L industria del latte alimentare italiana: Comportamenti di consumo e analisi della struttura di mercato

L industria del latte alimentare italiana: Comportamenti di consumo e analisi della struttura di mercato Castellari Elena * Dottorato in Economia e Management Agroalimentare Università Cattolica del Sacro Cuore, Piacenza Scuola estiva per i Dottori e Dottorandi di ricerca Bari, 2-4 Settembre

Framework for the Presentation Research goals Methodology Overview of the Italian milk market Demand specification Estimation and Results Conclusion

1.1) Research Goals First, -Explain : consumer behavior for milk purchasing decisions. - Estimate : how a change in Price shifts consumption between different brands, and milk categories. Second, -Analyze : the relationship between maor players in the market (Granarolo, Parmalat, and PL). -Calculate : Learner Index and profit margins, considering different price strategies.

2.1) Methodology Research Structure I. Estimate demand, using a discrete choice model; II. III. use the estimated coefficients to calculate consumers substitution patterns between different products; assuming a fixed marginal cost, calculate the profit margin for different products, and test for different price strategies and market structure.

2.2) Methodology Demand Model Lancaster (1979): a product is a bundle of characteristics; characteristic space demand estimation, from maximization of the consumer utility function; U i = x ~ β αp + ξ + ε i i ν = + i x k σ k ζ ik ε i k ~ β = β + ik k σ k ζ ik δ = x β αp + ξ

2.3) Methodology Demand Model ),,,, ( ),,,, ( d i k k k d i p x U p x U θ ν ξ θ ν ξ > { } k A i, k i i + > + = ν δ ν δ ν δ / ) ( = ) ( ),, ( ), ),,, ( ( δ σ ν ν ν θ ξ δ A d x f x p x s

2.4) Methodology Nested Logit model Divide the market into G+1 groups (exhaustive and mutually exclusive) Where : u i = δ + ε i δ = X β ap + ξ ε i = ζ ig + ( 1 σ ) v i Utility common to all the products within a group Measure of the correlation of the utility within a nest Extreme-value distribution Type I

2.5) Methodology Nested Logit model s ( e δ ) = σ ( 1 σ Ω ( Ω ) g δ g /( 1 σ g ) ) Where: Ω g = exp( δ k /(1 σ )) k g Normalizing the outside good to zero, we have: S δ S 0 = e Taking the log of both sides provides a linear functional form.

2.6) Methodology Nested Logit model linear specification of the Berry (1994) Nested Logit model: ( s ) ( s ) x β p ( s ) ln ln = α + σ ln + ξ 0 g relaxes the IIA assumption; allows consumer taste to be correlated across a particular group of products;

2.7) Methodology Elasticities Own-Price elasticity: η = α p g ) 1 σ ( 1 σs (1 σ S ) = αp ( 1 S ) Cross-Price elasticities: η k = αpk S α P k(-σs 1 σ g -( 1-σ)S ) if if k g k g

2.8) Methodology Profit margins and Lerner Index Following Nevo (2001), assumptions are: pure Nash-Strategy in prices constant marginal cost Π f = χ f ( p mc ) Ms ( p) C f Where: =1,,J products of each firm f, which belong to set χ f ; M is the size of the market; s (p) is the market share of product ; C f is the fixed cost of production;`

2.9) Methodology Profit margins and Lerner Index Each firm maximizes its profit independently, choosing the price for any product J in the market. FOC: s ( P ) + r χ f ( P r mc r ) s r ( P ) P = 0 S r = s r / p,, r = 1... J Ω * { } r r r * if f : r, χ f r =, otherwise 1, 0 Ω = Ω * S

2.10) Methodology Profit margins and Lerner Index s ( p ) Ω ( p mc ) = 0 Profit margins: p mc = Ω 1 s( p) We can assume different market structures and calculate profit margins by building different matrices * Ω r

3.1) An Overview of the Italian Milk Market Division of the Market Parmalat 16.1% Total milk Market 63% UHT Granarolo 14.5% Private Label 37% Annual per-capita consumption: 57 liters (14.1 gallons) 37% Parmalat 17% Refrigerated Granarolo 33% Private Label 6.6%

3.2) An Overview of the Italian Milk Market Sub-categories Milk Market share Within the group Price % value merchandising UHT Whole 15.08% 0.99 17.37% Semi-skimmed 71.61% 0.76 40.63% skimmed 6.34% 0.93 11.50% Special 6.97% 1.49 16.34% Refrigerated Whole 53.20% 1.45 1.70% Semi-skimmed 33.30% 1.42 1.50% Micro-filtered 9.00% 1.34 3.40% Special 3.80% 1.79 12.30%

3.3) An Overview of the Italian Milk Market Market Structure Shelf life: 120-90 days Price: 1.0 euro/liter Total Milk Market Shelf life: 6-10 days Price: 1.5 euro/liter UHT Refrigerated Whole Milk Semi- Skim Skim Whole Milk Microfiltered Semi- Skim

4.1) Demand Specification Data o IRI (Information Resources InfoScan Data), Food Marketing and Policy Center, U Conn o 48 quarters, from 2004-07 o o o 17 regions 6 kinds of milk x 3 brands = 18 products 13,815 observations Total market = (monthly per-capita consumption, 4.7 liter) X (Regional population)

4.2) Demand Specification Demand Specification ( s ) ( s ) x β p ( s ) ln ln = α + σ ln + ξ 0 g : 18 products S : Share of product P : Price of product g : 2 nests S o : Share of outside good S g : Share inside the nest X volume/unit (packaging size) number of items (each combination of processor-brand-milk-type is one item) regional fixed effect, Label fixed effect, special milk fixed effect

4.3) Demand Specification Descriptive statistics Variables Mean St. Dev Market Share 0.0116 0.0167 Price (euro/liter) 1.0282 0.3387 Volume per unit (liter/unit) 0.9116 0.1635 Conditional market share 0.1181 0.1365 Number of Items 2.3002 1.4007

4.4) Demand Specification Endogeneity and Instruments ( s ) ( s ) x β p ( s ) ln ln = α + σ ln + ξ 0 g Solving the endogeneity problem using an IV approach: Instruments for P : Price of gasoline Instruments for S g (Defining an item as a combination of processor-brand-milk-type): the number of items of a brand s product over the total number of items used to estimate demand inside the nest; the number of items over the total number of items in the market which belong to the same nest;

5.1) Estimation and Results Results The model is estimated using the GMM Generalized Method of Moments. Most of the coefficients, except two regional dummies, are significant and have the expected sign. R-squared:0.5744 Hansen's J-Chi-test: 0.0593 (p = 0.8076)

Variables Coefficient Standard Error P> z Constant -2.9008 0.3515 0.0000 Price -5.8011 0.2308 0.0000 Sg 0.2288 0.0664 0.0010 Volume/unit 1.1471 0.1769 0.0000 Number of Items 0.3230 0.0281 0.0000 Dummy fat content Skimmed -0.1286 0.0730 0.0780 Whole 0.1476 0.0475 0.0020 Brand dummy Granarolo 1.4732 0.0754 0.0000 Parmalat 1.5423 0.0780 0.0000 Characteristic dummy Micro-filtered 0.3924 0.0580 0.0000

5.2) Estimation and Results Consumer Behavior results Milk is a normal good; Consumers consider UHT and Refrigerated milk to be differentiated; Consumers value the national brand more than Private Label; Consumer utility increases with bigger packaging, more variety in the market, for more variety within brand lines;

5.3) Estimation and Results: Own-Price Elasticity All the own-price elasticities are negative; inside any category of product, the PL s elasticity is lower than Parmalat s or Granarolo s; the own-price elasticities within the refrigerated nest are usually greater than the UHT elasticities.

5.4) Estimation and Results: Own-Price Elasticity (contd.) In the UHT nest, consumer fidelity is higher in the semi-skimmed milk category; In the Refrigerated nest, consumer fidelity is higher in the whole milk category.

5.5) Estimation and Results: Cross-Price Elasticity All are positive, meaning that the products are substitutes; nested cross-price elasticities are usually bigger than non-nested cross-price elasticities ;

5.6) Estimation and Results: Cross-Price Elasticity, in UHT UHT nest: The consumers of a national brand are usually fairly reluctant to shift their consumption to a PL product, especially in the whole milk category. Cross-Price elasticities in the skimmed are fairly small; Inside each category, national brand elasticities are higher than PL elasticities.

5.7) Estimation and Results: Cross-Price Elasticity, in Refrigerated Refrigerated nest: The consumers of a national brand prefer to shift their consumption to another national brand rather than to a PL brand; Inside each category, the national brands elasticities are higher than the PL s; Across refrigerated products, shifts between products are fairly homogenous;

5.8) Estimation and Results: Profit margin (PM) assumptions Product-specific PMs calculated, assuming: companies set the price of each product separately; considering the portfolio effect, assuming no firm considers the crossprice effect of products outside its nest; Collusion between National Brands.

5.9) Estimation and Results: Profit Margins, results Milk Whole Semi Skimmed Skimmed Brand Single Product Pricing UHT Portfolio Pricing Fully Collusion by Gr and Par G 0,134 0,232 0,364 PA 0,135 0,187 0,253 PL 0,136 0,173 0,173 G 0,145 0,148 0,162 PA 0,146 0,148 0,161 PL 0,145 0,146 0,146 G 0,133 0,437 0,839 PA 0,134 0,214 0,306 PL 0,135 0,179 0,179

5.10) Estimation and Results: Profit Margins, results Milk Whole Semi Skimmed Micro filtered Brand Single Product Pricing Refrigerated Portfolio Pricing Fully Collusion by Gr and Par G 0,146 0,152 0,169 PA 0,140 0,148 0,184 PL 0,134 0,183 0,183 G 0,139 0,183 0,233 PA 0,138 0,166 0,234 PL 0,133 0,394 0,394 G 0,137 0,244 0,338 PA 0,135 0,189 0,327 PL 0,137 0,142 0,142

6.1) Conclusion: Milk is a differentiated product, where the consumer values different characteristics; UHT and Refrigerated nests are considered to be fairly different from each other, from both the consumers and processors point of view; Prices do influence consumer milkpurchasing decisions;

6.1) Conclusion: The consumer value more a national brand product respect to a PL; We can say that there is a sort of loyalty to national brands ; The market seems to be pretty competitive???(and the portfafolio pricing strategy is the most credible scenario);

References - Nevo Aviv (2001), Measuring Market Power in the Ready To Eat Cereal Industry, Econometrica, 69, 307-342. - Marina Di Giacomo(2008), GMM Estimation of a Structural Demand Model for Yogurt and the Effects of the Introduction of New Brands, Empirical Economics, 34, 537-565. - - Li Tian, Ronald Cotteril (2005), Constrained Price, Address, or Logit Brand Demand Models: an Econometric Comparison in the Boston Fluid Market, Agribusiness, Vol 21(2), 149-166. - - Berry S. (1994), Estimating Discrete-Choice Models of Product Differentiation, Journal of Economics, Vol.25, No.2, pp. 242-262.

Thanks for your attention! Questions??