Masaryk university



Download 3.14 Mb.
Page9/11
Date30.04.2018
Size3.14 Mb.
1   2   3   4   5   6   7   8   9   10   11

Women’s market


Second market affected by the P&G/Gillette merger was the market for production and distribution of women’s deodorants (women’s market). According to FTC’s decision, the size of merger’s expected anticompetitive effect on this market was not as large as on men’s market, however some competition concerns existed. In the end, FTC ordered merging parties to offer an option to the buyer of divested men assets to acquire selected assets also from the women’s market. This led to two brands (Dry Idea, Soft&Dri) previously owned by Gillette being bought by Henkel. In this chapter I investigate the efficiency of this divestment and compare it to alternative scenarios varying in the assets sold and the acquirer of these assets.
      1. Descriptive statistics


At the time of the merger, stick format deodorants represented the largest shares of sales on the market with 65 % market share. Deodorants made in aerosol on roll-on form both had comparable level of sales constituting between 16-19 % of the market (see Table 37). Ranking formats on the women’s market according to their level of sales corresponds to their ranking on the men’s market, even though higher preference for roll-ons and lower preference for stick deodorants at women’s market relative to the men’s market is observed. Solid deodorants on women’s market are also almost equally divided between standard deodorants and deodorants specially designed for leaving no marks on dark-colored clothes (“invisible”).

Table 37 – Overview of product differentiation






Shares

(%)

Price

(mean)

Size

(mean)

Gel

(%)

Invisible

(%)

Stick

65.34

1.18

2.53

26.59

59.02

Roll-on

16.05

1.35

2.71

0.12

42.51

Spray

18.61

0.70

5.86

0.00

0.47

The following table (see Table 38) shows descriptive statistics for price, market share and container size – variables used in demand estimation.

Table 38 – Descriptive statistics






Mean

SD

Min

Max

Price ($)

1.12

0.35

0.15

5.40

Market share

0.005

0.006

0.00

0.60

Size (oz.)

3.18

1.40

0.35

9.60
      1. Market structure


The market for production and distribution of women’s deodorants was less concentrated before the merger compared to its men’s counterpart. Market shares for period 2002-2006 of the companies present on the market are shown in Table 39. Prior to the merger, there were seven producers with P&G and Unilever being the two largest companies (each with approximately 31 % market share). They were followed by Church & Dwight, Colgate, Gillette and Kao - competitors with comparable market shares between 8-12 %. Revlon had almost marginal market presence with share below 3 %.

Table 39 – Market shares for companies on women’s market



Company

2002

2003

2004

2005

2006

Church & Dwight

11.97

12.85

11.19

10.85

9.99

Colgate

9.83

8.81

8.38

8.72

9.16

Gillette

9.36

8.62

8.28

-

-

Kao

8.79

8.2

7.7

6.98

6.81

P&G

30.64

30.77

31.39

37.60

22.57

Revlon

2.23

2.27

2.41

2.59

2.66

Unilever

27.19

28.48

30.66

33.26

35.3

Henkel

-

-

-

-

5.97

Innovative Brands

-

-

-

-

7.53

Note: Based on units of production (in %)

The merged entity would amount to 37.6 % market share, with the closest rival – Unilever – having only slightly lower market share (33 %). This would effectively turn the market into a duopoly as the closest rival to the merged entity and Unilever would have approx. one third of their individual market share.

As in the men’s market, the reported market shares are based on production (rather than sales). This enables to include the transition year 2005, where P&G was still in charge of producing divested deodorant brands (Dry Idea, Soft&Dri), but the approved buyer (Henkel) was in charge of all business activities connected with those assets. Market share of P&G in 2005 reflects that. Only in 2006 did Henkel fully took over the technological production and became completely independent of P&G.143

Table 40 shows market shares between 2002 and 2006 for individual brands. Approximately 22 % of all deodorants were sold under the brand Secret, followed by Degree (14 % market share) and several smaller brands (Arrid, Dove,Mennen, Sure) with comparable market shares around 8 %.

After the merger, both brands previously owned by Gillette were sold to Henkel as part of remedies approved by FTC. P&G was therefore left with assets it owned prior to the merger. Henkel was not present on the market before the merger and used this opportunity for a market entry.

Table 40 – Market shares for women’s brands



Brand

Owner

2002

2003

2004

2005

2006

Arrid

Church&Dwight

11.97

12.85

11.19

10.85

9.99

Ban

Kao

8.79

8.2

7.7

6.98

6.81

Degree

Unilever

13.93

13.96

14.02

15.68

17.2

Dove

Unilever

9.55

9.81

9.89

10.75

10.98

Dry Idea

Gillette

3.73

3.48

3.44

2.9

2.08

Mennen

Colgate

9.83

8.81

8.38

8.72

9.16

Mitchum

Revlon

2.23

2.27

2.41

2.59

2.66

Secret

P&G

22.23

22.66

23.27

22.31

22.57

Soft&Dri

Gillette

5.63

5.14

4.84

4.59

3.89

Suave

Unilever

3.7

4.7

6.75

6.83

7.12

Sure

P&G

8.41

8.11

8.12

7.8

7.53

Note: Based on units of production (in %)

In September 2006, P&G sold Sure deodorants to a private investment company Innovative Brands (Berkmann and Byron, 2006). Sure brand of deodorants was known for being unscented, whereas all other brands owned by P&G (including the ones on the men’s market) were scented deodorants. Given the ongoing market strive for gender segmentation, scented deodorants can be more easily tailored to gender-specific tastes of consumers and were therefore considered more strategically valuable to P&G. It is questionable, whether P&G planned this step at the time of the merger and communicated its intent of selling Sure to FTC. This question is relevant for correct definition of the counterfactual, against which the merger was judged. However, in order to make the intent of selling Sure credible (and therefore permissible into the counterfactual analysis), FTC would likely include the sale as an extra condition in its Decision. Since there is no such condition mentioned, I assume that the correct counterfactual to the P&G/Gillette merger on women’s market were pre-merger conditions.

Comparing the market structure for men’s deodorants and women’s deodorants (compare Table 9 and Table 39), it is apparent that the competitive environment at both markets significantly differs. The number of market players, their market shares or brands they offer differ between both markets. Some companies were also present only one of the gender-specific markets and most brands were gender specific and present only on one the markets market. The fact that the competitive environment significantly differs between men’s and women’s market for deodorants supports the existence of gender-specific markets (see section 4.3.1 for more arguments supporting gender segmentation).

Using the observed market shares, I calculate HHI and concentration ratios – indicators of market structure described in U.S. Horizontal Merger Guidelines regularly used by FTC for screening merger’s likely effects.144 The pre-merger HHI was 2 254 points and increased to 2 774 points after the merger. According to the U.S. Horizontal Merger Guidelines, such markets are considered highly concentrated and any merger leading to a higher change in HHI than 200 points is presumed to likely enhance market power. In case of the investigated P&G/Gillette merger, the increase in HHI amounted to 520 points. This makes the merger anticompetitive by presumption. Concentration ratios C2 and C3 show that the two largest players account for 62 % of the pre-merger market and the three largest players represent 73 % of the entire women’s market. According to the structural indicators, the merger would lead to significant anticompetitive effects and therefore raises a competitive concern.

Table 41 – Market structure indicators for 2004


Indicator

Value

Number of firms pre-merger

7

HHI (pre-merger)

2 254

HHI (post-merger)

2 774

HHI delta

520

C2

62.05

C3

73.24
      1. Demand estimation


Demand on the market for women’s deodorants was estimated using the structural model described in section 3.2. Similarly as in men’s market, I use format as a nesting variable and assume consumers’ preferences differ between stick, roll-on and spray deodorants. I also follow the same logic when choosing instrumental variables as when estimating demand on the men’s market. The values of correlation coefficients between inputs necessary for deodorants’ production (i.e. cost shifters) and endogenous variables are shown in the top part of Table 42. The bottom part of the table refers to the correlation of endogenous variables with the actual instruments used in the demand estimation. More detailed information on both the nesting structure and the choice of instruments are provided in section 5.2.3.1. as part of men’s market analysis.

Table 42 – Correlation of endogenous variables and IVs









Price

Within share

Vegetable and Animal Fats

0.008

-0.032

Coconut Oil

0.143

0.082

Plastic and crude oil

-0.083

-0.152













AVG price in other areas

0.743

0.129

Sum of rival cont.size

0.285

0.250

Sum of rival cont.size in nest

0.184

0.285

Sum of cont.size in nest

0.414

0.372

Demand estimation results are shown in Table 43. Four models are presented - the main model used in further analyses (Model 1) and three other models used as robustness checks of the main model. In order to verify the sensitivity of Model 1 to certain changes, I re-estimated the model using different sets of instrumental variable (Model 2), more aggregate product definition (Model 3) and larger potential market size (Model 4). More details on specification of each model are described in Table 44.

Estimated parameters α and σ have expected values (i.e. parameter α is negative and for nesting parameter is true that ), both are statistically significant at 1 % significance level and have comparable values across all four specifications. Comparing the results from the presented models, we see that the estimates of the initial model (Model 1) are robust to the choice of instrumental variables, product definition and potential market size.



Table 43 – Demand estimation results for women’s market




Model 1

Model 2

Model 3

Model 4

Model 5

VARIABLES

market size: r=1

r=4

OLS



















Price coeff. α

-1.178***

-1.166***

-1.132***

-1.178***

-0.461***




(0.235)

(0.237)

(0.253)

(0.235)

(0.124)

Nesting parameter σ

0.302***

0.315***

0.252**

0.305***

0.868***




(0.0889)

(0.0943)

(0.110)

(0.0888)

(0.020)

Constant

-2.856***

-2.775***

-3.125***

-4.222***

-0.806




(0.887)

(0.894)

(1.012)

(0.886)

(0.344)



















Geographical market FE

Yes

Yes

Yes

Yes

Yes

Time FE

Yes

Yes

Yes

Yes

Yes

Brand FE

Yes

Yes

Yes

Yes

Yes

Label FE

Yes

Yes

Yes

Yes

Yes



















Obs

39320

39320

35005

39320

39329

F-stat

110.2

109.0

179.6

108.0

1608.7

RMSE

0.992

0.975

1.073

0.987

0.448



















First stage F-test Price

1440

1370

608.7

1440




First stage F-test Share_nest

21.50

18.40

15.60

21.50




Sheas Part.R2 Price

0.659

0.659

0.646

0.659




Sheas Part.R2 Share_nest

0.188

0.188

0.163

0.188




Sargan test PV

0.390

0.128

0.828

0.285




Notes: Robust standard errors (clustered by product) in parentheses; *** p<0.01, ** p<0.05, * p<0.1

Table 44 – Models’ description



Model

Brief description

Model 1

Initial model

Instrumental variables used: average price of the product in other metropolitan areas, sum of container size for all rivals’ products on the market, sum of container size for all and rival products within the nest



Model 2

Different set if IVs used

average price of the product in other metropolitan areas, number of rivals’ products on the market, number of all and rivals’ products within the nest



Model 3

More aggregate product definition used

Product characteristics “invisible” and “gel” were aggregated over



Model 4

Parameter r for potential market size = 4

Model 5

OLS estimation

The comparison of Model 1 and Model 5 give a first indication of the strength of instruments. Given that the endogeneity bias in estimation of nested-logit models is known (price coefficient is biased towards zero and nesting parameter towards one), we can see that results of instrumental estimation indeed shifted the resulting estimates in the correct direction. This is the first indication of successful choice of instrumental variables. The statistical significance of instruments is also indicated by the reported F-test from a first stage. According to Staiger and Stock (1997), the value of the F-test can be judged against a simple rule of thumb: the higher the F-stat is above value of 10, the better are instruments modelled in the first stage. Secondly, it is important to look at whether the correlation between instruments and endogenous variables is high enough. The value of correlation coefficients is reported in Table 42 and is tested by reported Shea’s partial R-squared (see Shea, 1997).

Other nesting structures were tried. Correlation of consumers’ preferences could be more precisely captured by using a two-level nesting of demand. Unfortunately, using two-level nested logit model did not return meaningful results in any attempted specification. Using deodorants’ format as nests and “invisible” characteristic for subnests returned parameter identifying correlation of preferences in nests greater than parameter identifying the correlation of preferences in subnests. Such results contradict the random utility theory underlying the demand model. Using nests based on whether the deodorant is solid or spray and then dividing solid deodorants into two formats (stick, roll-on) did not return reasonable as well - the nesting parameter was outside the interval . Other one-level nesting possibilities (e.g. by “invisible” characteristic) offered smaller amount of nests, hence lower ability to model consumers’ preferences and were generally outperformed by the main model (Model 1).
      1. Demand elasticities


Using the available data and acquired demand estimates, I calculated own-price cross-price elasticities on product level (see Equation 3 and 4). Descriptive statistics for obtained product-level elasticities are shown in Table 45. All own-price elasticities are negative and cross-price elasticities between products inside the same nest are higher than between products from two different nests. This is in line with expectations and properties of the estimated demand model.

Table 45 – Elasticities for the main demand model






N

Mean

SD

Min

Max

Own-price elasticity

39 329

-2.131

0.866

-9.12

-0.256

Cross el. inside nest

39 329

0.023

0.039

0.00

0.334

Cross el. outside nest

39 329

0.007

0.009

0.00

0.077

Table 46 shows own-price elasticities on a brand level referring to a change in given brand’s demand if all the products sold under that brand’s name increase by 1 % in price. Results are presented for two parameters of potential market size (i.e. Model 1 and Model 4).145

Table 46 – Brand-level own-price elasticities of demand



Brand

Company

Own-price el. (r=1)

Own-price el. (r=4)

Arrid

Church&Dwight

-1.31

-1.33

Ban

Kao

-1.91

-1.93

Degree

Unilever

-1.60

-1.62

Dove

Unilever

-1.77

-1.79

Dry Idea

Gillette

-2.47

-2.50

Mennen

Colgate

-1.81

-1.83

Mitchum

Revlon

-2.67

-2.69

Secret

P&G

-1.69

-1.71

Soft&Dri

Gillette

-1.73

-1.74

Suave

Unilever

-1.12

-1.13

Sure

P&G

-1.46

-1.47

Demand elasticity on the market-level is calculated as a percentage change of the total volume of sales assuming all products sold on the market increased in price by 1 % (see Table 47).

Table 47 – Market elasticity for women’s market






R = 1

R= 4

Market elasticity

-0.467

-0.774

Even though the presented results are calculated for two different parameters of potential market size, further analyses use only smaller value of the parameter . As described in men’s market analysis (see chapter 5.2.4), available market reports consider the U.S. market for deodorants as saturated, with almost 100 % adoption rate of the product amongst U.S. consumers (Euromonitor, 2012a). Based on this evidence, it is more reasonable and in line with market evidence to assume the smaller value of potential market size parameter.

Important analysis in investigating horizontal mergers with differentiated products is the analysis of substitution patterns. If the merging parties are close competitors with high level of substitution between their products, it is likely that they will be able to profitably increase prices post-merger by a significant amount. Table 48 shows the cross-price elasticities between any two brands on the women’s market. Using this information, we can identify the closest substitutes to the merging brands and see how closely are the merging parties competing.

The investigated merger involves P&G, which owns Secret and Sure brands pre-merger, and Gillette owning Dry Idea and Soft&Dri. Cross-price elasticities show that the second closest rival to the Secret brand is Soft&Dri brand owned by Gillette. The same brand is the first closest competitor to Sure. Gillette’s Soft&Dri is therefore exerting a lot of competitive pressure on both P&G’s brands and merger between Gillette and P&G would internalize these competitive constraints.

Table 48 – Cross price elasticities between brands



Brand

Owner

Arrid

Ban

Degree

Dove

Dry Idea

Mennen

Mitchum

Secret

Soft&Dri

Suave

Sure

Arrid

Ch&Dw

-

0.134

0.208

0.162

0.063

0.120

0.059

0.314

0.084

0.065

0.106

Ban

Kao

0.146

-

0.182

0.204

0.102

0.116

0.053

0.296

0.055

0.042

0.068

Degree

Unilever

0.153

0.119

-

0.148

0.057

0.134

0.059

0.352

0.073

0.059

0.094

Dove

Unilever

0.159

0.195

0.195

-

0.100

0.117

0.057

0.306

0.063

0.051

0.076

Dry Idea

Gillette

0.135

0.197

0.168

0.200

-

0.104

0.046

0.278

0.045

0.033

0.062

Mennen

Colgate

0.120

0.104

0.195

0.130

0.050

-

0.059

0.383

0.068

0.050

0.081

Mitchum

Revlon

0.177

0.145

0.251

0.183

0.064

0.180

-

0.426

0.100

0.081

0.102

Secret

P&G

0.128

0.102

0.194

0.128

0.050

0.141

0.054

-

0.066

0.050

0.086

Soft&Dri

Gillette

0.204

0.122

0.242

0.165

0.051

0.160

0.078

0.386

-

0.084

0.116

Suave

Unilever

0.172

0.109

0.227

0.148

0.049

0.143

0.067

0.373

0.087

-

0.105

Sure

P&G

0.160

0.093

0.195

0.122

0.045

0.119

0.050

0.327

0.072

0.055

-

The profitability of a potential price increase is determined not only by cross-price elasticities, but also by the absolute size of demand. For that reason, it is useful to also look at diversion ratios. Table 49 shows the number of units caught by a brand in a particular row if a price of the column brand increases in such a way that it loses 100 units of sales. Using this information, we can see that despite relatively high cross-price elasticities, Soft&Dri is able to retain only very few products lost by Sure or Secret due to a price increase. On the other hand, if either of the two brands produced by Gillette increased in price, most units would be caught P&G’s brand Secret. This indicates that post-merger entity would likely increase prices of Gillette brands more compared to price increase of P&G’s brands.146

Table 49 – Diversion ratios between brands



Brand

Company

Arrid

Ban

Degree

Dove

Dry Idea

Mennen

Mitchum

Secret

Soft&Dri

Suave

Sure

Arrid

Church&Dwight

-100

11

11

11

9

9

12

10

14

15

12

Ban

Kao

7

-100

7

10

10

6

7

6

6

7

5

Degree

Unilever

13

11

-100

12

9

12

13

13

14

16

12

Dove

Unilever

10

12

9

-100

11

7

9

8

9

10

7

Dry Idea

Gillette

3

4

3

4

-100

2

3

3

2

2

2

Mennen

Colgate

7

6

8

7

5

-100

8

9

8

8

7

Mitchum

Revlon

3

2

3

2

2

3

-100

3

3

4

2

Secret

P&G

18

15

20

17

13

20

20

-100

21

22

18

Soft&Dri

Gillette

5

3

4

4

2

4

5

4

-100

7

4

Suave

Unilever

5

3

5

4

3

4

5

4

6

-100

4

Sure

P&G

7

4

6

5

4

6

6

6

7

8

-100

Market Loss

22

29

24

24

32

27

12

34

10

1

27

The analysis of substitution patterns has indicated that brands offered by P&G and Gillette are very close substitutes. A merger between these companies would internalize a significant amount of competitive pressure and would likely lead to anticompetitive effects in the form of a price increase. The size of these anticompetitive effects under different scenarios is now calculated in the following sections.
      1. Inferred marginal costs


Following the structural model described in chapter 3.3, I infer the level of marginal costs for each market product from the model. Values of average marginal costs for each company are displayed in Table 50. Because the demand estimates are based on retail data, the level of reported marginal costs accounts for both wholesale cost of deodorants’ production and retailers’ margin (for more details see section 3.4.5).

Table 50 – Inferred marginal costs by company



Company

Marginal costs

Church & Dwight

18.96 %

Colgate

43.53 %

Gillette

45.77 %

Kao

45.14 %

P&G

28.10 %

Revlon

60.71 %

Unilever

22.77 %

Note: Marginal costs are expressed as % of retail price

Similarly as in the case of men’s market (see section 5.2.5), the level of marginal costs inferred from the model depend on the price and market share of the company (see Equation 16). Companies with lower prices and higher market shares are therefore assumed to have lower marginal costs based on the first order conditions of the model. This is for example the case of the market leader P&G, who has the lowest inferred level of marginal costs. Regarding the relationship between market share and inferred marginal costs, one can think about it in terms of economies of scale, which indeed exist in this market. The larger the scale of production, the lower are the unit costs of production.

It would be desirable to check the level of inferred marginal costs against the level of marginal costs obtained from the market. However, cost information is rarely made public, esp. at such level of detail as necessary in the structural model. Moreover, companies do not use marginal costs as one of the cost standards in their internal systems very often. I am therefore unable to validate the values of inferred marginal costs against market data.

Based on the data on prices, volume of sales and inferred marginal costs, I calculate profit for all market players in the year directly preceding the merger. Results are reported in Table 51.

Table 51 – Estimated profits of companies on women’s market in 2004


Company

Profit ($)

Church & Dwight

9.11 mil

Colgate

6.46 mil

Gillette

5.57 mil

Kao

5.97 mil

P&G

27.97 mil

Revlon

1.71 mil

Unilever

27.22 mil


      1. Unilateral effects


Using estimates of demand parameters and information about the level of marginal costs, I simulate merger effects in various scenarios. Similarly as in the analysis of men’s market, I start by simulating unilateral effects assuming no remedies were imposed in the merger and compare it to simulation results showing the effect of remedies imposed in the FTC’s decision. Lastly, I evaluate alternatives involving different subject to remedies and different acquirers of divested assets.
        1. Counterfactual – No remedies


Firstly, I simulate the effects of an unconditional approval of the P&G/Gillette merger. Results are reported in Table 52. First two columns refer to an average price increase for each company (weighted by pre-merger and post-merger sales respectively) and the third column reports changes in companies’ market shares. As indicated in the analysis of diversion ratios, results show that Gillette would increases its prices significantly (by 13 % on average) after the merger and much smaller price increase is invoked by P&G (only 3.2 % on average). The closest competitor to the merging parties is Unilever, who could profitably increase prices post-merger by 0.42 % solely because the merging parties increased their prices as well. Unilever would also acquire the largest share of the sales lost by P&G and Gillette. Overall, the average market price would increases by 2.20 %, which would decrease the size of the market 1.31 %.

Table 52 – Unilateral merger effects of unconditional merger



Company

Price change 1 (%)

Price change 2 (%)

Market share

change (ppt)

Church & Dwight

0.34

0.34

0.43

Colgate

0.15

0.15

0.32

Gillette

13.12

12.86

-1.47

Kao

0.12

0.12

0.28

P&G

3.17

3.16

-0.54

Revlon

0.02

0.02

0.09

Unilever

0.42

0.42

0.89

Market total

2.20

1.97

-1.31

Table 53 shows the size of unilateral effects broken down by brands. We can see that the two brands produced by Gillette (Soft&Dri and Dry Idea) increase in price substantially (by 8 % and 16 % respectively), brands produced by P&G (Sure and Secret) increase in price only by 4 % and 3 % respectively. This indicates that the P&G/Gillette merger would be detrimental to consumers due to the closeness of substitution between P&G and Gillette deodorants and led to significant price increase.

Table 53 – Unilateral merger effects on brand level



Brand

Company

Price change 1 (%)

Price change 2 (%)

Market share

Change (ppt)

Arrid

Church&Dwight

0.34

0.34

0.43

Ban

Kao

0.12

0.12

0.28

Degree

Unilever

0.40

0.40

0.45

Dove

Unilever

0.36

0.36

0.30

Dry Idea

Gillette

8.85

8.58

-0.57

Mennen

Colgate

0.15

0.15

0.32

Mitchum

Revlon

0.02

0.02

0.09

Secret

P&G

2.95

2.94

-0.41

Soft&Dri

Gillette

16.71

16.70

-0.90

Suave

Unilever

0.61

0.61

0.14

Sure

P&G

3.84

3.83

-0.14

Market total




2.20

1.97

-1.31

Using Equation 11, I calculate merger’s welfare effects (see Table 54). Unconditional approval of the investigated merger would lead to a decrease in consumer surplus by -0.75 % relative to its pre-merger values. Some of that surplus would be captured by increased producers’ profit. This also implies that consumers would face annual increase in expenditures for women’s deodorants of 3.37 million dollars due to higher prices post-merger. Despite the women’s market being larger by one fifth relative to the men’s market, the change in consumer expenditure is about one third of by how much would the unconditional approval of the P&G/Gillette merger change consumers’ expenditures on the men’s market. This shows (along with results reported in Table 23 and Table 53) that merger’s anticompetitive effects were much larger on the men’s market compared to its effect on the women’s market.

Table 54 – Welfare indicators






Change

Consumer surplus

-0.75 %

Producers surplus

1.94 %

Total surplus

-0.53 %

Consumer exp.(annually)

3.37 mil USD

Because the size of unilateral effects obtained from merger simulation is based on the demand estimates characterized by their standard errors, I re-simulate the size of unilateral effects using random draws of demand estimates. I assume demand estimates follow a normal distribution with a mean equal to their point estimate and a standard deviation equal to the parameter’s estimated standard error.147 The following graphs (see Figure 9) show the distribution of the price effects for each brand on the women’s market using 1 000 simulations.

Figure 9 – Distribution of predicted price increase of each brand













Figure 10 – Distribution of predicted market price increase



Based on the performed Monte Carlo simulations, mean price effects and its percentiles (confidence intervals) were calculated (see Table 55). The results confirm high price increases in case of brands produced by Gillette (i.e. Dry Idea and Soft & Dry) and relatively small price effects on Sure and Secret – brands owned by P&G. These results remain valid even when only the lower boundary of 90 % confidence interval is considered. The average overall market price effect of the merger is 2.07 % price increase.



Table 55 – Percentiles and descriptive stats for predicted price effect (in %)













Percentiles

Brand

Company

Mean

SD

10 %

50 %

90 %

Arrid

Church&Dwight

0.29

0.08

0.19

0.28

0.39

Ban

Kao

0.12

0.03

0.08

0.11

0.15

Degree

Unilever

0.37

0.09

0.27

0.36

0.49

Dove

Unilever

0.34

0.08

0.25

0.33

0.44

Dry Idea

Gillette

8.33

1.81

6.22

8.14

10.89

Mennen

Colgate

0.14

0.04

0.10

0.14

0.19

Mitchum

Revlon

0.02

0.01

0.02

0.02

0.03

Secret

P&G

2.76

0.57

2.10

2.67

3.52

Soft&Dri

Gillette

15.41

3.12

11.87

14.91

19.53

Suave

Unilever

0.57

0.13

0.41

0.56

0.75

Sure

P&G

3.72

0.77

2.85

3.60

4.73

Total market




2.07

0.45

1.58

2.00

2.64

Simulation results indicate that approving the P&G/Gillette merger without any remedies would lead to significant anticompetitive effects in form of a price increase. Even though the effects are smaller in size compared to the unilateral effects of the same scenario on men’s market, they remain significant enough to require a regulatory action. It is questionable, why FTC required only an option of divestments and did not require full remedies on women’s market as it did on men’s market.
        1. Efficiencies


Previous results demonstrate the size of the unilateral effects arising from an unconditional approval of the P&G/Gillette merger. Could such effects be fully offset by a decrease in marginal costs induced by the merger? How large would a decrease in marginal costs have to be to fully offset the predicted price increase? Compensating marginal cost reduction was calculated using Equation 18.

Table 56 presents the change in market price, consumer surplus and total surplus caused by the merger assuming different levels of marginal cost efficiencies. In order to fully offset the negative price increase caused by the merger, the merger would have to decrease marginal costs on average by 17.6 %.

Table 56 – Price effects with different marginal cost efficiencies


MC reductions

Market price change (%)

CS change (%)

TS change (%)

0 %

2.20

-0.76

-0.53

5 %

1.69

-0.53

-0.25

10 %

0.70

-0.29

0.04

15 %

0.05

-0.04

0.33

17.6 %

0.00

0.00

0.46

Comparing the obtained level of necessary marginal cost efficiencies with information about the efficiencies claimed by the merging parties (see section 4.3.3.3), it is unlikely that the claimed efficiencies would reach a sufficient magnitude. Most of the publically efficiencies were savings of indirect costs and are unlikely to reach a sufficient level to fully compensate merger’s price effect on a particular market.

Based on the performed analysis, the assessed merger would lead to a significant price increase that would not be countered or deterred by an existing market force (entry, buyer power, efficiencies). Regulatory intervention from the FTC and requiring merging parties to divest some of their assets was therefore a necessary step in order to fulfil the goal of merger regulation policy.


        1. Imposed remedies


The following analysis evaluates the market effect of remedies imposed by the FTC in the P&G/Gillette merger decision. According to the decision, brands formally owned by Gillette (Dry Idea and Soft&Dri) had to be offered to the acquirer of Right Guard deodorants on men’s market. When Henkel became the approved buyer for the assets divested on men’s market, it also acquired Gillette’s assets on women’s market and used this opportunity to enter the women’s market. After the merger, Henkel owned all assets from women’s market previously owned by Gillette, while P&G’s assets remained unchanged.

Given the structural model used in merger simulations, simulation results of this scenario would return zero changes in prices and market shares. There is no change in the ownership matrix and hence the model returns zero merger effects. Such results are intuitive as there is no change in competitive environment on the market. No company internalizes any competitive pressure it was previously facing and no company is facing a newly created competition that it did not face. The set of remedies required by FTC in this case prevented any negative effects the P&G/Gillette merger might have had on the market of women’s deodorants.


        1. Alternative subject of remedies


How would alternative scenarios involving different divested assets or different approved buyers compare to the remedies imposed by the FTC? Which scenario has the smallest effect on consumers’ surplus?

Let’s begin by defining possible alternatives. Firstly, all former Gillette assets could have been bought by an existing player and not by a new entrant. Under this scenario, one of the companies present on the market would acquire both Dry Idea and Soft&Dri brands. Given that P&G’s assets do not change under the simulated scenario, the price effect induced by P&G is only a reaction to the primary price change caused by the acquirer of Dry Idea and Soft&Dri. Consistently with previous methodology, I report merger’s price effect which corresponds to the price effect of assets involved directly in the merger (i.e. P&G’s assets), but also report buyer’s price effect (i.e. average increase in price of assets owned by the acquirer of Dry Idea and Soft&Dri).

Simulation results (presented in Table 57) show that making Church&Dwight or Unilever an approved buyer when divesting all Gillette assets would lead to relatively high negative price effects. Due to these effects, FTC would unlikely approved Church&Dwight or Unilever as possible buyers. This result does not change even if 10 % cost efficiencies are assumed.

Unilateral effects arising from choosing Colgate or Kao as approved buyers are relatively small. Simulation shows that the price of any brand does not increase by more than by 3.5 % and overall market price increases by almost 3 % (assuming no efficiencies). With some level of cost efficiencies assumed, the size of negative effects caused by selling Gillette to an existing buyer diminishes. It is questionable how FTC would assess the situation in which Colgate or Kao would be proposed as potential buyers of Gillette assets. However, the option of selling Gillette to a new entrant and therefore preventing any negative price-effects remains the most efficient scenario.

Due to the very small market share of Revlon on the market before the merger, the size of negative price effects induced from its acquisition of Gillette is negligible. Revlon can therefore be considered as a potential buyer when divesting Gillette. More detailed results for each simulated scenario can be found in Annex II.

Table 57 – Merger’s effects assuming divestment of Dry Idea and Soft&Dri









Buyer of Dry Idea and Soft&Dri

MC reduction




Church & Dwight

Colgate

Kao

Revlon

Unilever

0 %

Merger‘s price effect

0.19

0.13

0.09

0.03

0.40

Buyer’s price effect

4.72

2.81

3.07

1.01

5.41

% Δ in market price

1.09

0.55

0.58

0.13

2.19

% Δ in market size

-0.58

-0.37

-0.41

-0.96

-1.31

% Δ in CS

-0.20

-0.20

-0.21

-0.06

-0.77

5 %

Merger‘s price effect

0.13

0.04

0.03

-0.04

0.30

Buyer’s price effect

3.48

0.85

1.09

-1.26

4.39

% Δ in market price

0.79

0.16

0.20

-0.16

1.77

% Δ in market size

-0.35

-0.07

-0.12

0.15

-0.96

% Δ in CS

-0.08

-0.05

-0.06

0.08

-0.57

10 %

Merger‘s price effect

0.07

-0.06

-0.04

-0.11

0.20

Buyer’s price effect

2.25

-1.11

-0.89

-3.51

3.38

% Δ in market price

0.49

-0.24

-0.18

-0.45

1.35

% Δ in market size

-0.10

0.24

0.18

0.41

-0.61

% Δ in CS

0.47

0.10

0.11

0.23

-0.37

The most efficient scenario represents the possibility of divesting assets from women’s market to a new entrant. This would eliminate any unilateral effects arising from combining divested assets with the assets of an existing market player. However, a new entrant can face certain inefficiencies arising from lack of experience with the market it just entered, imperfect expectations about the behaviour of other market player etc. In order to evaluate this inefficiency, I calculate a maximum potential efficiency loss of a new entrant, which refers to the increase in marginal costs that a new entrant might experience, such that divesting former Gillette assets to a new entrant is still preferable to selling the assets to an existing market player. In other words, if marginal costs of a new entrant would increase by more than the threshold values presented in Table 58, selling Dry Idea and Soft&Dri brands to a particular market player (or any other player with smaller threshold) would have led to smaller unilateral effects.

Table 58 – Efficiency loss for Dry Idea and Soft&Dri relative to each market player


Market player

Efficiency loss

Church&Dwight

14.3 %

Colgate

6.2 %

Kao

5.9 %

Revlon

2.2 %

Unilever

22.1 %

The second possibly in terms of divested assets was divesting only Dry Idea. It could have been bought by either an existing player, or by a new entrant. The most problematic fact with divesting Dry Idea is that P&G acquires Soft&Dri. And before the merger, it is P&G who exerted the strongest competitive pressure on pricing of Soft&Dri deodorants. Regardless of whoever buys Dry Idea deodorants, P&G would increase the price of Soft&Dri deodorants post-merger by approximately 15 %. This is shown in first half of Table 59. Because of internalizing strong competitive pressure between Sure and Soft&Dri, the price of the latter increases significantly. This is therefore a very ineffective set of remedies.

What would happen if only Soft&Dri deodorants were divested and P&G could keep Dry Idea? Under this scenario, P&G internalizes competitive pressure between Dry Idea and its own brands (Sure, Secret), leading to a price increase of Dry Idea post-merger by almost 8 %. Simulation results for a scenario are shown in the second half of Table 59. The resulting price increase is unacceptable by standards of merger regulation and given that there are no market forces that would counter or deter this price effect, it requires regulatory action from the competition authority.



Table 59 - Merger’s effects assuming partial divestment from Gillette







Divesting Dry Idea

Divesting Soft&Dri

Brand

Owner

% Δ in P

Δ in shares

% Δ in P

Δ in shares

Arrid

Church&Dwight

0.25

0.24

0.06

0.14

Ban

Kao

0.03

0.14

0.07

0.11

Degree

Unilever

0.23

0.25

0.13

0.15

Dove

Unilever

0.14

0.16

0.18

0.11

Dry Idea

Gillette

-0.90

0.14

7.84

-0.56

Mennen

Colgate

0.08

0.18

0.05

0.11

Mitchum

Revlon

0.01

0.05

0.01

0.03

Secret

P&G

1.78

-0.26

1.11

-0.21

Soft&Dri

Gillette

15.55

-0.89

-1.03

0.13

Suave

Unilever

0.38

0.08

0.17

0.05

Sure

P&G

2.58

-0.11

1.20

-0.05

Market Total




1.35

-0.73

0.65

-0.45

Overall results show that it was important that P&G acquires neither of the two women’s brands previously own by Gillette. P&G exerts a significant competitive pressure on both Gillette brands and by acquiring either one it internalizes this pressure and could significant increase its prices.

Theoretically, P&G could also keep one of its original brands and acquire one of the brands previously owned by Gillette. Under such scenarios, P&G keeps either Sure or Secret and acquires one of the originally Gillette brands (either Dry Idea or Soft&Dri). However, the transfer of technology, know-how, business strategies specific to the given brand, key employers etc. connected with acquiring any brand of deodorants is quite complex.148 It seems unlikely (and unreasonable) that P&G would be willing to undergo this twice - once when acquiring one brand from Gillette and once when simultaneously divesting one of its own brands. The scenario, in which P&G keeps its original brands and divests both Gillette brands is therefore much more attractive and significantly less complicated than divesting one brand to acquire another. For the low relevance and low likelihood of occurrence of these scenarios, their post-merger market equilibria were not simulated.

Upon the approval of the P&G/Gillette merger, FTC required an option for the divestment of both Gillette brands of women’s deodorants to an acquirer of divested assets on the men’s market. In the end Henkel – a new entrant – acquired divested assets on both men’s and women’s market. The analyses has shown that if the assets owned pre-merger by Gillette were sold to an existing market player (excluding Revlon), significant negative price effects would arise. This would defeat the main goal of imposed structural remedies. Similarly, if P&G was allowed to divest only one of Gillette’s brands and acquire the other, the merger would lead to significant anticompetitive effects. Due to complications connected with assets transfer, it is unlikely that P&G would be willing to sell one of its own brands in order to acquire one of the original Gillette ones. This shows that remedies imposed in the final decision were the most effective in terms of selecting the subject of the remedies (i.e. both Gillette brands were divested) and their approved buyer (both were acquired by a new entrant).

        1. Conclusion


Simulation showed that without any remedies on the women’s market, the P&G/Gillette merger would lead to significant anticompetitive effects in the form of price increase. Prices of Dry Idea and Soft&Dri, brands owned originally by Gillette, would increase by 8.85 % and 16.71 % respectively. P&G would also increase prices of its brands – prices of Sure deodorants would increase by 3.84 % and prices of Secret deodorants by almost 3 %. The average price of the merging parties would increase by 5.2 %. The analyses of countervailing forces (potential entry, significant buyer power, efficiencies) showed that there no market force would offset or deter these anticompetitive price effects. Regulatory intervention in the form of remedies was therefore necessary.

I evaluated the efficiency of the remedies imposed by FTC on the particular market. According to FTC’s decision, merging parties had to provide an option for the acquirer of assets divested on men’s market to acquire also selected assets (Dry Idea, Soft&Dri) on the women’s market. In the end, Henkel bought assets divested on the men’s market and used this opportunity to enter the market for women’s deodorants. Given that there was no change in competitive environment on the market, this set of remedies maintained the level of market competition on pre-merger level.

The divested assets could have been sold not only to a new market entrant, but also to an existing market player. Simulation results indicated that Colgate, Kao or Revlon were other potential buyers who could have acquired Dry Idea and Soft&Dri without the risk of significant negative price effects. However, FTC’s choice of approving Henkel, a new entrant, remains the first-best solution.

Requiring a divestment of other assets was also possible. The merging parties could have divested only Dry Idea deodorants and kept the Soft&Dri brand. However, this scenario would have led to significant anticompetitive merger effects as the price of Soft&Dri would increase by approx. 15 %. Similarly, if the merging parties kept Dry Idea and divested only Soft&Dri, the prices of Dry Idea would increase by almost 8 %. Table 60 compares changes in consumer surplus and annual expenditures that consumers would face under each possible scenario. Results confirm that merger scenarios involving no remedies or a divestment of only one of the brands are detrimental to consumers and therefore not in line with goals of competition policy.

Table 60 – Comparison of relevant scenarios





Δ in P

Δ in CS

Δ cons.expenditure

No remedies

5.17 %

-0.76 %

3.37mil USD

Divesting Dry Idea to a new entrant

3.62 %

-0.37 %

1.86 mil USD

Divesting Soft&Dri to a new entrant

1.83 %

-0.32 %

1.20 mil USD

Selling both to a new entrant

0 %

0 %

0 USD

Note: first column reports change in average price of the merging parties, last change in annual consumers’ expenditure

Based on the results obtained from investigating various relevant scenarios, I conclude that FTC’s decision was the most efficient one in terms of pursuing economic goals of merger regulation.


      1. Robustness analysis


In order to test and verify the validity of the obtained simulation results, I perform several robustness checks. Results for two of them (out-of-sample test for the demand estimation and the IIA assumption test) are reported here, others are incorporated into previous sections (e.g. robustness analysis of demand estimates for changes in instrumental variables and product definition can be found in section 5.3.3).
        1. Out-of-sample test


Similarly as in case of the men’s market (see section 5.2.7.1), I conduct an out-of-sample test for demand estimation. Out-of-sample testing is a standard procedure used for cross-validation of obtained estimation results. In my case, I use the out-of-sample test to verify simulated merger effects of the P&G/Gillette merger assuming no remedies (see section 5.3.6.1). In this scenario, P&G had to sell its Dry Idea and Soft&Dri brands of deodorants to Henkel. Based on the underlying structural model of the market, this scenario does not generate any change in competitive environment of the market, it keeps the first order conditions without unchanged and hence should not lead to any change in prices and market shares.

In September 2006, P&G sold its Sure brand of deodorants to a private investment company Innovative Brands. This altered the competitive environment as Sure was previously owned by the same entity that also owned Secret deodorants. P&G’s divestment led to both brands competing against each other. To exclude potential pricing effects caused by this structural change on the investigated market, perform the out-of-sample test using only data for the first six months of 2006.

Due to the characteristic of ownership changes on the market, analysis performed here is not truly an out-of-sample test as there is no model prediction that would be compared to observed market data. However, I consider this analysis important as it illustrates whether market changes expected by the model correspond to the changes observed on the market – even if the expected change in prices and market shares is zero.

Comparing observed market values with simulation results for the post-merger time period is also important because it serves as a counterfactual to simulating the effects of P&G/Gillette merger assuming no remedies. By comparing results simulating merger effect with remedies and with no remedies, I determine the need for remedies and the efficiency of those remedies (see sections 5.3.6.1 and 5.3.6.3). It is therefore necessary that the results simulating merger effect with remedies correspond to the observed market values.

Results of the out-of-sample test are reported in Table 61. We can see that most observed price changes are relatively small, around 2-3 %. The largest price change observed belongs to the divested assets. The price of Dry Idea increases by 4 % and the price of Soft&Dri deodorants decreases by 5 %. This can be caused by the lack of Henkel’s know-how about the market as there is no change in competitive environment and thus no need to change prices. Henkel was not present on the market before the merger and it is likely that this put him in a short-run information disadvantage.

Table 61 – Out-of-sample test results









Price change (%)

Market share change (ppt)

Brand

Owner

Predicted

Real

Difference

Predicted

Real

Difference

Arrid

Church&Dwight

0

-1.34

-1.34

0

-2.39

-2.39

Ban

Kao

0

1.87

1.87

0

-1.65

-1.65

Degree

Unilever

0

0.69

0.69

0

2.32

2.32

Dove

Unilever

0

2.22

2.22

0

1.15

1.15

Dry Idea

Gillette

0

4.01

4.01

0

-1.31

-1.31

Mennen

Colgate

0

-1.42

-1.42

0

0.04

0.04

Mitchum

Revlon

0

-2.31

-2.31

0

0.33

0.33

Secret

P&G

0

2.01

2.01

0

-0.69

-0.69

Soft&Dri

Gillette

0

-5.04

-5.04

0

-0.12

-0.12

Suave

Unilever

0

-2.21

-2.21

0

1.42

1.42

Sure

P&G

0

2.91

2.91

0

-0.11

-0.11

The reported changes in prices and market shares are not large and are likely caused by short-run fluctuations around the underlying market equilibrium. This is supported by the fact that change in overall average market price is only -0.03 %. Assuming zero change in prices and market shares as the outcome of simulation involving merger remedies seems therefore appropriate.
        1. IIA test


One of the properties implied by the nested-logit model is the restriction of substitution patterns between products belonging to the same nest. These substitution patterns are constrained by the IIA assumption that dictates that the ratio of choice probabilities between any two products remains the same after excluding a third product from the given nest. Based on the IIA assumption inside each nest, cross-price elasticities from one product to all the others are the same (i.e. ).

I follow the methodology of Hausman and McFaden’s test (“HF test”) for testing the IIA assumption in discrete choice models. The intuition behind the HF test is that if the ratio of choice probabilities between any two products should remain the same after excluding a third product (as dictated by the IIA assumption), by re-estimating the original unrestricted demand model with a restricted choice set, parameter estimates should not significantly change. More information on the IIA assumption within the nested logit model and the methodology of the HF test are described in section 5.2.7.2, which performs the same test in the market for men’s deodorants.

Similarly as in the men’s market analysis, I estimate a full and restricted logit model for each nest and calculate p-values for the HF test statistics. When performing the HF test for stick deodorants, I exclude all Degree deodorants which amount to almost 15 % sales within the nest. In the second nest for roll-on deodorants, I exclude Ban deodorants. In both instances is the test statistic too high, making the p-value practically zero and indicating that IIA assumption does not hold within the two nests. I did not perform the HF test for the third nest of spray deodorants as the price parameter α in the unrestricted model had a positive sign indicating increasing utility with increasing prices. Such notion contradicts basic economic theory.

Table 62 – Results of HF test on men’s market



Format

Sales

excluded

Products

excluded

Chisq st.

PValue

Passed IIA

assumption

Stick

14.87 %

13/98

28 930

0

No

Roll-on

26.87 %

3/19

4 403

0

No

Spray

-

-

-

-

-

Results of the HF test show that the IIA assumption of the nested-logit does not hold within the nests on the market for women’s deodorants. This indicates that there are products that consumers consider closer or further substitutes and choice probabilities between any two products are not independent from excluding a third product. However, I was unable to improve the specification of the nested-logit in a way that would account for correlation of consumer preferences within each nest. Several two-level nesting specification were tried, but none led to any meaningful results. One possible solution might be using a full random coefficients model (for example a BLP model) instead of a nested-logit model.

    1. Download 3.14 Mb.

      Share with your friends:
1   2   3   4   5   6   7   8   9   10   11




The database is protected by copyright ©sckool.org 2020
send message

    Main page