Chapter VI advertising and Market Share Dynamics and DurahiCity of Advertising Effect on Sales
CHAPTER VI ADVERTISING AND MARKET SHARE DYNAMICS AND DURABILITY OF ADVERTISING EFFECT ON SALES 6.1 Introduction Economists have produced numerous studies of the effectiveness of advertising on firms market shares and durability of advertising s effect on sales. The motivation for these studies has been provided by the advertising durability hypothesis. This hypothesis asserts that periodic advertising contributes to a long lived asset called goodwill which positively affects future market share (Martin; 1993). Kelly Bird (2002) estimated the effect of advertising expenditure on market shares using Error correction model. Own and rival s advertising elasticities for the short run and long run were estimated along with durability of advertising effects. Empirical results showed that the effect of advertising on market share is relatively short lived, appearing to be substantially depreciated within one year for all firms. It is also found that advertising is less effective for foreign firms in the domestic market. The widely found positive relationship across industries between sales and advertising intensity has led to Elisabeth M.Landes and Andrew M.Rosen field (1994) to find that holding firm specific factors constant substantially increases current advertising s estimated effect on current sales in a Koyck lag model. She also found that holding firm specific factors constant dramatically reduces advertising s estimated useful life. This chapter presents the results and discussion of the durability of advertising effect on market share as well as on sales. 6.2 Previous Research Works Metwally (1975) and Roberts and Samuelson (1988) found that advertising in the cigarette industry is relatively long-lived, while Thomas (1989), Boyd and Seldon (1990) and Landes and Rosenfield (1994) found that advertising appears to be short lived. Clarke (1976) summarized the early studies of advertising's durability in his survey article. These found a wide range of depreciation rates for advertising, some very 181
high (e.g., 90 percent of the advertising's effectiveness was spent within two months) and some very low (e.g., 90 percent was spent over a period of several years). Clarke showed that the length of the estimated 90 percent duration interval was itself closely correlated with the periodicity of the data used in the study -a problem he termed "data interval bias." On the basis of studies using monthly, bimonthly, or quarterly data -studies less likely to be affected by data interval bias than those based on annual data -Clarke concluded that "the duration of cumulative advertising effect on sales is between 3 and 15 months; thus this effect is a short-term (about a year or less) phenomenon, (p. 355) The studies Clarke surveyed all used a simple Koyck lag model (or a variant), in which a firm's sales in one period depend on advertising undertaken in the current and previous periods. Many of the studies have continued to rely on the Koyck lag model. For example, Abdel-Khalik (1975), Picconi (1977), Falk and Miller (1977), and Thomas (19891 all applied Koyck lag models to estimate advertising's useful life. These studies analyzed time-series data for firms producing a variety of consumer goods and tended to find that advertising was short-lived. Only Abdel-Khalik (1975) found any evidence of long-lived advertising, and then only for two of the five products he studied. Empirical studies testing the advertising durability hypothesis have reported mixed results. Hence, this chapter attempts to analyse the durability of advertising effect on market share as well as on sales in selected Indian consumer durable products. 6,3 Koyck Lag Model The Koyck model relates a firm s market share to current period advertising expenditure and lagged period market share. The model is estimated using double logarithmic data for the period 1991-2006. MSt = f (At, MS,-0 Where MSt = market share in period t MSm = market share in period t-1 At = advertising expenditure in period t 182
The estimated equation is MSt = bo+bimst-i+b2at+u, bq, bj, and bj are the parameters to be estimated where bj, b2, > 0 and t>o will be zero or positive. 6.4 Durability of advertising effect on market share 6.4.1 Implied annual advertising depreciation rates in Colour TV Table 6.1 presents the results of the implied annual advertising depreciation rates using Koyck lag model. It is evident from the table that the estimated effect of advertising on current market share was positive in BPL and Sharp India. In the remaining four companies the estimated effect of advertising on current market share was negative and statistically significant in Bestavision, Videocon group and Onida group, which implies that advertising had counter productive effect on market share of these companies. The coefficients on lagged market share were positive in all companies. In two companies viz., BPL and Sharp India, the coefficient on lagged market share was greater than one, the instantaneous depreciation rate is computed as the inverse of the coefficient on lagged market shares minus one (Rao; 1986). Thus the coefficients greater than one yielded negative depreciation rates. Even though the current advertising reduced the market share of BPL, it is suggested that advertising had infinite life in these companies. Current advertising did not have a statistically significant effect on current market share in most of the companies. Four out of six companies implied annual deprecation rates were 50 per cent and above and hence, the simple Koyck lag model implied that advertising is short lived in the reference companies of Colour TV. 183
Table 6.1 Implied annual advertising depreciation rates in Colour TV Company Constant MS,., At 2 R D.W Implied annual depreciation rate (%) Bestavision 1.213 (3.118)** 0.446 (0.974) -0.876 (-2.148)** 0.65 2.246 71.13 BPL 0.029 (0.877) 1.159 (4.379)* 0.281 (1.631) 0.61 2.304 0.00 Sharp India 0.288 (3.532)* 2.105 (11.493)* 0.107 (0.551) 0.92 2.035 0.00 Philips 0.083 (0.997) 0.538 (2.998)* -0.644 (-1.565) 0.59 1.895 57.63 Onida Group 0.017 (1.890)** 0.462 (1.988)** -0.070 (-1.928)** 0.66 1.762 68.79 Videocon Group 0.103 (3.648)* 0.221 (0.108) -0.156 (-3.205)* 0.59 1.735 97.05 * 1 per cent level of significance ** 5 percent level of significance Note: Annual depreciation rate is calculated as 1-exp (-p) where p = the inverse of the regression coefficient of lagged market share (MS,.,) minus one (l/b,-l). 6.4.2 Implied annual advertising depreciation rates in Refrigerator Table 6.2 presents the results of the implied annual advertising depreciation rates in Refrigerator using Koyck lag model. It is evident from the table that the estimated effect of advertising on current market share was positive (but statistically not significant) in Electrolux Kelvinator and Whirlpool. In the remaining three companies, the estimated effect of advertising on current market share was negative but statistically significant in Voltas and Godrej group which implies that advertising had counter productive effect on market share of these companies. 184
The coefficients on lagged market share were statistically significant in four out of five companies. In two companies viz., BPL Refrigeration and Voltas, the coefficient on lagged market share was greater than one, and the instantaneous depreciation rate is computed as the inverse of the coefficient on lagged market shares minus one (Rao; 1986). Thus the coefficients greater than one yielded negative depreciation rates. Even though the current advertising reduced the market share of these companies, it is suggested that advertising had infinite life. In Electrolux Kelvinator and Godrej Group, the coefficient on lagged market shares were greater than 0.75. Even though, current advertising did not have a statistically significant effect on current market share in most of the companies, the simple Koyck lag model implied that advertising is long lived in the reference companies of Refrigerator except in Whirlpool, as the implied depreciation rate in whirlpool was 84.50 percent. Table 6.2 Implied Annual Advertising Depreciation Rates in Refrigerator Company Constant MS,., A, 2 R D.W Implied Annual Depreciation Rate (%) BPL Refrigeration 0.113 (1.185) 1.462 (3.886)* -0.487 (-1.413) 0.72 1.951 0.00 Electrolux Kelvinator 0.394 (2.674)* 0.791 (3.861)* 0.487 (1.413) 0.87 2.27 23.20 Voltas 0.033 (1.683) 1.286 (5.637)* -0.298 (-2.281)** 0.71 1.978 0.00 Whirlpool 0.154 (3.482)* 0.349 (0.181) 0.009 (0.635) 0.59 2.199 84.50 Godrej Group 0.007 (0.227) 0.931 (8.861)* -0.039 (-2.348)** 0.87 1.923 7.10 * 1 per cent level of significance ** 5 percent level of significance Note: Annual depreciation rate is calculated as 1-exp (-p) where p = the inverse of the regression coefficient of lagged market share (MS,./) minus one (l/hrl). 185
6.4.3 Implied annual advertising depreciation rates in Washing Machine Table 6.3 presents the results of the implied annual advertising depreciation rates in washing machine. The estimated effect of advertising on current market share was positive in four out of six companies, but statistically significant in IFB and Electrolux Kelvinator. The implication is that current advertising significantly contributed to the market share of these companies. The estimated effect of advertising on current market share was negative in Videocon Appliances and Onida Savak, which implies that advertising had counter productive effect on market share of these companies. In Onida Savak, the coefficient on lagged market share was greater than one, and hence, the implied depreciation rate was 0.00 per cent. Since, the instantaneous depreciation rate is computed as the inverse of the coefficient on lagged market shares minus one (Rao, 1986). Thus the coefficients greater than one yielded negative depreciation rates. Even though, current advertising reduced the market share of this company, it is suggested that advertising had infinite life in Onida Savak. The coefficient on lagged market share was negative in IFB and hence, the implied depreciation rate was 100 per cent. In five out of six companies the implied depreciation rates were greater than 50 per cent and above, which implies that the advertising is relatively short lived in these companies. The simple Koyck lag model implied that advertising is relatively short lived in the reference companies of Washing machine except in Onida Savak. 186
Table 6.3 Implied annual advertising depreciation rates in Washing Machine Company Constant MS,., A, 2 R D.W Implied annual depreciation rate (%) BPL Sanyo 0.055 (0.119) 0.335 (1.108) 0.994 (0.596) 0.62 2.010 86.26 IFB 0.056 (1.167) -0.472 (-2.546)* 0.559 (6.035)* 0.76 1.623 100.00 Electrolux Kelvinator 0.617 (2.041)** 0.315 (2.284)** 0.586 (2.373)** 0.60 1.693 88.63 Videocon Appliances 0.103 (3.196)* 0.558 (4.161)* -0.013 (-1.921)** 0.55 1.671 54.71 Onida Savak 0.891 (3.846)* 1.045 (1.818)** -1.012 (-2.403)** 0.68 1.826 0.00 Whirlpool Group 0.009 (0.771) 0.401 (1.361) 0.005 (0.149) 0.52 1.338 77.55 * 1 per cent level of significance ** 5 percent level of significance Note: Annual depreciation rate is calculated as 1-exp (-p) where p = the inverse of the regression coefficient of lagged market share (MS,.i) minus one (l/brl). 6.4.4 Implied annual advertising depreciation rates in Air Conditioner Table 6.4 presents the results of the implied annual advertising depreciation rates in Air Conditioner. The estimated effect of advertising on current market share was positive in two out of six companies. It was negative in four companies and statistically significant in Voltas, Blue Star and Carrier Aircon, which implies that current advertising had counter productive effect on market share of these companies. 187
In Carrier Aircon, the coefficient on lagged market share was greater than one, and hence, the implied depreciation rate was 0.00 per cent. Since, the instantaneous depreciation rate is computed as the inverse of the coefficient on lagged market shares minus one (Rao; 1986). Thus the coefficients greater than one yielded negative depreciation rates. Even though, current advertising reduced the market share of this company, it is suggested that advertising had infinite life in Carrier Aircon. In three out of six companies, viz., Videocon International, Amtrex Hitachi and Blue Star, the implied depreciation rates were greater than 50 per cent and above, which implies that the advertising is relatively short lived in these companies. In two companies, the implied depreciation rates were less than 50 per cent, indeed Carrier Aircon was effectively zero suggesting that advertising had an infinite life in these companies. The simple Koyck lag model implied that advertising had mixed effect on the reference companies of Air Conditioner. Table 6.4 Implied annual advertising depreciation rates in Air Conditioner Company Constant MSm A, 2 R D.W Implied annual depreciation rate (%) Voltas 0.007 (0.694) 0.850 (12.486)* -0.108 (-2.200)** 0.92 1.588 16.18 Videocon International 0.126 (1.141) 0.212 (0.717) 0.169 (0.516) 0.67 2.179 90.97 Amtrex Hitachi 0.0006 (0.087) 0.471 (6.042)* -0.015 (-1.350) 0.78 1.757 67.47 Carrier Aircon 0.035 (1.465) 1.121 (5.887)* -0.064 (-1.951)** 0.72 1.794 0.00 Blue Star 0.037 (1.760) 0.733 (3.049)* -0.074 (-2.058)** 0.59 2.337 74.69 Fedders Lloyd 0.163 (1.282) 0.604 (2.493)* 0.082 (0.331) 0.66 1.597 48.09 * 1 per cent level of significance ** 5 percent level of significance Note: Annual depreciation rate is calculated as 1-exp (-p) where p = the inverse of the regression coefficient of lagged market share (MS,./) minus one (l/hrl). 188
6.5 Durability of advertising effect on sales 6.5.1 Implied annual advertising depreciation rates in Colour TV Table 6.5 presents the results of the implied annual advertising depreciation rates using Koyck lag model. It is evident from the table that the estimated effect of advertising on current sales was positive in Philips and Videocon group but statistically significant in Videocon group. In the remaining four companies the estimated effect of advertising on current sales was negative and statistically significant in BPL, which implies that advertising had counter productive effect on current sales of these companies. The coefficients on lagged sales were positive in all companies. In three companies viz., Bestavision, BPL and Onida group, the coefficient on sales was greater than one, the instantaneous depreciation rate is computed as the inverse of the coefficient on lagged sales minus one (Rao; 1986). Thus the coefficients greater than one yielded negative depreciation rates. Even though the current advertising did not have a statistically significant positive effect on current sales in Bestavision, BPL and Onida group, the simple Koyck lag model implied that advertising is long lived in these companies. All of the implied annual depreciation rates were 25 per cent or less except Philips (since, it was 86.26 per cent). Indeed three companies implied annual depreciation rates were effectively zero, suggesting that advertising has an infinite life. 189
Table 6.5 Implied annual advertising depreciation rates in Colour TV Company Constant s,., A, ~R D.W Bestavision BPL Sharp India Philips Onida Group Videocon Group 2.492 (1.765) 0.434 (2.689)* 0.031 (0.178) 0.242 (6.030)* 0.018 (0.168) 0.080 (2.360)** * 1 per cent level of significance ** 5 percent level of significance 10.199 (2.052)** 2.144 (5.027)* 0.897 (1.848)** 0.335 (2.180)** 1.065 (4.177)* 0.820 (10.704)* -0.034 (-0.114) -0.440 (-2.343)** -0.031 (-0.406) 0.123 (1.034) -0.051 (-0.997) 0.062 (1.982) Implied annual depreciation rate (%) 0.38 1.784 0.00 0.75 1.654 0.00 0.50 1.574 10.85 0.71 2.345 86.26 0.90 2.313 0.00 0.93 2.091 19.71 Note: Annual depreciation rate is calculated as 1-exp (-p) where p = the inverse of the regression coefficient of lagged sales (S,,) minus one (l/hrl). 6.5.2 Implied annual advertising depreciation rates in Refrigerator Table 6.6 presents the results of the implied annual advertising depreciation rates in Refrigerator using Koyck lag model. It is evident from the table that the estimated effect of advertising on current sales was positive in 4 out of 5 companies, but statistically significant in Electrolux Kelvinator and Godrej group. The implication is that the current advertising had positive and significant effect on current sales of these companies. In Voltas, the estimated effect of advertising on current sales was negative and statistically significant, which implies that advertising had counter productive effect on sales of this company. 190
The coefficients on lagged sales were statistically significant in four out of five companies. In Voltas, the coefficient on lagged sales was greater than one, and the instantaneous depreciation rate is computed as the inverse of the coefficient on lagged sales minus one (Rao; 1986). Thus the coefficients greater than one yielded negative depreciation rates. Even though the current advertising did not have a positive effect on current sales, the simple Koyck lag model implied that advertising is long lived in this company. The coefficients on lagged sales were greater than 0.50 in four companies. All of the implied annual depreciation rates were 25 per cent or less except Godrej group (since, it was 80.44 per cent). Indeed in Voltas, the implied annual depreciation rates were effectively zero, suggesting that advertising has an infinite life. Table 6.6 Implied Annual Advertising Depreciation Rates in Refrigerator Company Constant Sm A, 2 R D.W Implied Annual Depreciation Rate (%) BPL Refrigeration 0.103 (0.568) 0.684 (1.323) 0.011 (0.114) 0.54 1.778 37.00 Electrolux Kelvinator 0.150 (4.701)* 0.594 (9.417)* 0.041 (1.970)** 0.87 1.819 49.52 Voltas 0.419 (2.300)** 2.163 (4.160)* -0.431 (-2.001)** 0.57 1.716 0.00 Whirlpool 0.156 (1.836)** 0.662 (3.513)* 0.006 (0.438) 0.57 2.301 39.98 Godrej Group 0.242 (2.954)* 0.380 (1.878)** 0.127 (2.263)** 0.54 1.701 80.44 * 1 per cent level of significance ** 5 percent level of significance Note: Annual depreciation rate is calculated as 1-exp (-p) where p = the inverse of the regression coefficient of lagged sales (S,_i) minus one (l/brl). 191
6.5.3 Implied annual advertising depreciation rates in Washing Machine Table 6.7 presents the results of the implied annual advertising depreciation rates in washing machine. The estimated effect of advertising on current sales was positive in 4 out of 6 companies, but statistically significant in IFB and Electrolux Kelvinator. The implication is that the current advertising had positive and significant effect on current sales of these companies. In the remaining two companies the estimated effect of advertising on current sales was negative, which implies that advertising reduced the current sales of these companies. The coefficients on lagged sales were positive in four companies. In BPL Sanyo, the coefficient on sales was greater than one, the instantaneous depreciation rate is computed as the inverse of the coefficient on lagged sales minus one (Rao; 1986). Thus the coefficients greater than one yielded negative depreciation rates. Even though the current advertising did not have a statistically significant positive effect on current sales in BPL Sanyo, the simple Koyck lag model implied that advertising is long lived in this company. The coefficient on lagged market share was negative in IFB and Electrolux Kelvinator and hence, the implied depreciation rate was 100 per cent. The implied annual depreciation rate was 21.76 per cent in Videocon Appliances and zero in BPL Sanyo, implying that advertising has long life. The implied annual depreciation rates were 50 per cent and above in Onida Savak and Whirlpool group. Indeed two companies implied annual depreciation rates were 100 per cent, suggesting that advertising has short life. 192
Table 6.7 Implied annual advertising depreciation rates in Washing Machine Company Constant S,., A, 2 R D.W Implied annual depreciation rate (%) BPL Sanyo 0.098 (1.026) 1.150 (6.408)* -0.028 (-0.101) 0.79 2.134 0.00 IFB 0.451 (6.382)* -0.351 (-1.654) 0.412 (4.826)* 0.86 1.738 100.00 Electrolux Kelvinator 0.007 (0.073) -0.323 (-2.942)* 4.822 (3.119)* 0.74 1.965 100.00 Videocon Appliances 0.099 (4.404)* 0.803 (15.511)* 0.026 (1.715) 0.93 2.002 21.76 Onida Savak 0.041 (1.047) 0.206 (2.436)** -0.052 (-1.464) 0.58 2.130 97.89 Whirlpool Group 0.182 (6.542)* 0.494 (5.662)* 0.018 (1.678) 0,93 2.089 64.09 * 1 per cent level of significance ** 5 percent level of significance Note: Annual depreciation rate is calculated as 1-exp (-p) where p = the inverse of the regression coefficient of lagged sales (S,.j) minus one (l/hrl). 6.5.4 Implied annual advertising depreciation rates in Air Conditioner Table 6.8 presents the results of the implied annual advertising depreciation rates in Air Conditioner. The estimated effect of advertising on current sales was positive in 3 out of 6 companies. The implication is that the current advertising had positive effect on current sales of these companies. In the remaining three companies, the estimated effect of advertising on current sales was negative and statistically significant in Carrier Aircon and Fedders Lloyd, which implies that advertising had counter productive effect on sales of these companies. 193
The coefficients on lagged sales were statistically significant and greater than 0.50 in 5 out of 6 companies. All of the implied annual depreciation rates were 25 per cent or less except Fedders Lloyd (since, it was 84.26 per cent), suggesting that advertising has long life in these companies. Table 6.8 Implied annual advertising depreciation rates in Air Conditioner Company Constant S,., At 2 R D.W Implied annual depreciation rate (%) Voltas 0.014 (0.330) 0.984 (9.079)* 0.023 (0.663) 0.89 1.894 1.61 Videocon International 0.137 (7.537)* 0.701 (8.531)* -0.154 (-1.277) 0.91 2.243 34.73 Amtrex Hitachi 0.113 (9.286)* 0.710 (18.873)* 0.003 (0.471) 0.92 1.629 33.53 Carrier Aircon 0.032 (1.415) 0.918 (15.448)* -0.044 (-3.094)* 0.95 1.581 8.55 Blue Star 0.043 (0.831) 0.934 (7.259)* 0.036 (1.010) 0.94 2.173 6.82 Fedders Lloyd 0.077 (1.892)** 0.351 (1.261) -0.117 (-1.961)** 0.55 1.868 84.26 * 1 per cent level of significance ** 5 percent level of significance Note: Annual depreciation rate is calculated as 1-exp (-p) where p = the inverse of the regression coefficient of lagged sales (S,./) minus one (l/b,-l). 194
6.6 Conclusion The conclusions drawn from the chapter are: P It is evident that the advertising effect on sales is relatively long lived in the reference companies of Colour TV, Refrigerator and Air Conditioner, and had mixed effect on sales of the reference companies of Refrigerator. P Current advertising had counterproductive effect on market share of Onida Group and Videocon in Colour TV, Voltas and Godrej in Refrigerator, Videocon Appliances and Onida Savak in Washing Machine and Blue Star and Voltas in Air Conditioner. P Current advertising had counterproductive effect on current sales of Bestavision, BPL, Sharp India and Onida Group and Videocon in Colour TV, Voltas in Refrigerator, BPL Sanyo and Onida Savak in Washing Machine and Videocon International, Carrier Aircon and Fedders Lloyd in Air Conditioner. 195
References 1. Bird, Kelly (2002), Advertise or die: advertising and market share dynamics revisited, Applied Economics, 2002, 9, 763-767. 2. Boyd, R. and Seldon, B. (1990), The fleeting effect of advertising: empirical evidence from a case study of the US cigarette industry, Economic Letters, 34(4), 375-79. 3. Clarke, Darral G, (1976), Econometric measurement of the Duration of Advertising Effect on sales, Journal of Marketing Research, Vol.13, No.4, (Nov, 1976), pp. 345-357. 4. Landes, E. M. and Rosenfield, A. M. (1994), Durability of advertising revisited, Journal of Industrial Economics, 43(3), 263-77. 5. Martin, S. (1993) Advanced Industrial Economics, Blackwell, Oxford. 6. Metwally, M. M. (1975), Advertising and competitive behaviour in selected Australian firms, Review of Economics and Statistics, 51,417-27. 7. Roberts, M. and Samuelson, L. (1988), An empirical analysis of dynamic, non-price competition in an oligopoly industry, Rand Journal of Economics, 19(2), 200-20. 8. Thomas, L. G. (1989), Advertising in consumer goods industries: durability, economies of scale and heterogeneity, Journal of Law and Economics, 32, 163-93. 196