Statistics for Marketing and Consumer Research

Mario Mazzocchi

• Verlag: SAGE Publications Ltd
• Jahr: 2008
• ISBN:
  9781473903524
• Auflage: 1
• Thema: Business
• Thema: Marketing Research

Balancing simplicity with technical rigour, this practical guide to the statistical techniques essential to research in marketing and related fields, describes each method as well as showing how they are applied.

The book is accompanied by two real data sets to replicate examples and with exercises to solve, as well as detailed guidance on the use of appropriate software including:

- 750 powerpoint slides with lecture notes and step-by-step guides to run analyses in SPSS (also includes screenshots)

- 136 multiple choice questions for tests

This is augmented by in-depth discussion of topics including:

- Sampling

- Data management and statistical packages

- Hypothesis testing

- Cluster analysis

- Structural equation modelling

PART I

Collecting, Preparing and Checking the Data

This first part of the book synthesizes the founding steps of the statistical analysis, those leading to the construction of a data-set meeting the necessary quality requirements.

Chapter 1 reviews the basic measurement issues and measurement scales, which are the elemental tools for measuring constructs whose quantification is not straightforward. It also introduces the distinction between metric (scale) and non-metric (nominal or ordinal) variables and the concept of error, central to statistical theory. Finally, it presents two sample data-sets which are used for examples throughout the book, together with a short review of the most popular statistical packages for data analysis. The distinction between secondary and primary data is explored in chapter 2 and chapter 3 respectively. Chapter 2 deals with the use of secondary data, that is, information not explicitly collected for the purpose of the research, and explores the main available sources for secondary data relevant to consumer and marketing research, bringing examples of official consumer surveys in the UK, Europe and the US. Chapter 3 gets into the process of primary data collection and provides a synthesis of the main steps of the data collection process, with an overview of planning and field work issues. This chapter emphasizes the role of non-random errors in collecting data as compared to random errors. Chapter 4 moves a step forward and begins with the applied work. First, an overview of data quality issues, diagnostics and some solutions is provided, with a special emphasis on missing data and outlier problems. Then, with the aid of the SPSS examples based on the two sample data-sets, it illustrates some graphical and tabular techniques for an initial description of the data-set.

CHAPTER 1

Measurement, Errors and Data for Consumer Research

THIS CHAPTER reviews the key measurement and data collection issues to lay the foundation for the statistical analysis of consumer and marketing data. Measurement is inevitably affected by errors and difficulties, especially when the objects of measurement are persons and their psychological traits, relevant to consumer behavior. The chapter looks at the distinction between systematic and random sources of error, then presents the main measurement scales and discusses the concept of latent dimension for objects – like quality – that cannot be defined or measured in an unequivocal way. Finally, this chapter introduces the two data-sets exploited for applied examples throughout the book and briefly reviews the main statistical packages commonly used in consumer and marketing research.

Section 1.1	introduces the problem of measurement in relation to consumer research
Section 1.2	introduces the fundamental measurement scales and data types
Section 1.3	presents the two data-set used in applied examples throughout the book
Section 1.4	provides a brief overview of commercial software for data analysis

---

THREE LEARNING OUTCOMES

This chapter enables the reader to:

• Focus on the main problems in measuring objects and collecting data
• Review the key data types and measurement scales
• Be introduced to the data-sets and statistical packages used later in the book

---

PRELIMINARY KNOWLEDGE: It is advisable to review the basic probabilistic concepts described in the appendix, with particular attention to frequency and probability distributions and the normal curve.

1.1	Measuring the world (of consumers): the problem of measurement

Measuring the World is the title of an enjoyable novel built upon the lives of the most famous of statisticians, Carl Friedrich Gauss, and Alexander von Humboldt, the naturalist and explorer who founded biogeography.¹ The two scientists have something in common besides their German roots: they both devoted their lives to the problem of measuring objects. Interestingly, the novel portraits Gauss as a quite unsociable mathematician who preferred working at home with numbers, while von Humboldt was traveling around Latin America and Europe with his instruments to measure scientifically lands and animal species. Consumer research needs good measurement and a bit of both personalities, that is, some ‘unsociable’ math and hard field work. These ingredients can create good marketing intelligence. The good news is that both scientific celebrities were successful with money … in the sense that their faces ended up on German banknotes.

This chapter lays the foundations for reading the rest of the textbook. While the math is kept to a minimum, those readers willing to refresh or familiarize with the key mathematical notation and the basic concepts of matrix algebra, probability and statistics are invited to read the appendix, which should help them get a better grasp of the more complex concepts discussed in the rest of the textbook.

However, before getting into quantification and data analysis, it is probably useful to go back to the origin of data and think about the object of measurement and the final purpose of consumer and marketing research. There are several definitions of marketing. A widely used one is that adopted by the Chartered Institute of Marketing (CIM), which states that ‘Marketing is the management process which identifies, anticipates, and supplies customer requirements efficiently and profitably.’

Little more is required to show how consumer research is the foundation of any marketing activity. While in the CIM definition the target is the customer, the focus here is rather on consumers in general, as one might argue that customers are only a subset of consumers. Politically correct language is now even suggesting that the term citizen should be preferred, although few publishers would consider a title on citizen research.

As a matter of fact, consumer research should ideally be able to explore even those behaviors and attitudes which consumers are not even consciously considering at the time of the study. The basis of product innovation is to create something which will be welcome by those consumers who are not yet customers.

Let us take one of the controversial topics in current consumer research – genetically-modified foods. This is one of the biggest challenges for those who want to measure and predict behavior relying on a scientific approach. What is the value of asking someone an opinion, an attitude or a statement on purchasing intentions for something which they don’t know or – in many circumstances – does not even exist? Are there procedures to ensure that the collected data is not completely useless? What is the point of running complex statistical analysis on huge samples if the single data in itself has little meaning?

A well-known example of consumer research failure is the New Coke case in 1985 (see Gelb and Gelb, 1986), when Coca Cola proposed a new formula based on blind sensory tests on 200,000 consumers, testing preferences versus Old Coke and Pepsi. Since tests were blind, research ignored the role of brand image and failed to predict consumer reaction, which was a wide rejection of the new product in favor of the old one, so that Coca Cola had to go back to a single product after 77 days, with a cost $35 billion. While some conspiracy theorists argue that the New Coke case was a successful marketing move to strengthen brand loyalty, this is an unlikely justification for the waste of resources in running the consumer research.

This sort of failure is one of the preferred arguments for those researchers who advocate qualitative research methods over quantitative and statistically-based consumer research.

While this whole textbook argues that only probability and rigorous statistics allow generalization of results from a sample to the target population, it is difficult to deny that ‘ten thousand times nothing is still nothing’ and big representative samples are useless if not misleading when the collected information is not appropriate.

1.1.1 Issues in measurement

Get a ruler and measure the width of the cover of this book. Repeat the exercise 20 times and write down the measurements. Then get a different instrument, a folding rule, and measure the book width again for 20 times. It would be surprising if you had exactly the same measurement 40 times, especially when your instruments are slightly different. More precise instruments and higher target precisions of the measure make it less likely that all observations return the same measure. There may be different reasons for this, for example a slightly different inclination of the ruler, an error in reading the measure or a minor difference between the ruler and the folding rule. Measuring objects is not that straightforward.

Things get much more complicated when the objects of measurement are persons or their tastes, attitudes, beliefs … Needless to say, it might be a waste of time to run complex statistical methods on badly measured information.

The key point here is that empirical measurements and the characteristics being measured are different things, although they are related. Measurement is usually defined as the assignment of numerals to objects or events according to rules (Stevens, 1946). Thus, a transparent solution to the problem of measuring requires a clear and explicit definition of the measurement rule including its mathematical and statistical properties. The rule is usually defined as the scale of measurement.

Before getting into the discussion of measurement scales, it is worth looking at other aspects of measurement theory, those more directly related with the statistical foundation of this book. For example, before using any statistical methodologies one should carefully check whether it is appropriate (Stevens uses the word ‘permissible’) for a given measurement rule. Another issue which follows from the initial example of this section regards the relationship between measurement and reality, with an eye at the concepts of precision and accuracy. While at a first glance it might look like a purely philosophical issue, the question is whether ‘numbers are real’ is quite important to obtain meaningful results from data analysis. For example, take a measurement of attitude toward math taken in three different years on a class of first-year undergraduates. If the measured value increases over time, does it mean that first-year students increasingly like math?

This example draws a line between the contributions of measurement and statistical theory to the knowledge of reality. On the one hand, more accurate measurements lead to better research conclusions. On the other hand, statistics is needed to deal with measurement error (unavoidable, to some extent), to generalize measuremen...