Wald indices. “Lagrange multipliers are printed in place of fixed parameters; they indicate how much better the model would fit if the related parameter was freely estimated. Wald indices are printed in the place of free parameters; these statistics tell how much worse the model would fit if the parameter was fixed at zero” (Wothke, 2010, p. 29). However, these indices are not very convenient to use in practice and sometimes are not necessary. These indices were not used in my work, so I am not going to discuss them in greater details.

2. Empirical testing

I will start this chapter by introducing my data sources and the methods which I have used to collect data. Next, I will come to the empirical testing of my model making the steps of explanatory factor analysis, confirmatory factor analysis and path analysis. I will also shortly describe a software used in the estimation process (STATA and R)

Data collection

In this section I will present data sources and collection methods separately for each variable used in the model.

I will start from the dependent variable (Sales) which indicates the number of races sold by the team during a race season. This information was collected from the official competition records. These records are kept by race organizers on their official websites. All records contain an information about participating drivers, teams in which they were driving and final results. That is, the number of races during the season for each team can be easily calculated by summing up all races in which its drivers had participated.

Next variable is the Price. It was quite difficult to receive this private information, as teams are not willing to share it. Moreover, they have no official price lists and can significantly differentiate their prices for different clients. For example, if they know that the driver is skilled, they can probably reduce price, and vice versa. These pricing strategies can be significantly different across different teams. Here it is also important to notice that Russian karting is not very big and the majority of people are familiar with each other. Particularly, they are familiar with me as a driver, and it can cause bias in the data. Therefore, to increase the objectivity of this data it is better to conduct this research by a third party not related to a karting at all. That is why, this research was ordered to the third party call-center. The instructions were to use the mystery shopping method to collect this data. That is, they were calling to team managers presenting themselves as potential clients. The story was the same each time to minimize bias in the data. Therefore, I have grounds to assume that this data was collected with maximum possible objectivity.

Data for WebSMM variable has been collected using teams' social network and website pages. The number of followers is always clearly seen in the Vkontakte or Facebook, while the website traffic can be seen using special counters such as KartingProfi (very popular one in this market), LiveInternet and Yandex Metrics.

Data for Content and Networking variables has been also collected using same online sources. The fact of presence of unique content or active online public communication was evaluated manually viewing each team's page. The period viewed was the last full race season.

The fact of usage of targeted advertising have been also evaluated manually using Google and Yandex websites. Almost all reasonable combinations of key words have been sorted in the different time of the day and different day of the week to capture all existing targeted advertising links. However, some advertising can be still missing using such way, but these advertising are unlikely to be efficient and, therefore, can be treated as the absence of targeted advertising.

The variable Identity have a positive value if 3 conditions are met: logotype, corporate colors and sustainable name. It is possible to evaluate this variable manually by also looking at teams' online pages and tents placed near the track during competitions. The presence of logotype and corporate colors can be determined in this way, but to determine the presence of a sustainable name it is necessary to take some long period of type. In my work the period of 2 race seasons has been taken. The data was collected from the official competition records using teams' license numbers which are also available.

Next variable is Banners. It can be evaluated in 2 ways: online (by looking at the teams' photos and videos from competitions and tests) and offline (by looking at the teams' tents and apparel during competitions). Both methods have been employed here to collect all necessary information.

Information about teams' results for the ResultsW variable has been also gathered from the official competition records which contain information about all participating drivers' final results.

The participation of the team in international competitions (variable International) has been evaluated using competition records of 2 organizers of international competitions: CIK FIA and World Series of Karting (WSK).

Finally, the data for Crashes variable has been also gathered from the official records. These records mark each driver's technical crash by the letters NC or DQ. Therefore, all technical crashes of the team's drivers can be easily counted and summed up.