MARKETING IN TIMES OF POLITICAL TURBULENCE

(1)

1

MARKETING IN TIMES OF POLITICAL

TURBULENCE

BY ELLA BURGOYNE

(2)

2

MARKETING IN TIMES OF POLITICAL

TURBULENCE

BY ELLA BURGOYNE

MSc Marketing Intelligence

Master Thesis

June 2019

University of Groningen Faculty of Economics and Business

First Supervisor: dr. ir. M.J. Gijsenberg Second Supervisor: dr. A.E. Vomberg

Student number: S3546632 Email: Ella.Burgoyne@outlook.com

(3)

3

ABSTRACT

This paper investigates how the effectiveness of the marketing instruments advertising, price promotions and innovations are affected by political turbulence, in comparison to economic turbulence. This explored through the analysis of a large dataset containing 6780 observations of several categories of durable consumer goods. A dynamic modelling approach is deployed, enabling some unique insights into the differential short run and long run effects of the interaction between turbulence and marketing effectiveness. Most significantly, political turbulence is found to negatively moderate the effectiveness of advertising and innovation in the long run. This is distinct from the findings for economic turbulence, for which turbulence is instead found to positively impact the effectiveness of price promotions in both the short and long run, but not alternative marketing instruments. The paper therefore carries important managerial implications in terms of the consideration of influential political events, such as the occurrence of a general election, in the scheduling of advertising campaigns and launch of new products.

(4)

4

PREFACE

This paper is the final piece for my graduation in MSc in Marketing, with a profile in Marketing Intelligence at the University of Groningen. Firstly, I would like to give a thousand thanks to my supervisor Maarten Gijsenberg. One of the first lectures that I attended at the University of Groningen was for the course Marketing Research Methods, given by Maarten. His enigmatic teaching style encouraged me to be excited by the power of data and statistics and therefore pursue the Marketing Intelligence track, a challenge which has been very much out of my comfort zone but incredibly rewarding. I have since found myself teaching SPSS to pre-master students and self-teaching myself how to code in R.

(5)

5

1. INTRODUCTION

Politics is undergoing an identity crisis on a global scale. The rise of far right nationalist and populist parties (BBC News, 2019), that appeal to the economic fears of working-class citizens, are reshaping the Western world’s established liberal and democratic ideologies (Financial Times, 2018), threatening the free movement of people, information and goods as we know it. It is these forces that have culminated in some of the most controversial political events to have occurred within the past few decades (Arnorsson & Zoega, 2018; Bieber, 2018). Namely, the election of Donald Trump as President of the United States and the UK’s referendum on its membership of the European Union, henceforth to be referred to as Brexit (Alabrese et al., 2016). Moreover, political tensions are also rising within nations, with right and left wing political views becoming further opposed. Factual evidence of this can be seen in the divisiveness of the results of both the election of Trump and the Brexit referendum, split 51.9% leave to 48.1% (BBC News, 2016a). The result of this is greater unpredictability in the outcome of political events, and therefore heightened levels of uncertainty and turbulence (Kotler & Casoline, 2009).

(8)

8

negative consequences on sales even after a period of turbulence has subsided, highlights the relevance of this topic to managers (Deleersnyder et al. 2004).

Furthermore, this existing body of marketing literature focuses solely on economic indicators of turbulence, with a distinct lack of research into whether these findings are applicable to the cases of political turbulence that we now see growing in impact and societal relevance. The evidently counterintuitive prevailing use of aggregate economic turbulence indicators can be explained from a methodological standpoint in that it is mainly a publicly available, readily accessible and easily modelled data source. In the case of assessing the political environment on the other hand, there exists no such concrete empirical indicator. Inferences can be made from the occurrence and outcome of key political events, such as votes concerning changes of leadership, but such data is rarely documented in one consistent format. As a consequence, there exists a lack of standardised and empirically validated political turbulence indicators, and therefore the area in general, and in particular the intersection of the political environment and marketing is greatly under researched.

1.1. RESEARCH AIMS

This research will thus seek to addresses this gap in the literature of how firms can most effectively tailor their use of marketing instruments during politically turbulent times through the following research questions:

- To what extent does the effectiveness of marketing instruments change during periods of political turbulence?

- Can a political turbulence indicator better explain changes in marketing effectiveness than traditionally used economic turbulence indicators?

- Do the findings differ in magnitude or direction across differently positioned product ranges?

These research questions will be explored through the analysis of a dataset containing monthly sales and marketing activity for several categories of durable consumer appliances, from one leading industry brand. This is in contrast to the dominant trend of existing

research in the field, which makes use of fast-moving consumer goods (FMCG) data to study the effects of turbulence on marketing mix effectiveness (Dekimpe & Deleersnyder, 2018). Whilst this body of research has found evidence of variation in the sales of FMCGs, it is important to note that these types of goods are generally essential items that are low priced and routinely purchased, regardless of external factors. In contrast, consumer appliances are characterised by higher levels of durability and comparably lower necessity, meaning that their purchase can be realistically postponed in response to external factors (Cook, 1999; Deleersnyder et al., 2004; Quelch & Jocz, 2009). This makes this research of particular interest, in the sense that the effects of turbulence are expected to be even more so

pronounced.

(9)

9

and finally discussing some of the complexities in interpreting research findings depending on the type of product that is studied. The chapter will conclude with synthesising the common themes of this literature into the conceptual model to be followed by this study. The next chapter will focus solely on the data underpinning this study, while the following methodology chapter will detail how the conceptual model is translated into an actionable model formula, to address the specific problem at hand with the data available. Subsequently, the research findings will be presented and evaluated, to be concluded with a discussion on the key underlying themes and implications for both future research and marketing managers.

2. THEORETICAL BACKGROUND

2.1. CONCEPTUALISING TURBULENCE

2.1.1. Market vs. Environmental Turbulence

In order to effectively review the literature on this topic, it is important to first conceptualise what the broad term of ‘turbulence’ should encompass, in terms of how research has approached it differently, and how this thesis shall define it going forth. Firstly, a distinction can be made between market turbulence and environmental turbulence (Mascarenhas, 2018). The former encompasses specific consumer driven changes in market demand, such as intra-year category demand cycles and seasonality effects, that by definition can last no longer than a year (Gijsenberg, 2017). In contrast, and of greater relevance to this research, the latter encompasses all aggregate changes in the external environment of a firm, such as the political, economic and social environment. This includes the widely researched economic form of turbulence, business cycles, which, unlike market turbulence cycles, are longer, lasting from two to eight years (Christiano & Fitzgerald, 1998). By nature, they are also therefore more unpredictable. Consequently, a wealth of empirical research exists exploring the dependency of various marketing instruments’ effectiveness on the economic business cycle (see Dekimpe & Deleersnyder, 2018 for an extended literature review). Political environmental turbulence, however, remains widely under researched within the marketing literature.

(10)

10

2.1.2. Economic Turbulence

Studies that have investigated economic turbulence can broadly divided into two streams: those focusing on the general pattern of business cycle fluctuations and a sub-stream of research investigating the specific business cycle event of recessions, otherwise referred to as ‘macroeconomic shocks’ (Barlevy, 2007). The key differentiator between these streams is the approach taken to defining the economic indicator. On the one hand, business cycle research takes a longer sighted, more holistic approach, as it is concerned with tracing the co-movement of macroeconomic indicators, such as output, employment and investment, over a longer time period (Christiano & Fitzgerald, 1998). While in contrast, research into recessions tends to span a shorter time frame, due to the inherent greater variability in economic indicators covered by the period before, during and after a recession. This is evident in its long-standing definition by the National Bureau of Economic Research as, “a [persistent] period of decline in total output, income, employment, and trade, usually lasting from six months to a year” (NBER, 2012). Having said this, it is important to highlight that more recent marketing scholarship has argued that turbulence in the form of extended periods of deep recessionary economic decline is fast becoming the new norm, suggesting that the traditional business cycle no longer exists, and therefore also the need to redefine what constitutes a recession (Kotler & Casoline, 2009). Consequently, this research will consider research findings relating to business cycle downturns and recessions jointly.

2.1.3. The distinguishing features of political turbulence

Having discussed the similarities of the two forms of turbulence, the existence of certain features that are unique to political turbulence make it an interesting phenomenon to investigate independently of economic forces. These features are namely disruptiveness, salience and involving nature.

Firstly, political turbulence can be argued to be more disruptive due to the fact that, unlike economic turbulence where the situation is generally expected to improve in time, it is not necessarily cyclical. Some cyclicality can be argued to exist in the form of regular election cycles (Smales, 2018), however, political events, such as changes in leadership, significant policy votes and referendums, still generally carry more dramatic and permanent effects. As a consequence, political changes are also far more salient in the media, and often hyper sensationalised and emotionally charged, in comparison to somewhat more difficult to comprehend and mundane changes to economic indicators, such as employment and GDP statistics. The prevalence of politics in the media is particularly evident in politicians’ exploitation of the media, to achieve their election agenda and campaign success (Scammell, 1996). The former British Tory Prime Minister Margaret Thatcher has been presented as a pioneer of this phenomena for her election campaign, during which she was assisted by the pro-Tory tabloid paper ‘the Sun’ and its readership of her target voters of skilled working-class readers.

(11)

11

be viewed as more concerning the consumer’s perceived uncertainty of future financial wellbeing.

Finally, cases of political turbulence can be further distinguished from economic turbulence by higher levels of consumer involvement, in the sense that individuals play a more active role in the determination of their outcome, by exercising their right to vote. Although individual differences do exist in the extent of members of society’s political engagement (Audit of Political Engagement, 2017), this in general means that they are more knowledgeable and aware of political turbulence, in comparison to economic. This has been argued to be a relatively new development in politics, coinciding with the rise of social media as a news source, resulting in not only less of a time delay in consumers receiving political news, but also increased engagement and mobilisation through the spread of news among peers (Jungherr, 2014; Kahne & Bower, 2018). Consequently, this forms the basis of this research’s prediction that it is logical to assume that political measures of turbulence can provide a better indicator than economic indicators of the impact of turbulence effectiveness of marketing instruments.

(12)

12

TABLE 1: LITERATURE OVERVIEW: TURBULENCE AND MARKETING Study Form Of Turbulence Measure Of Turbulence Marketing

Instrument Key Findings

Devinney (1990) Business cycles Gross Net Product (GNP),

inflation, investment Innovation

Introduction of new product innovations slightly leads the business cycle. Variation found depending on the metric used.

Axarloglou (2003) Business, demand & seasonality cycles

Growth rates of (non-seasonal) market demand & supply

Innovation

Timing of new product introductions most influenced by aggregate demand fluctuations, followed by market demand, business cycles, and lastly least by seasonality.

Barlevy (2007) Business cycles &

R&D investments GNP growth Innovation

R&D is more effectively carried out during recessions, despite current procyclical trend.

Fabrizo &

Tsolmon (2014) Demand cycles

Industry level output growth ( market demand)

Innovation

Timing of investments is procyclical, depending on the rate of

obsolescence. Argue for countercyclical R&D investments and procyclical new product introductions.

Jindal & McAlister

(2015) Market turbulence

Market sales growth volatility

Advertising & Innovation

Turbulence increase the effectiveness of R&D in the short run, in terms of reduced bankruptcy risk, but to decrease the effectiveness of advertising, which is more effective when the conditions are stable.

Deleersnyder et

al. (2004) Business cycles Quarterly cycle volatility

GNP Price Promotions

Sales of durable goods found to fluctuate with the business cycle, more so than alternative aggregate economic indicators, moderated by the nature of the durable good, PLC stage and inherent price volatility.

Lamey et al.

(2012) Business cycles GDP

Advertising, Innovation & Price Promotions

Pro-cyclicality effectiveness of all 7 marketing instruments, except for price premium. Greatest variability found for price promotions Gordon et al.

(2013)

Macroeconomic turbulence

Aggregate & quarterly

GDP growth Price Promotions

Overall consumer price sensitivity is countercyclical, and positively correlated with the average level of price sensitivity within a product category

Steenkamp &

Fang (2011) Business cycles GDP

Advertising & Innovation

R&D and advertising investments are more effective during contractions than expansions, moderated by the degree of industrial cyclicality Van Heerde et al.

(2013) Business cycles GDP

Advertising & Price Promotions

During a business cycle expansion, long-run price sensitivity is found to decrease, whereas advertising effectiveness is suggested to increase.

This thesis Political turbulence

Monthly aggregated indicator of key political events

(13)

13

2.2. RESEARCH FINDINGS ON THE OBSERVED & EFFECTIVE USE OF MARKETING INSTRUMENTS DURING TURBULENCE

The following section will focus in on discussing research findings relating to three of the seven marketing instruments identified by Borden in his seminal paper conceptualising the marketing mix (1964). These are advertising, price promotions and innovation. A majority of the research into economic turbulence considers either or both advertising and pricing simultaneously, while several authors identify them as the two most influential marketing tools (Bijmolt, Van Heerde, and Pieters 2005; Sethuraman, Tellis, and Briesch 2011). In addition, innovation is also considered an important marketing tool to explore in this research, due to its particular relevance to contexts of environmental turbulence, with it being argued that firms which view turbulence as an opportunity to develop and capitalise on new market opportunities ultimately achieve superior sustained competitive performance (Gulati et al., 2010).

2.2.1. Advertising

The primary purpose of advertising is commonly conceptualised as the acquisition of new customers, through both growing brand awareness and appeal to attract new customers, as well as strengthening the brand image to maintain existing customers (Bendixen, 1993). More specifically, it has been shown to play a critical role in developing long-term brand equity by increasing the brand familiarity and strengthening associations, thereby reducing price sensitivity (Nijs et al., 2001). However, in opposition to this broad perception of the value of advertising, Tellis (2004) proposes that the effects of much advertising are actually relatively short-lived, and although they can be strengthened through repeated exposures, this only works if the advertisement is effectively executed in the first place. Hence, it is often considered in practice as an expensive and unprofitable form of marketing investment. It should therefore come as no surprise that, of all the operating expenses, advertising is most commonly presented in the literature as the first to be cut during financially challenging or uncertain times (Srinivasan et al., 2005; Deleersnyder et al. 2009; Tellis and Tellis, 2009). Consequently, adverting investments have been shown to follow a pro-cyclical pattern, with firms reactively adjusting available budgets in line with economic expansions and contractions (Deleersnyder et al. 2009; Srinivasan, Rangaswamy & Lilien, 2011). This has been attributed to the relative ease of cutting advertising budgets compared to other marketing activities, in order to generate short-term cash flow (Lamey et al., 2007). In addition, this approach to dictating advertising investments can be explained by the difficulties in demonstrating advertising effectiveness, inherent in its definition as a tool for developing long-term brand equity and customer loyalty (Nijs et al., 2001), both being concepts that are difficult to measure in practice.

(14)

14

profitability (Steenkamp & Fang, 2011; O’Malley, Story & O’Sullivan, 2011) than those that decrease investments.

A logical explanation for this, is that with the dominant strategy during a recession being to cut, firms that do advertise benefit from the opportunity of reduced advertising clutter and competitive interference, thus making the same investments more effective (Steenkamp & Fang, 2011; Srinivasan et al., 2011; Van Heerde et al., 2013). However, the counter side to this logic is that as more firms adopt this strategy, these benefits diminish, therefore reducing the validity of this argument. This is supported by the well-established notion that competitive actions moderate the effectiveness of advertising on reducing price sensitivity (Gatignon, 1984). Consequently, pursuing such logic would suggest that if all firms adopted the alternative strategy of abstaining from advertising during a recession, all firms would benefit in the long-term, from reduced consumer price sensitivity.

The more compelling arguments for the effectiveness of increasing advertising investments during turbulent times stem from understanding the psychological effects of turbulence on consumers’ attitudes towards consumption and consequent behaviour. The existing literature suggests that advertising investments are more effective during recessions (Van Heerde et al., 2013), due to a combination of increased risk aversion and consequent price consciousness, resulting in reduced consumption inertia (Tellis & Tellis, 2009; Steenkamp & Fang, 2011). This suggests that in an economically turbulent environment, consumers will pay more attention to advertising during the information search stage and are therefore more receptive to advertising. Furthermore, it is known that that well-crafted advertising plays an important role in reducing the perceived risk of the purchase, either through making consumers more aware of the product’s benefits and quality (Wiggens & Lane, 1983), or through simply increasing the exposure and thereby familiarity of the product (Erdem & Keane, 1996). Consequently, the combination of this evidence suggests that during turbulence, firms can capitalise on the unique opportunity of both increased consumer information seeking and risk aversion, by increasing advertising investments and tailoring it towards combatting the negative psychological effects of turbulence. This suggests that the receptiveness of consumers to advertising and therefore its effectiveness increases during turbulent times, which are characterised by greater levels of salience and consequent negative psychological effects, meaning that advertising is more important in terms of reassuring consumers. The effectiveness of advertising investments is therefore expected to be positively moderated by political turbulence.

2.2.2. Price Promotions

(15)

15

goods. Furthermore, promotional elasticities have been found to be on average 10-20 times greater than advertising elasticities, therefore suggesting that the effective use of this instrument is even more important (Bijmolt et al., 2005; Sethuraman, Tellis & Briesch, 2011). In general, frequent and shallow price promotion has been found to be the most effective strategy (Srinivasan et al., 2004). However, during economically turbulent times, the frequency and extent of their usage is observed to increase significantly across a variety of industries. The key reason that firms will make greater use of price promotions during economic contractions is that it allows firms to generate immediate sales and cash flow at a time of financial hardship when survival of the firm may be the priority (Quelch & Jocz, 2009). However, research has shown that that price promotions generate little to no enduring effect on long-term sales (Nijs et al., 2001; Srinivasan et al., 2004). In contrast, any positive effect has been found to disappear after an average of just 10 weeks (Nijs et al., 2001), therefore suggesting that for firms that can afford it, this is not the most effective use of marketing budgets. To summarise, this suggests that price promotions can be very effective in the short run, but less so in the long run.

On the one hand, several longitudinal studies have established that price sensitivity is closely related to the business cycle, with consumers on average becoming more price sensitive and therefore responsive to price promotions as the economy weakens (Estelami, Lehmann, Holden, 2001; Goldfarb & Yang, 2013; Hampson & McGoldrick, 2013; Steenkamp & Maydeu-Olivares, 2015). Further support for the existence of this effect can be seen in the increased market-share of specialist discounter retailers and supermarket private-label branded goods during recessions, demonstrating that consumers actively exhibit low price seeking behaviour during turbulent times (Lamey et al., 2012; Lamey, 2014). A commonly cited logical explanation for the occurrence of this phenomena is that the average consumer’s available or anticipated future disposable income is reduced, and hence they become more risk averse and conscious of price in the decision making process (Estelami et al., 2001; Kamakura & Du, 2012).

Alternatively, Hampson and McGoldrick (2017) propose that social factors play just as, if not more, of an important role in this increased price sensitivity, with smart and frugal consumption choices becoming socially desirable during recessions, and frivolous spending rebuked. In applying this research to situations of political turbulence, it is important to remember that political events, for example elections, are not necessarily reflected in immediately perceivable changes to disposable income as with economic turbulence. However, they are characterized by high levels of uncertainty regarding future financial stability, hence the economic findings in this area are still very much applicable.

(16)

16

lower levels of inherent risk, for example categories where repeat purchasing is common. Consequently, it can be inferred from the existing economic research findings that the effectiveness of price promotions in stimulating short-term sales will increase as the political environment becomes more turbulent.

2.2.3. Innovation

In contrast to price promotions, the successful introduction of new product innovations to market during a recession has been empirically associated with permanent increases in sales and long-term profitability (Nijs et al., 2001). However, following the same logic discussed underpinning advertising investments; its use is also likely to be reduced during turbulence, due to its objective of generating long-term sustainable competitive advantage, as opposed to immediate cash flow injections (Srinivasan et al. 2011).

When considering ‘innovations’, it is first useful to make a distinction between investing in research and development (R&D) activities, and the subsequent stage of introducing the new innovation to market. This is because the effectiveness of the latter component is arguably affected by turbulence to a greater degree, as consumers become more risk averse and consequently likely less willing to try new products (Steenkamp & Fang, 2011). Furthermore, the purchase of products newly introduced to the market assumed to be driven by different purchase motivations to more mature goods that have been on the market for some time. In the context of durable goods, the purchase of goods in earlier stage of their life cycle are primarily driven by the desire to experience novel product benefits, such as new technological developments, as opposed to out of replacement necessity, such as the need to replace a broken household appliance, which is more commonly associated with mature products (Bayus & Steffens, 1998; Guiltinan, 2010). Consequently, it is important to account for the timing that a product is introduced to market in this form of research, independently and in addition to R&D investments.

(17)

17

therefore harder to predict and more disruptive, meaning that the marketing environment might permanently change, making the conditions unfavourable to launch an innovation.

The decision to delay innovations can once again be explained by the financial constraints and uncertainty that firms face during turbulent economic times, resulting in the short-term focused decision to delay investments in both R&D and new product launch campaigns until periods of stability or growth (Barlevy, 2007; Ouyang, 2011). An alternative explanation presented by Fabrizio and Tsolmon (2014), is that the observed reduced number of new innovations launched during turbulent economic times is an intentional strategic decision to better align new product introductions with periods of economic prosperity, when consumer demand and spending power is higher. The validity of this approach is further supported by the argument that consumers become more risk averse during economically turbulent times, and consequently less receptive to trying unfamiliar new products (Steenkamp & Fang, 2011, Erdem & Keane, 1996). This is because trying a new product will always come with a greater inherent risk as the consumer lacks the experience to directly evaluate the product’s attributes.

However, research in this area is lacking and inconclusive, with a more recent 12-year longitudinal study by Steenkamp & Maydeu-Olivares (2015) finding no significant economic cyclical variability in consumer openness to trying new market innovations. Consequently, the most convincing existing research in this domain essentially recommends the effectiveness of taking a balanced approach to investments in innovations. Fabrizio and Tsolsmon (2014) ultimately argue that firms should seek to maximise the utility of their available marketing budgets by taking a differentiated strategic approach to R&D and the subsequent introduction of new innovations to market. Specifically, R&D is more effectively carried out continuously during turbulence, with the smoothing of investments across the business cycle in order to sustain competitive advantage (Barlevy, 2007; Fabrizio & Tsolsmon, 2014). In contrast, the subsequent market introductions of the innovation is more effectively withheld until the turbulence has passed, when consumer demand will be higher and risk-aversion irrefutably lower (Fabrizio & Tsolsmon, 2014). Consequently, the effectiveness of new product introductions is expected to be negatively moderated by political turbulence.

2.3. INFLUENCE OF THE PRODUCT TYPE

2.3.1. Differences between product categories: Durables compared to FMCGs

(18)

18

(Berger & Vavra, 2015) to televisions (Deleersnyder et al., 2004). Moreover, Dutt and Padmanabhan (2011) make a further distinction between durables and semi durables, as well as non-durables and services. In addition to finding that durables are the most sensitive product category to economic crises, they also further investigate intra-category variation within durable goods, with the necessity of the good being the common overriding factor for all categories of goods in determining their sensitivity to economic turbulence. This finding is of particular relevance to the broad category classification of durable goods, for which additional research also finds support for the notion that the nature of the durable good, in terms of convenience and leisure, moderates the extent to which sales are sensitive to micro-level fluctuations in economic turbulence (Deleersnyder et al.,2004). Consequently, it is important to take into consideration the specific type of durable goods that the research has studied when interpreting the findings.

In comparing the predominantly researched categories of FMCGs and durables, the latter are arguably most affected by economic turbulence, because their purchase carries greater performance and financial risk and they are easier to realistically postpone, in contrast to more habitually purchased and essential FMCGs (Cook, 1999; Deleersnyder et al., 2004). As a result, it is evident that different product categories are affected by business cycle fluctuations to differing extents (Deleersnyder et al., 2004).

2.3.2. The role of product positioning

In addition to having a differential impact on marketing effectiveness between product categories, environmental turbulence can also be argued to have a varying impact on marketing effectiveness within categories, in terms of interacting with the positioning of products within the category. A common strategy of retailers and brands alike is to balance the positioning of their product or brand portfolio to capture more of the market, by fulfilling a range of consumer segments’ needs. This is commonly observed in the form of: good, better and best vertical brand extensions; price fighter ranges; and, in the context of FMCGs, private-labels and national A and B-brands (Keller & Aaker, 1992; Mohammed, 2018).

(19)

19

On the opposite end of the spectrum, demand for premium, high-end positioned products has also been found to increase during economically turbulent times, particularly in the categories of personal care and fashion (Nunes et al., 2008; Biciunaite, in Euromonitor

International, 2013; Sharma & Alter, 2012). This can be explained by the symbolic social status

and image benefits that they carry being even more so important to consumers during turbulent times, where consumers face reduced financial and psychological wellbeing (Millet et al., 2012), leading to a desire to compensate (Sharma & Alter, 2012). For example, evidence from designer handbag sales before, during and after the 2008 recession reveals that products displayed the brand logo far more prominently during the recession, therefore suggesting an increase in conspicuous motivated consumption during economic turbulence (Nunes, et al., 2008). In addition, risk aversion is known to increase in the face of uncertainty, meaning that consumers will favour established and trusted brands, over low-end non-branded. Consequently, this research expects that high-end positioned ranges will benefit during turbulent periods, as a means of self-image maintenance and a way of compensating for the heightened psychological effects of political turbulence, in comparison to economic. Finally, synthesising the above research findings suggests that the mid-end suffers twice during turbulent times: losing both consumers that detract to low-end positioned products to reap the price advantages, and those that shift to higher-end offerings to reap the compensatory image and self-esteem benefits. Existing consumers of high-end positioned product ranges however, are unlikely to detract to middle-end goods for price reasons during turbulence, due to being affluent enough not to feel the financial effects (Quelch & Jocz, 2011). This therefore suggests that marketing instruments will be most effective at increasing the sales of high and low-end ranges during political turbulence, while the sales of middle-end ranges are expected to suffer the most.

(20)

20

FIGURE 1: CONCEPTUAL MODEL

(21)

21

3. DATA COLLECTION & TREATMENT

The panel data underpinning this analysis has been drawn from a total of five different sources: GfK sell-out data warehouse; internal advertising activation records; publically available website archive pages and external political and economic indicator secondary research sources. The data from each of these sources has been joined by month and the product’s stock keeping unit (SKU). However, the variety of sources means that an extensive amount of pre-processing of the data was first required. The decisions taken in this stage will shape the validity of the outcomes of the analysis, hence the following section will provide detail on the specific steps taken to prepare and combine these datasets.

3.1. INTERNAL DATA SOURCES

3.1.1. Sell-out data

The extracted commercial data which formed the starting point for this research was provided by GfK and consists of in depth SKU-level sell-out datafor one leading brand in the UK consumer goods market. The key benefit of using sell-out data for this type of research is that it provides scanner collected information on the actual units of goods bought by consumers and for what price, therefore providing an accurate representation of consumer demand within a given time period. This is in contrast to sell-in data, which provides information on the number of units bought in by retailers, but not necessarily sold in the time period. This data is aggregated on a monthly basis, spanning a time period from May 2014 to December 2018. Furthermore, the extracted data includes all channels and touchpoints in the UK, combining both online and in-store sales. Three distinctive categories of durable goods have been selected: kitchen appliances, garment care and male grooming. This is then further split into sub-categories, creating the potential for additional insights into the type of product, in terms of the necessity of the good. The sub-categories of accessories and replacement parts were removed, due to the judgement that purchase motivation for goods from these categories is primarily driven by necessity and therefore not influenced by the political environment. Table 2 provides an overview of all the included categories and corresponding number of observations.

(22)

22

Subsequently, a second binary variable was created to indicate the occurrence of new product introductions (NPI). This was coded so that the first two months of sales observations for an SKU are indicated with a 1, with a benchmark of 0 taken to denote that the product is not considered a new introduction to the market. Observations that occurred within the first month of the observation period (May-2014) were not considered in this variable creation, as it is fair to assume that sales for the vast majority of SKUs continued before this observation period began.

TABLE 2: OVERVIEW OF INTERNAL SKU LEVEL DATA

Category Sub Category # of SKUs # of Sales Observations

1. GARMENT CARE 1. Iron 75 1448 2. Steam Generator 55 968 3. Handheld Steamer 4 27 4. Stand Steamer 2 19 2. KITCHEN APPLIANCES 5. Liquidiser 16 377 6. Juice Extractor 14 364 7. Food Processor 8 231 8. Hand Blender 8 158 9. Fryer 7 169 10. Chopper 3 68 3. MALE GROOMING 11. Men’s Shaver 67 1260 12. Beard Trimmer 21 502 13. Multi-Grooming Kit 19 440 14. Hair Clipper 15 385 15. Body Groomer 9 247 16. Nose Trimmer 5 117 TOTAL 16 328 6780 3.1.2. Advertising Campaigns

(23)

23

TABLE 3: TV ADVERTISING FREQUENCY PER PRODUCT CATEGORY

Product Category Advertising Frequency _{(# of months with a campaign)} Advertising Contribution _{(% of months with a campaign)}

Garment Care 127 4.59

Kitchen Appliances 78 3.78

Male Grooming 170 4.76

Overall 375 4.51

3.2. EXTERNAL DATA SOURCES

3.2.1. Website archives

The product range names; the positioning classifications in terms of good better or best; and Recommended Retail Prices (RRPs) were extracted from the brand’s publically available product listing archive pages. The listed RRP in the launch year of each product’s retail page is taken to reliably represent the regular base price of an SKU. Table 4 provides an overview of the average RRPs per a subcategory, revealing that Steam Generators are on average the most expensive category, but also exhibit the most price variation between SKUs, while the opposite is true for Nose Trimmers.

TABLE 4: AVERAGE SELLING PRICES

& USE OF PRICE PROMOTIONS BY SUB CATEGORY

(24)

24

Subsequently, this RRP and the previously extracted actual selling price could be combined into one variable to represent temporary price promotion effects at the time the consumer bought the product, commonly referred to as the price index (Van Heerde et al., 2004). This is calculated as the actual selling price divided by the regular RRP, therefore a 1 indicates that the product was bought at the full recommended retail price, and a value less than 1 and above 0 indicating that the product was purchased at a promotional price. For example, table 4 shows that Men’s Shavers were on average the most heavily discounted subcategory, with an average price promotion across all SKUs of 36% (price index = 0.64). The descriptive statistics presented in table 4 also reveal that Steamers are the least discounted products, with Stand Steamers selling for on average 17% more than the RRP. Although abnormal, this is entirely possible as the RRP is, by definition, the recommended price that the brand sets, but is open to interpretation by retailers that may chose to raise prices for goods that are selling well and in high demand. Consequently, it is important to consider these outliers in the analysis, as they may carry important information about when consumer demand is highest, in relation to the political and economic environment.

Finally, a categorical variable consisting of three tiers (low, mid and high end) was created from the range names and their associated good, better or best brand classification, to represent the product positioning. Figure 2 demonstrates that there are sufficient observations per a positioning tier, with a fairly even portfolio split to make this a viable method of investigating whether this variable moderates the relationship between turbulence and marketing effectiveness as the literature suggests.

3.2.2. Political Turbulence Indicator

The observation period May 2014 to December 2018 encompasses significant variation in the political landscape of the UK, with changes in leadership in both directions of right and left wing, as well as the significant referendum on the country’s membership of the European Union in June 2016. This timeline is summarised on a monthly aggregation level, with the use of dummy coding to represent the occurrence of influential political events. A benchmark of zero is taken to indicate that no substantial political event occurred in that month, with a 1 indicating that an important event did occur.

(25)

25

Date Event Description Sources Agreement

May-14 European Elections Support for the anti-European Union UK Independence Party (UKIP) surges in local _{and European elections.} BBC News (2018a) 100%

Sep-14 Scottish Referendum Scottish voters narrowly opt to remain part of the UK with 55% of the vote, to 45% _{favouring independence.} BBC News (2018a) 100%

May-15 General Election Conservative Party, led by David Cameron, wins majority, confounding all pre-election _{polls and making it the first time the party has been in power since 1992.} BBC News (2018a) 100%

Oct-15 Immigration Crisis Calais migrant crisis reaches peak media attention and disruption as migrants headed _{for the UK break through fences into the terminal.} BBC News (2015a) 66%

Dec-15 International Relations UK launches air strikes in Syria, as overwhelmingly voted for by Members of _{Parliament (MPs) but to much public controversy.} BBC News (2015b) 66%

Feb-16 Key Brexit Event David Cameron (PM) announces the EU referendum date – 23 June 2016. BBC News (2018a);

Walker (2019) 100%

Jun-16 EU Referendum Political crisis after voters in a referendum opt to quit the European Union. David _{Cameron resigns.} Walker (2019) 100%

Jul-16 New PM Theresa May becomes the new Conservative Party PM. Walker, (2019) 100%

Nov-16 Global Event Donald Trump is elected President of the US. BBC News (2016b) 100%

Jan-17 Key Brexit Event Theresa May delivers her first Brexit speech and publishes the Notification of _{Withdrawal Bill, revealing her intentions to pursue a 'hard Brexit'.} Walker (2019) 66%

Mar-17 Key Brexit Event Theresa May triggers Article 50, formally beginning the Brexit negotiations. Walker (2019) 100%

Jun-17 General Election* Theresa May calls a snap general election to strengthen her negotiating position. This _{results in a hung parliament and fragile Conservative party minority.} BBC News (2018a); _{Walker (2019)} 100%

Dec-17 Key Brexit Event An agreement is reached on the ‘divorce bill’, concerning the Northern Irish border _{backstop and EU and UK citizens’ rights. Negotiations progress to second phase.} Walker (2019) 100%

Jul-18 Key Brexit Event

Theresa May’s Cabinet agree on a Brexit plan, proposing an independent trade policy; separation of UK courts from the EU, and giving the UK control over the movement of people.

Walker (2019); BBC

News (2018b) 66%

Jul-18 Key Minister

Resignations

Two key Conservative ministers, including Foreign Secretary Boris Johnson, resign in

protest of the government's plans for free trade with the EU. BBC News (2018a) 33%

Dec-18 Vote of Confidence in

PM

Theresa May narrowly wins a vote of confidence in her leadership of the Conservative

Party. Walker (2019) 100%

(26)

26

In order to minimise the subjectivity of what constitutes an ‘influential’ or ‘important’ political event, a panel of three expert judges were consulted. Each judge composed the timeline independently, with no-predefined list of events in order to minimise bias in the selection of events. Clear instruction was given to focus largely on BBC News articles as sources, due to this being a reliable, politically unbiased source, but also the most popular source of news consumption across all ages of adults in the UK, as per a report commissioned by the UK communications body Ofcom in 2018 (Jigsaw Research, 2018). This was an especially important methodological consideration given this research’s aim of determining how political events affect consumers’ response to marketing instruments. Hence, it is important that the constructed political indicator focuses on events that received media attention and consumers were therefore aware of.

The resulting timeline is detailed in table 5, with only events that were independently identified by at least two judges (66% agreement or above) being included in the final indicator. In support of the validity of the resulting indicator, all three judges independently identified with 100% agreement nearly three quarters (71%) of the 14 included events (indicated in green in table 5). Just one event, in July 2018 was identified by only one judge and therefore excluded. A further single event, the appointment of a new Prime Minister (PM) in July 2016, was excluded due to the methodological constraint that events should not overlap. The Brexit referendum in the previous month, June 2016, was therefore included instead due to it being considered relatively more influential.

3.2.3. Economic Turbulence Indicator

This study will utilise GDP as a broad indicator of economic turbulence, in line with the majority of research in this area (see table 1). Monthly GDP data was sourced from the publically available UK government database (GOV.UK, 2019). The first step to determine from the raw data whether the economy was in a contraction or expansion, is to decompose the series into its trend component and business cycle component (Mintz, 1969). This was carried out in EViews, using the Christiano & Fitzgerald band-pass filter, a widely deployed method within the business cycle research; hence this paper will follow the same methodology to maintain consistency (Christiano & Fitzgerald, 1998; Van Heerde et al., 2013).

(27)

27

FIGURE 3: DECOMPOSITION OF GDP USING THE CHRISTIANO-FITZGERALD FILTER

3.3. DATA CLEANING

The following section will discuss steps taken to deal with abnormal and missing values, firstly from joining the data from the above varied data sources, secondly from transforming the data into a balanced panel dataset with 54 regular monthly observations for each SKU.

3.3.1. Prior to transformation

The dataset contains a total of just 64 cases of missing values for only two independent variables, representing a small percentage of less than 1% all 6780 rows of observations. Moreover, the observations belong to the same two independent variables: selling price and price index. Closer investigation of these cases reveals that they are missing due to the sales units of the SKUs in that month being zero. Considering that the variable selling price was created from dividing sales by volume, it therefore makes sense that dividing by zero sales units will produce NAs. Consequently, the decision was taken to recode these sales observations to NAs rather than zero and exclude from the analysis, so that the log of sales and price promotions can later be taken to enable their interpretation as elasticities.

3.3.2. After transformation

(28)

28

exhibiting continuous sales for the full observation period and a total of 62% of rows containing NAs. Consequently, it was very important to the outcome of the analysis to investigate and deal with them appropriately. Overall, three distinct cases of abnormal sales observations became apparent: instances of negative sales; very short observation periods of sales and missing sales observations for entire product categories.

Firstly, instances of negative sales value and units were taken to indicate that there were more product returns than sales within that month. This is logical, as for the vast majority of cases these observations occurred at the end of product life cycles, often several months after the product had been phased out from the market, therefore likely representing quality issues that are still covered by the product warranty. Consequently, the decision was taken to replace any negative values with NAs, as the motivations for returning products with quality issues do not relate to the specific aims of this research. This will also once again enable the log transformation of the associated variables for later ease of interpretation.

The second situation of short periods of observed sales for an SKU, specifically two months or less, were also treated as abnormal observations and replaced with NAs. This is because more in-depth investigation of these specific SKUs revealed that these sales represented loyalty and incentive bundling deals with other products, and therefore exhibited inflated sales that could confound the results.

(29)

29

4. METHODOLOGY

The research problem of isolating the effects of political turbulence on marketing effectiveness comes with several challenges that need to be accounted for in the model form and specification and therefore form the basis of this chapter. Firstly, a key challenge in investigating political turbulence is that it requires the consideration of both the lag and lead effects of political events, in order to effectively isolate the effect of just current political turbulence on marketing effectiveness. This is because there is evidence of consumers anticipating political changes and therefore displaying stockpiling consumption behavioural patterns, for example in the lead up to polling days (Smales, 2017). Likewise, research also suggests that political events continue to have enduring effects on consumer psychology and therefore receptiveness to marketing activities in the following weeks. However, it would be wrong to assume that these effects occur symmetrically in magnitude and duration surrounding events, with the actual observed consequences of the event likely to have a much greater and more enduring effect than the lead up to events. Consequently, the model needs to be able to account for asymmetrical dynamic effects. Secondly, the literature review also reveals that the model should allow for differential short and long run effects of turbulence on marketing effectiveness. Given these challenges and the research aims, of isolating the interaction effect of turbulence on marketing variables, the dynamic effects model is preferred, due to its flexibility in accounting for differential effects over time.

4.1. MODEL SPECIFICATION

The final model specification, presented below (formula __), takes the form of a finite distributed lag model. This means that the specified lags and leads have predetermined fixed end points, as the effects of turbulence or not anticipated to have a significant direct effect on sales for more than several months before or after. It is also autoregressive in the sense that a trend to account for sales of the previous month is included, in order to combat the issue of autocorrelation and allow for the calculation of long-run effects. The turbulence indicator is specified so that the model can be applied to both cases of political and economic turbulence, for which the model will be estimated three times separately for each of the product positioning tiers: low, medium and high.

(30)

30

Where…

𝑆𝑎𝑙𝑒𝑠_𝑡 Sales value in month 𝑡

𝑇𝐼_𝑡= 0 Turbulence indicator benchmark:

Case 1: no influential political event took place in month t Case 2: the economy was in an expansion in month t

𝑇𝐼_𝑡= 1 Turbulence indicator denoting:

Case 1: an influential political event occurred within the month t

Case 2: the economy was in an economic contraction in month 𝑡

𝑗 Effect window

𝑗 Start of the effect window i.e. the earliest lagged month t

𝑗 End of the effect window i.e. the last lead month t

𝐴𝑑𝑣_𝑡 Advertising in month 𝑡

𝑃𝑟𝑖𝑐𝑒𝑡 Price index in month 𝑡

𝑁𝑃𝐼_𝑡 New product introductions in month 𝑡

𝑆𝐷_𝑖 Dummy variable to account for fixed SKU effects, where 𝑛 is equal to the total number of SKUs minus 1

The remainder of this section will explore in more depth the key challenges and modelling considerations that resulted in this model specification (equation 1).

4.1.1. Variable transformations

Firstly, the natural logarithm is taken of all continuous variables that cannot be zero, namely sales and price. This carries the advantage that they can be interpreted as elasticities, and thus comparisons across product categories and ranges at different points in time can be made (Leeflang et al., 2015).

4.1.2. Determining the lag & lead window

Secondly, the decision of how many lead and lag event effect terms to include is made by specifying several models with differing lengths of effect windows and comparing the resulting information criteria, a common technique in the business cycle research literature (Van Heerde et al., 2004). Table 6 compares the adjusted R squared and Bayesian Information Criterion (BIC) of several models with differing lag structures. Both measures of fit penalise the model for adding parameters, and therefore for increasing the number of lag or lead parameters. The BIC is focused on as opposed to the adjusted R squared or alternative information criteria, as it is stricter in this penalty.

(31)

31

anticipate. Furthermore, the anticipation effects of stockpiling discussed in the literature is far more applicable to FMCGs, as opposed to the durable goods in this dataset.

Furthermore, the inclusion of a lagged turbulence indicator term does not justify the improvement in explanatory power for the economic model. However, it is important for the purposes of this research that both turbulence indicator models the same standardised schedule, to ensure that the results are comparable and one does not hold greater explanatory power than the other. Consequently, the schedule of one lagged term only is chosen but the notation of in the formula is left flexible, to enable the application of the formula to further research.

TABLE 6: CRITERIA FOR DETERMINING OPTIMAL LAG & LEAD SCHEDULE

none 1 lag 1 lead 2 lags

Adjusted R Squared Political _0.544 _0.544 _0.543 _0.541 Economic _0.545 _0.545 _0.543 _0.542 BIC Political 13562.2 13564.2 13273.1 13355.6 Economic _13566.6 _13564.7 _13278.0 _13363.0

4.1.3. Allowing for differential effects across time

Finally, the model is specified so that the carry over effect interacts with the turbulence indicator, thereby allowing for differences between the short and long run impact of turbulence on marketing effectiveness. The long run effectiveness of a marketing instrument during a period of political stability can therefore be calculated according to formula 2, where 𝛽_𝑥 represents beta for the main effect of the marketing instrument of interest:

(2) 𝑦 = _1−𝜆𝛽𝑥

Considering that the carry over effect is equal to the beta of the lagged sales term, this equation is equivalent to:

(3) 𝑦 = 𝛽𝑥

1−𝛽5

During a period of political turbulence, this formula can be extended to include the betas for the interaction terms of both the marketing variable of interest and lagged sales with the turbulence indicator. This is depicted in formula 4, where 𝛽𝑧 is equal to the beta of the

interaction between the variable of interest and the turbulence indicator, with 𝛽9

representing the interaction between lagged sales and the turbulence indicator.

(4) 𝑦 = 𝛽𝑥+𝛽𝑧

(32)

32 4.2. CHECKING MODELLING ASSUMPTIONS

The following section will proceed to check the underlying assumptions of the model, which in turn have had implications for the model’s specification. The associatiated tests for these assumptions haven been carried out on the political indicator version of the model only, due to this being the main focus of the research problem. However, the results are expected to be similar.

4.2.1. Pooling

A first key assumption of the model is that the observations for all categories and SKUs can be stacked, resulting in the estimation of a single intercept and set of coefficients that are applicable to all products. While this does mean the loss of information, it is the preferred approach for this research case, as the data involves a restricted observation period of a limited number of product categories for just the one brand. Pooling across categories and SKUs therefore means that more observations are available to estimate the coefficients with greater statistical significance (Leeflang et al., 2015). In order to formally test whether observations in this panel dataset can be pooled, the Goldfeld-Quandt test is appropriate (Croissant & Millo, 2008). The test results in a Goldfeld-Quandt test statistic of 0.9482 that is insignificant at 3042 degrees of freedom (p=1.434). The null hypothesis that the variance between SKUs does not significantly differ should therefore be accepted. This means that the same set of coefficients is applicable to all SKUs and that pooling is acceptable.

However, it is still considered important to account for unobserved differences in the relationships between SKUs and the levels of the independent marketing variables, SKU effects should be accounted for in the model. For example, this can account for one product being more heavily advertised than another. This is especially important to consider due to the limitation in the available data that advertising activity is specified as a binary dummy indicator, therefore giving no indication of the extent of advertising investment. SKU effects can be accounted for in the form of either fixed or random effects. The inclusion of fixed effects in dynamic models means that the intercept may vary with individual SKU effects and the timing of the observation, while the estimation of all other parameters is fixed and therefore assumed to be constant across all SKUs, while the random effects approach is more flexible, allowing for each SKU to have an individually estimated error term (Croissant & Millo, 2008). This reduces the degrees of freedom and makes the interpretation of parameters more complex, therefore the inclusion of fixed effects is preferred. The Hausman Test can be used to formally test for whether modelling fixed effects is acceptable, with the null hypothesis that allowing for random effects provides a better fit for the data, and the alternative that fixed effects provide a sufficient or better fit (Croissant & Millo, 2008). The resulting chi-squared test statistic is highly significant at 9 degrees of freedom (chi square statistic = 208.988, p-value < 0.001), therefore confirming that alternative hypothesis should be accepted and that the use of fixed effects to account for differences between SKUs is recommended.

4.2.2. Autocorrelation

(33)

33

estimates of the variance of effects. It is especially important to investigate the issue in the context of this research, as it is dealing with time-series sales data within which observations are commonly correlated with the historical sales trend. The most widely utilised method for testing for autocorrelation is the Durbin Watson test (Durbin, 1970). However, this test statistic has been found to be unreliable when lagged independent variables are included as in this model (Leeflang et al., 2015). Alternative methods appropriate for modelling panel data that do capture dynamic effects utilise the Langrage Multiplier, such as the Breusch-Godfrey/Wooldridge test, which is particularly adept at identifying autocorrelation in the errors of relatively short panel datasets (Wooldridge, 2002; Croissant and Millo, 2008). The results of the Breusch-Godfrey test for serial correlation of an order up to 1 on the pooled model with fixed SKU effects gives an chi-square statistic of 13.9 at 1 degrees of freedom, which is significant on the 0.01% level (p=0.0003). This means that the null hypothesis of no serial correlation should be rejected and the alternative hypothesis that of significant correlation in the error terms should be accepted. As a consequence, the sales trend variable was included in the model, both as a main effect and as an interaction term with the turbulence indicator, a recommended method for combatting autocorrelation (Leeflang et al., 2015).

4.2.3. Multicollinearity

Analysis of the variance inflation factors (VIF) scores can help to determine whether the variables included in the model are sufficiently independent, therefore enabling the model to accurately determine their individual effects on sales. From table 7, it can be seen that all of the independent variables in the political turbulence model exhibit low levels of multicollinearity, with VIF scores of below 4 (Malhotra, 2004). However, the current period and one month lagged economic turbulence indicator variables in the second model can be said to carry moderate levels of multicollinearity, as they are both greater than 4 yet below 10. This makes sense when considering how this indicator has been constructed, as a binary variable that is either a one or zero for several months in a row, to indicate a period of consecutive economic growth or decline. It therefore makes sense that there is a moderate degree of similarity in terms of the evolution of their impact on sales. The benefits of including the lagged turbulence indicator term, primarily in that it enables the comparison of the two models, therefore outweigh the moderate risk of not being able to distinguish whether it is economic turbulence of the current or previous month that is having an impact on sales. Consequently, all of the specified independent variables should be included in the model.

TABLE 7: VIF SCORES FOR INDEPENDENT VARIABLES

TI TI t-1 Adv PI NPI Sales t-1 TI: Adv TI: Price TI: NPI

Political 2.583 1.102 1.357 1.364 1.240 2.583 1.410 2.856 1.248

(34)

34

4.2.4. Seasonality

A final consideration in the model specification was whether the sales exhibit any seasonal trends. To investigate this, the sales observations were aggregated on a quarterly level. Subsequently, an analysis of variance (ANOVA) was carried out to test for whether the variation in sales between the four groups is significant. The results presented in table 8 reveal that there are significant differences between the observed sales in each season, with an F-value of 25.26, that is significant at 3 degrees of freedom (p=<0.001). However, decomposing the sales by product category reveals that only the sales of Male Grooming exhibit significant seasonality; also clearly visible in figure 4 depicting the aggregated sales trend over time per a category. Consequently, it makes sense to account for seasonality in the inclusion of the fixed SKU effects dummies.

TABLE 8: ANOVA TO TEST FOR SEASONALITY

Degrees of Freedom

Sum of Squares

Mean Sum of

Squares F-value P-value

Quarter 3 685100000 228365342 25.26 < 0.001

(35)

35 BREXIT REFERENDUM ARTICLE 50

5. RESULTS

5.1. MODEL FREE INSIGHTS

5.1.2. Descriptive analysis

FIGURE 4: TIME SERIES OF MONTHLY SALES TREND PER CATEGORY

(36)

36

5.1.3. Correlation analysis

Analysis is of the correlations between the key variables provides some further initial insights into the direction and magnitude of relationships (see table 9).

TABLE 9: CORRELATION MATRIX FOR KEY VARIABLES

(1) (2) (3) (4) (5) (6) (7) (8) (1) Positioning 1.00 (2) Sales Value -0.26 1.00 (3) Sales Units -0.39 0.95 1.00 (4) Price Index -0.37 -0.48 -0.28 1.00 (5) NPI -0.05 -0.34 -0.35 0.07 1.00 (6) Advertising -0.12 0.27 0.19 -0.33 -0.25 1.00 (7) Political TI -0.10 -0.13 -0.14 -0.18 -0.21 -0.12 1.00 (8) Economic TI -0.4 -0.16 -0.18 -0.03 -0.17 -0.14 -0.03 1.00

With regards to the effectiveness of marketing instruments, interpretation of the correlation coefficients reveals that there exists a strong, positive correlation between advertising and sales, support the notion that advertising is generally effective at increasing sales. The marketing instruments new NPI and price index also are also positively correlated. However, in interpreting this relationship it is important to take into account the meaning of the price index variable, in that a higher observation closer to one indicates that the product was sold for closer to the recommend full retail price. This relationship therefore suggests that new products are often introduced to the market at a higher price, and are not put on promotion until later in their lifecycle. Similarly, NPIs are also negatively related to sales, confirming the expectation that the sales of products newly introduced to the market are on average lower than later in the products lifetime, as consumers becoming more aware of the product and its benefits. Finally, the marketing variables price index and advertising also exhibit a relatively strong, negative relationship, therefore indicating that price promotions are commonly deployed in combination with key advertising campaigns.

MARKETING IN TIMES OF POLITICAL TURBULENCE

MARKETING IN TIMES OF POLITICAL

TURBULENCE

BY ELLA BURGOYNE

MARKETING IN TIMES OF POLITICAL

TURBULENCE

BY ELLA BURGOYNE

MSc Marketing Intelligence

Master Thesis

June 2019

ABSTRACT

PREFACE

CONTENTS

1. INTRODUCTION

2. THEORETICAL BACKGROUND

3. DATA COLLECTION & TREATMENT

4. METHODOLOGY

5. RESULTS