Consumer Purchase Behaviour and Nash Equilibrium
4.3. Consumer Purchase Behaviour
4.3.1. Survey data cleansing
4.3.3.3. Resultant Functions
Following part-worth conjoint analysis, the consumer utility, of the ith segment for the jth product and the mth player, , is assumed to be a linear function. Three market segments, , were defined using cluster analysis and hypothesis testing. One of the part-worth binary independent variable
U
mis brand, where a player, m, where is part of a pre-set brand type as defined by binary attributes, , , and empty.
For a product, j, there is a set defined group of attributes that determine that utility of the product. These attributes are represented in a binary fashion with the variables as defined in section 4.3.2, and a full set of the defined products, j, can be found in Appendix H. The attributes for UK Motor Insurance are the Insurance Level, Brand, Price, and Add-ons. Each attribute has various attribute levels that contribute as the independent variables in the customer utility function.
From the conjoint analysis of each market segment, using JMP conjoint analysis, it was found that not all segments have the same utility function. Each has a different set of product attributes that contribute in a statistically significant way to the utility of the product. The base-line used for regression and conjoint analysis were: Third Party Only, Small/Medium Size Brand, High Price Point and No Add- ons.
The brand is defined by the player brand group (Big Insurance, High Street, Small/Medium Size and Other). Therefore, The brand attribute is a variable in the market, but not within a player strategy. The product, and its attributes, presented by the player, determines the final utility.
For the three segments the following binary independent variables, or attribute levels, are defined as follows:
High-Income Earners:
!
Students:
Remaining Population:
The Consumer Utility Functions are the regressive linear functions derived from the conjoint analysis. There is one linear function for each market segment discovered during the cluster analysis. Not all consumer utility functions contain the same independent variables. This demonstrates that clusters differ in their purchase behaviour and product attributes that impact the purchase choice of one
bm 1j bm 2j bm3j Um 1j = 0.52rm1j+ 0.29r2mj 0.37rm3j+ 0.34gmj Um 2j = 1.04lm1j 0.25lm2j+ 0.13bm1j 0.57bm2j + 0.76bm3j 0.24rm1j + 1.17rm2j+ 0.28rm 3j+ 0.79cmj Um 3j = 1.71l1 0.63l2+ 1.48r1+ 1.14r2 0.85r3+ 0.19g+ 0.62c+ 0.46k
market segment, may have no impact on another. The size market segments was found to be fairly represented in the respondent data. In the data there are 7 high income respondents, 9 students and 99 from the remaining population. Significance analysis of the coefficients for each of these market segments, as well as all respondent data, can be found in Appendix J (Images J.1-J.4). It can be said to be remarkable that such small numbers of respondents could have several statistically significant
coefficients, but with the use of DOE for the conjoint choice analysis and common economic markers of these market segments it is expected that the differentiation between data points would be small, allowing for a raised significance with a small sample (Burns & Burns, 2008).
High Income Earners are unique in that they do not appear to have a brand or insurance level
preference and have only a notable desire for the legal cover add-on. The price preference profile for the High Income Earners is unlike the other groups in that their is no dislike for the lowest price point. It can actually be seen that the lowest price point is the highest preference with the mid-high price point being the point of disinterest. This is most likely due to the customer perception that a higher price point would provide greater value beyond what may be obvious. This behaviour is frequently seen in many food markets and a point of debate in new product management (Krishnan, et al., 1999).
The student market segment can be seen to have a reliable brand preference. Specifically there is a preference for the Other brands category (Car manufacturers, Banks etc.) and yet are noticably against purchasing from the High Street Brands (M&S, WHSmith etc). The students have a dislike, or distrust, for the lowest price point, and demonstrate the strongest preference for the second to lowest price point. The students have a statistically significant preference for the Protected No-Claims Bonus Cover add-on. This add-on protests the insurers No-Claims Bonus from incidents where the other driver is uninsured or it is a “hit-and-run” or a similar type incident. This is an understandable preference for individuals in this student category given their usual age and driving experience, where there would be a dramatic increase in insurance premiums if they needed to claim on insurance.
The behaviour model of the remaining population shows a very strong preference for the lower price point and then a sudden dislike for the next to highest price point. This market segment demonstrates a similar preference to insurance levels as the Student population as well as the No-Claims Bonus
Protections add-on. This group, however, contains statistically significant preferences for a greater list of add-ons than the other two groups. The add-ons include the Protected No-Claims Bonus, as seen with the Students, the Legal Cover as seen with the High Income Earners, but also includes a preference for Key Cover. The Key Cover option is the inclusion of insurance for the loss or theft of the car’s electronic key fab. This is a surprisingly pricy item to replace. The larger number of add-ons included
as statistically significant could be due to size of sample, for the coefficients are not very high and the larger sample sizes create more accurate results (Jolson & Hise, 1973).