Russian karting market

In total, 111 observations have been collected for all 11 variables. The sample of such size can be treated as acceptable for SEM method.

Testing of the Model for a Race Team

In this section I will present the structural equation model specially developed for a race team case. I will start from the presenting of my hypothesized model. Next, I will conduct the explanatory factor analysis where I will find the appropriate number of latent factors to include in the model, find factor loading for all variables in the model and will name the resulted factors based on the relationships among variables with greatest loadings on that factor. Next, I will turn to the confirmatory factor analysis where I will test my hypothetical model for fit. I will find the best model with greatest chi-square p-value and NNFI / CFI indices. I will also compare these results to the results of EFA. As a result, I will receive a conclusion about the presence of latent factors influencing studied variables. Next, I will move to the measurement model which is a multivariate regression. I will evaluate it using an ML approach. Finally, I will interpret the received results.

Description of the Hypothetical Model

In my hypothesis there are 2 unobservable factors influencing the Sales. These constructs are: Quality and Marketing. Quality is the factor related to the level of team's services quality, while Marketing is the factor related to the level of team's marketing intensity. For sure, both these factors cannot be measured. That is, to determine these factors we should use some observable variables which we have.

I assume that the Quality factor influences Price, Results, Crashes, International and WebSMM variables. The relationship of Quality with Results and Crashes is quite obvious - high quality teams should be the ones who achieve better results and have less technical crashes. Quality should also influence International, as it less costly for high quality teams to take part in international competitions as described earlier. The positive relationship of Quality with Price seems also reasonable as high quality teams can afford to charge greater prices. Finally, Quality is likely to affect the number of online followers and visitors (WebSMM) as high quality teams should be the most famous ones. Therefore, the public should have more interest in them. I assume also, that ResultsW, International and Crashes are correlated with each other.

Speaking about the Marketing factor, I assume that it influences WebSMM, Identity, Banners, Content, Networking and Targeted variables. Moreover, variables Content, Networking and Targeted affect the variable WebSMM. The first relationship seems obvious as WebSMM, Identity and Banners represent different marketing instruments. That is, they are related to the Marketing factor. The second relationship is also quite reasonable as the number of online followers and visitors should be affected by the parameters related to the quality of team's social networks and websites such as the presence of unique content, active online public communication and targeted advertising.

After development of this hypothesis we should move further to test its validity and probably to improve this model.

Explanatory Factor Analysis

I have started from the EFA to determine some latent factors which influence variables in the model and then to compare them to hypothesized factors. This part of the analysis has been done using STATA software. In the first step I have determined the appropriate number of factors to include. For this purpose, Cattell's (1966) scree plot. The point of substantial drop was clearly visible on the level of 2 factors. Therefore, 2 factors were included in the model. Next, I have used an ML method together with Varimax rotation to construct an arbitrary model where all variables are loaded on all factors. It gave me the following result:

Next, we should determine variables with greatest loadings on each factor. For the first factor these variables are: Price, ResultsW and Internationsl. For the second factor these variables are: Targeted, WebSMM and Content. We have taken top 3 variables for each factor.

As we can see, these factors seem close to the hypothetical factors Quality and Marketing. The factor 1 here mostly influences Price, ResultsW and International, while the hypothetical Quality factor influences Price, ResultsW, International, Crashes and WebSMM. The main difference here is in variables Crashes and WebSMM. Their loading on factor 1 in my EFA model is quite low: 0.389 and 0.151 respectively. Therefore, it is possible that either the hypothesis is partly invalid or the results of EFA are partly invalid. We are going to determine it later in the CFA.

The factor 2 here mostly influences Targeted, WebSMM and Content, while the hypothetical Marketing factor influences WebSMM, Identity, Banners, Content, Networking and Targeted. The main difference here is in variables Identity, Banners and Netwotking. They have low factor loading of 0.282, 0.124 and 0.548 respectively. However, the factor loading of Networking is very close to the loading of Content. That is, it can be also related to the factor 2. Final implication is the same as with previous factor - either the hypothesis is partly invalid or the results of EFA are partly invalid.

To sum up, EFA has showed a very important thing for my analysis. There exist exactly 2 latent factors which influence some variables, and the structure of these factors is very close to the structure of hypothetical factors. It can be treated as a good result. Next, we are going to evaluate this model fit using CFA.

Confirmatory Factor Analysis

CFA showed slightly different results. CFA was conducted using modelling in R. The “lavaan” package has been downloaded to work particularly with SEM.

We can see here that this model is a good fit as it passes the chi-square test as the null hypothesis of a good model is not rejected (with p-value of 0.129) and it has high both CFI and TLI indices of 0.987 and 0.976 respectively.

As we can see, this model is a bad fit as it does not pass the chi-square test. The null hypothesis of a good model is rejected (with p-value close to 0). CFI and TLI indices are also low with the values below 0.9. Therefore, this model should not be used further.

Before moving further to the path analysis we should try to find some possibility to improve the model. To do it I have sorted all possible reasonable combinations of variables to receive model with the best fit.

This model has the following form: factor Quality influences variables Price and International, while factor Marketing influences factors WebSMM and Identity. As we can see, this model gives the best fit as its p-value (chi-square) is significantly greater than that of EFA-determined. Its fit indices are also good and very similar to those of EFA-determined model. CFI is almost the same, while TLI is slightly less than 0.9, but it can be still considered a good result (Hu and Bentler, 1999). One more important thing to notice here is the covariance between latent factor. It is equal to 0.132.

Measurement Model

Our final step is to construct the measurement model according the previously identified structure of factors. This step can be also done using the “lavaan” package in R. At first, I use the relationship from my hypothetical model using determined earlier latent constructs. Sales variable depends on 2 latent factors: Quality and Marketing (represented by associated with them observed variables: Price, International and WebSMM, Identity). Price depends on the Quality factor only, WebSMM depends on the Content, Networking and Targeted variables, while Results, Crashes and International are correlated with each other.

As we can see, this model is a good fit. It passes the chi-square test (but only on the 10% significance level) and has high CFI and TLI indices with the value of 0.977 and 0.963 respectively. However, we can see that the model can be improved by getting rid of insignificant variable Networking.

As we can see, the p-value (chi-square) has increased as well as both CFI and TLI indices. That is, the model has definitely become better. This is the final model which I am going to interpret further.

Interpretation

In this section I will separately interpret each equation in the final multivariate regression model.

Equation 1. An increase in the number of social network followers and monthly website visitors by 1 person on average increases the number of races sold during the season by 0.349 races holding other variables constant. Participation in international competitions on average increases the number of races sold during the season by 0.214 races holding other variables constant. Usage of a corporate identity on average increases the number of races sold during the season by 0.259 races holding other variables constant. An increase in the price of a single race by 1 000 rubles on average increases the number of races sold during the season by 0.295 races holding other variables constant. All these results except the last one are consistent with the initial hypothesis. The last result seems quite strange, but it can be explained by the fact that people may treat high prices as a signal of a good team. One more possible explanation for it is the multicollinearity of Price with some other variable.

Equation 2. An increase in the performance index by 1 unit on average increases the price level by 0.879 thousand rubles per race holding other variables constant. Participation in international competitions on average increases the price level by 0.111 thousand rubles per race holding other variables constant. All these results are consistent with the initial hypothesis.

Equation 3. Participation in international competitions on average increases the performance index by 0.383 units holding other variables constant. An increase in the crash index by 1 unit on average increases the performance index by 0.019 units holding other variables constant. All these results except the last one are consistent with the initial hypothesis. However, the influence of Crashes on ResultsW is extremely small and can be caused by multicollinearity.

Equation 4. Creation of the unique content on average increases the number of social network followers and monthly website visitors by 0.113 persons holding other variables constant. Usage of the targeted advertising on average increases the number of social network followers and monthly website visitors by 0.729 persons holding other variables constant. All these results are consistent with the initial hypothesis.

3. Results and their application

In this section I will compare the results of SEM with the results of standard OLS. Next, I will summarize the statements of the targeted race team management method and will give a short introduction to the practical usage of this method in the business model of MSKS K.P. race team. Finally, I will name some issues for the possible further work and will make a conclusion.

As we can see, this model does not seem to be a good fit. Both R-squared and adjusted R-squared are low with the values of 0.5683 and 0.5251 respectively. Many variables such as Price, Content, Targeted, ResultsW and Crashes have insignificant coefficients in this model. Moreover, some of them such as Content and ResultsW have unreasonable signs of coefficients. It all can be caused by OLS problems outlined earlier, and it is completely different from the SEM results.

Management Method

On the basis of results received with help of the SEM method and assuming all made assumptions are true we can state some statements which characterize the method for the efficient race team management. The main insight here is that some marketing instruments are efficient. There are some instruments that directly affect sales such as WebSMM and Identity. Moreover, WebSMM depends on the variables Content and Targeted. Therefore, we can conclude that a race team to be efficient should use the social networks and websites to promote its services and also it should have a particular corporate identity (including logotype, corporate colors and sustainable name). However, the greater effort should be placed on WebSMM instruments as they ate tend to be more important in increasing sales. Moreover, to efficiently use WebSMM instrument a team should also use a targeted advertising and create a unique content. Here the targeted advertising is of greater importance. Therefore, factors named in this section are considered to be the constituents of the efficient race team management.

Example of the MSKS K.P.

In this section I would like to say few words about the practical usage of the developed method. As soon as I have developed it, I have started to implement it into the business model of my race team. Of course, such projects always take a lot of time, and my project have not been finished yet. Therefore, I have not collected a good data that will prove the efficiency of the developed method. However, I have a few interesting things to present. My team have been always participating in international competitions, and the price level have not been changed after the method implementation. Therefore, in the equation which predicts sales only the variables of WebSMM and Identity have been changed. The corporate identity was introduced and social network and website pages were created. These changes have resulted in the increase in the number of sales by 5 races (approximately 30% increase) that can be considered as a quite good result. I am going to finally analyze the results of this method implementation after the end of an appropriate period to make a final conclusion about the validity of this method.

Further Work

Except the study of the practical usage of this method, it will be also useful to make the same research but accounting for the time changes. Therefore, to collect and use panel data to construct the regression within different time periods to get rid of any possible effects related to the specific time period used.

Conclusion

In my work I have analyzed all factors probably influencing the race teams' sales of races. I have developed a numerical model to predict this number of sales using the Structural Equation Modelling (SEM) approach as it turned out to be the most efficient one. The analysis of this model has shown the importance of marketing instruments for the increasing of sales. It has also shown that the most important instrument are social networks, website pages, targeted advertising, unique content and corporate identity. Therefore, race teams which want to increase their sale should use these instruments.

Размещено на Allbest.ru