Abstract
The assessment of the way distributive shocks, such as increased polarization or higher inequality, affect vertically differentiated markets has been severely hampered by the standard reference to uniform distributions. In this paper we offer the first proof of existence of a subgame perfect Nash equilibrium in a vertically differentiated duopoly with uncovered market, for a large set of symmetric and asymmetric distributions of consumers, including, among others, all logconcave distributions. The proof relies on the ‘income share elasticity’ representation of the consumers’ density function. Some illustrative examples are also provided to assess the impact of distributive shocks on market equilibrium.
Similar content being viewed by others
Notes
See also Cremer and Thisse (1991).
In an oligopolistic setting the existence of an equilibrium in prices for given qualities has been proved by Bonnisseau and Lahmandi-Ayed (2007) when consumers’ distribution is concave.
Notice that the possibility of ruling out full market coverage, corner solutions or only one firm being active in the market depends on our assumption on the support of the distribution of \(\theta \), namely that the minimum willingness to pay is 0. We are restating in our general framework a well-known property highlighted in the uniform case by, e.g., Wauthy (1996) and Motta (1993).
It should be noticed that the \(\pi \)-formulation of the density often allows simpler representations of the relevant features of the distribution. For example, \(\pi (\theta )=1\) identifies the modal value of \(\theta \), while “the Pareto, Gamma and Normal density functions correspond to constant, linear and quadratic elasticities, respectively” (Esteban 1986, p. 442).
It is easy to check that Eq. (4) actually states this condition, once the demand for L is written as \(D_{L}=\left( 1-F\left( \theta _{L}\right) \right) -\left( 1-F\left( \theta _{H}\right) \right) \). Notice that \(\eta \left( \theta _{L}\right) \) is the \(p_{L}\)-elasticity of overall market demand—the elasticity of demand with respect to \(\theta \) evaluated at \( \theta _{L}\) multiplied by the (unit) elasticity of \(\theta _{L}\) with respect to \(p_{L}\)—while \(\eta \left( \theta _{H}\right) \varphi _{L}\) is the \(p_{L}\)-elasticity of the demand accruing to H—the elasticity of \( 1-F\left( \theta \right) \) evaluated at \(\theta _{H}\), multiplied by the elasticity of \(\theta _{H}\) with respect to \(p_{L}\).
The distribution F used in this paper obviously refers to the consumers’ willingness to pay: a natural question is then that of the relationship between F and the actual income distribution—a relationship which is of course conditioned by the consumers’ preferences. We are grateful to an anonymous referee for drawing our attention to this point, a short discussion of which is presented at the end of this section.
Second-order stochastic dominance is well known to have noteworthy normative implications in terms of inequality rankings. In particular, equal-mean, second-order stochastic dominance amounts to Lorenz dominance (Atkinson 1970).
The Beta function with parameters (p, q) is given by \(B(p,q)= \int _{0}^{1}u^{p-1}\left( 1-u\right) ^{q-1}du\), while symmetry requires \( p=q=\gamma \); in the text \(\beta \left( \gamma \right) =B\left( \gamma ,\gamma \right) \). On the Beta distribution see Johnson et al. (1995, ch. 25). It is easily seen that \(\gamma =1\) delivers the uniform distribution.
In particular, \(K\left( \delta \right) =\left[ \Gamma \left( 3/\delta \right) \Gamma \left( \left( \delta -1\right) /\delta \right) /\Gamma \left( 2/\delta \right) \right] ^{\delta }\), where \(\Gamma \left( \cdot \right) \) is the Gamma function such that \(\Gamma \left( x\right) =\int _{0}^{\infty }e^{-z}z^{x-1}dz\). On the Dagum distribution and its properties see Kleiber (2008).
This is the formulation suggested by Tirole (1988, p. 97, fn. 1), on the basis of a separable representation of the consumer’s preferences.
This is true for all cases considered in our examples, subject to a finite variance constraint for the Dagum distribution requiring \(\delta \rho >2\). It may be also worth noting that our mean preserving shifts of the distribution of the willingness to pay are associated to lower or higher mean income, depending on the income concavity or convexity of \(\theta \): \( \rho \) smaller (larger) than one would imply a smaller (larger) mean income.
References
An YM (1998) Logconcavity versus logconvexity: a complete characterization. J Econ Theory 80:350–369
Anderson SP, Goeree JK, Ramer R (1997) Location, location, location. J Econ Theory 77:102–127
Atkinson AB (1970) On the measurement of inequality. J Econ Theory 2:244–63
Benassi C, Chirco A, Colombo C (2005) A model of monopolistic competition with personal income dispersion. Metroeconomica 56:305–317
Benassi C, Chirco A, Colombo C (2006) Vertical differentiation and the distribution of income. Bull Econ Res 58:345–367
Benassi C, Chirco A, Scrimitore M (2002) Income concentration and market demand. Oxf Econ Pap 54:584–596
Bertoletti P, Etro F (2017) Monopolistic competition when income matters. Econ J 127:1217–1243
Bonnisseau JM, Lahmandi-Ayed R (2007) Vertical differentiation with non-uniform consumers’ distribution. Int J Econ Theory 3:179–190
Cremer H, Thisse J (1991) Location models of horizontal differentiation: a special case of vertical differentiation models. J Ind Organ 39:383–390
Dixit A, Stiglitz J (1977) Monopolistic competition and optimum product diversity. Am Econ Rev 67:297–308
Esteban J (1986) Income share elasticity and the size of distribution of income. Int Econ Rev 27:439–444
Furth D (2011) Duopoly models with vertical product differentiation. Int Game Theory Rev 13:121–140
Gabszewicz JJ, Thisse JF (1979) Price competition, quality and income disparities. J Econ Theory 20:340–359
Gabszewicz JJ, Thisse JF (1980) Entry (and exit) in a differentiated industry. J Econ Theory 22:327–338
Johnson NL, Kotz S, Balakrishnan N (1995) Continuous univariate distributions, vol 2. Wiley, New York
Kleiber C (2008) A guide to the Dagum distribution. In: Chotikapanich D (ed) Modeling income distributions and Lorenz curves. Springer, Berlin, pp 97–117
Motta M (1993) Endogenous quality choice: price vs. quantity competition. J Ind Econ 41:113–131
Mussa M, Rosen S (1978) Monopoly and product quality. J Econ Theory 18:301–317
Osharin A, Thisse JF, Ushchev P, Verbus V (2014) Monopolistic competition and income dispersion. Econ Lett 122:348–352
Shaked A, Sutton J (1982) Relaxing price competition through product differentiation. Rev Econ Stud XLIX:3–13
Tirole J (1988) The Theory of Industrial Organization. MIT Press, Cambridge
Wauthy X (1996) Quality choice in models of vertical differentiation. J Ind Econ 44:345–353
Yurko AV (2011) How does income inequality affect market outcomes in vertical differentiated markets? Int J Ind Organ 29:493–503
Acknowledgements
The authors thank Luca Lambertini and two anonymous referees for helpful comments and suggestions. The usual disclaimer applies.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A. The second order conditions at the price stage
In this and in the following appendices we shall make use of the following simplifying notation: \(f_{i}\equiv f\left( \theta _{i}\right) \), \( F_{i}\equiv F\left( \theta _{i}\right) \), \(\eta _{i}\equiv \eta \left( \theta _{i}\right) \), \(\pi _{i}\equiv \pi \left( \theta _{i}\right) \), \( i=H,L \).
The SOC of firm H requires:
By using (1) and recalling the definitions of \(\eta \left( \theta \right) \) and \(\pi \left( \theta \right) \) it can be rewritten as
which is indeed the case for all \(\theta _{H}>0\), if Property 3 holds.
The SOC of firm L is satisfied if
i.e.
Equation (3) allows to reformulate \(\Delta /s_{L}\) and \(p_{L}/\Delta \) in terms of \(\theta _{H}\) and \(\theta _{L}\) only:
so that using (5\('\)) and the definitions of \(\eta \left( \theta \right) \) and \( \pi \left( \theta \right) \), the inequality (A1) can be rewritten as
At equilibrium, the first two terms are positive since Eq. (3) ensures \( \eta _{H}<1\) and Property 3 ensures that \(\left( \pi _{H}-1\right) >-2\eta _{H}\). The last term is positive since Property 2 and Eq. (3) imply (a) \(\eta _{L}<\eta _{H}<1\), and (b) \(\pi _{L}>-\eta _{L}>-1\)
Appendix B. The comparative statics of the price stage
By totally differentiating (1) and (2), we obtain the following linear system:
where the above derivatives can be written as:
-
\(\dfrac{\partial ^{2}\Pi _{H}}{\partial p_{H}^{2}}=-2f_{H}\dfrac{1}{ \Delta }-\dfrac{p_{H}}{\Delta ^{2}}f_{H}^{\prime }\); \(\dfrac{\partial ^{2}\Pi _{H}}{\partial p_{H}\partial p_{L}}=f_{H}\dfrac{1}{\Delta }+\dfrac{ p_{H}}{\Delta ^{2}}f_{H}^{\prime }\);
-
\(\dfrac{\partial ^{2}\Pi _{L}}{\partial p_{L}\partial p_{H}}=f_{H} \dfrac{1}{\Delta }-\dfrac{p_{L}}{\Delta ^{2}}f_{H}^{\prime }\);
-
\(\dfrac{\partial ^{2}\Pi _{L}}{\partial p_{L}^{2}}=-\dfrac{2}{\Delta } f_{H}-\dfrac{2}{s_{L}}f_{L}+\dfrac{p_{L}}{\Delta ^{2}}f_{H}^{\prime }-f_{L}^{\prime }\dfrac{\theta _{L}}{s_{L}}\);
-
\(\dfrac{\partial ^{2}\Pi _{H}}{\partial p_{H}\partial \Delta } =f_{H}\left( \dfrac{\theta _{H}}{\Delta }+\dfrac{p_{H}}{\Delta ^{2}}\right) + \dfrac{p_{H}}{\Delta }f_{H}^{\prime }\dfrac{\theta _{H}}{\Delta }\);
-
\(\dfrac{\partial ^{2}\Pi _{H}}{\partial p_{H}\partial s_{L}} =-f_{H}\left( \dfrac{\theta _{H}}{\Delta }+\dfrac{p_{H}}{\Delta ^{2}}\right) -\dfrac{p_{H}}{\Delta }f_{H}^{\prime }\dfrac{\theta _{H}}{\Delta }\);
-
\(\dfrac{\partial ^{2}\Pi _{L}}{\partial p_{L}\partial \Delta } =f_{H}\left( \dfrac{p_{L}}{\Delta ^{2}}-\dfrac{\theta _{H}}{\Delta }\right) + \dfrac{p_{L}}{\Delta }f_{H}^{\prime }\dfrac{\theta _{H}}{\Delta }\);
-
\(\dfrac{\partial ^{2}\Pi _{L}}{\partial p_{L}\partial s_{L}} =f_{H}\left( \dfrac{\theta _{H}}{\Delta }-\dfrac{p_{L}}{\Delta ^{2}}\right) - \dfrac{p_{L}}{\Delta }f_{H}^{\prime }\dfrac{\theta _{H}}{\Delta }+2\dfrac{ \theta _{L}}{s_{L}}f_{L}+\dfrac{\theta _{L}^{2}}{s_{L}}f_{L}^{\prime }\),
and are evaluated at equilibrium.
Denoting with A the coefficients matrix, its determinant can be written as:
Property 2 ensures that \(\left( 1+\pi _{H}\right) >0\), provided that \(\eta _{H}<1\), which is indeed the case at the price stage equilibrium. Since \(\theta _{L}<\theta _{H}\), Property 2 also ensures that at equilibrium \(\eta _{L}<1\) so that \(\left( 1+\pi _{L}\right) >0\). Therefore \(\left| A\right| >0\) if:
Recalling Eq. (A3), at equilibrium we can reformulate this inequality as
which is indeed the case under Property 3.
We can now perform the following comparative statics exercises.
The effect on \(p_{L}^{*}\) of a change in \( s_{H}\)
First notice that for given \(s_{L}\), \(dp_{L}^{*}/ds_{H}=dp_{L}^{*}/d\Delta \). By applying Cramer’s rule:
where the RHS is evaluated at equilibrium, and is positive under Property 2.
The effect on \(p_{H}^{*}\) of a change in \( s_{L}\)
By Cramer’s rule and using the definition of \(\pi \left( \theta \right) \), we have:
where the RHS is evaluated at equilibrium. By using (A2) for \( \Delta /s_{L}\), (A3) for \(p_{L}/\Delta \), \(p_{H}/\Delta =p_{L}/\Delta +\theta _{H}\), the definitions of \(\eta \) for f, the definition of \(\pi \) for \(f^{\prime }\), and then Eq. (5\('\)) for \(\left( 1-F_{L}\right) \), we obtain that at the price equilibrium equation (B1) collapses to:
which under Properties 2 and 3 is unambiguously negative.
The sign of \(\left( \frac{dp_{L}^{*}}{ds_{L}}-\theta _{L}^{*}\right) \)
Consider now \(dp_{L}^{*}/ds_{L}\). By applying Cramer’s rule and using the definition of \(\pi \left( \theta \right) \), we obtain
By using (A2) for \(\Delta /s_{L}\), (A3) for \(p_{L}/\Delta \), the definitions of \(\eta \) for f, the definition of \(\pi \) for \(f^{\prime }\), and then Eq. (5\('\)) for \(\left( 1-F_{L}\right) \), we obtain that at the price equilibrium equation (B3) collapses to:
Since under Property 3 the ratio in (B4) is surely less than one, we get:
where the RHS is evaluated at equilibrium.
The sign of \(\left( \frac{dp_{H}^{*}}{ds_{H}}-\theta _{H}^{*}\right) \)
Consider now \(dp_{H}^{*}/ds_{H}=dp_{H}^{*}/d\Delta \). By applying Cramer’s rule and using the definition of \(\pi \left( \theta \right) \), we obtain:
By using (A2) for \(\Delta /s_{L}\), (A3) for \(p_{L}/\Delta \), the definitions of \(\eta \) for f, the definition of \(\pi \) for \(f^{\prime }\), Eq. (5\('\)) for \(\left( 1-F_{L}\right) \), and noting that \(p_{H}/\Delta =p_{L}/\Delta +\theta _{H}=\theta _{H}/\eta _{H}\), we obtain that at the price equilibrium equation (B6) collapses to:
Since under Property 3 the ratio in (B7) is surely greater than one, we get:
Appendix C. The second order condition at the quality stage
The proof that the Second Order Condition at the quality stage for firm L is satisfied, relies on three preliminary steps.
Step I It is useful to calculate \(\Psi ^{\prime }\left( \theta _{H}\right) \) at \(\Psi \left( \theta _{H}\right) =0\). By deriving Eq. (11) and recalling that \(\theta _{L}=g\left( \theta _{H}\right) \) according to (5\('\)), we get that at \(\Psi \left( \theta _{H}\right) =0\):
Collecting terms the above expression can be simplified into:
Using now the definitions of \(\eta \) and \(\pi \), we can write:
which used into \(\Psi ^{\prime }\left( \theta _{H}\right) \) leads to:
Step II. Notice that at the price equilibrium the effects of changes in \(s_{L}\) on the value of \(\theta _{H}\) and \(\theta _{L}\) can be written as follows:
The sign of \(\theta _{LL}\) can be easily established from (B5). As far as \( \theta _{HL}\) is concerned, we recall from Lemma 2 that for (5\('\)) to be satisfied:
so that, given \(g^{\prime }>0\), \(\theta _{HL}<0\).
Step III For future reference, it is useful to solve the FOC of firm L at the quality stage in terms of \(p_{LL}\). From Eq. (8), this FOC can be written as:
By using (C4) for \(\theta _{LL}\) and (C3) for \(\theta _{HL}\), the above can be reformulated as:
Consider now the ratio at the RHS of the above expression. Recalling Eqs. (5\('\)) and (A3), the expression for \(g^{\prime }\) and the definitions of \(\eta \), we obtain:
so that, using Eq. (10) for \(\left( p_{HL}+\theta _{H}\right) \) we get:
Now we proceed to evaluate the SOC for firm L at the quality stage. The first order derivative of the profits of firm L with respect to quality (Eq. (8) in text) can be written as:
Therefore, given Eq. (2), the second order derivative is:
Using (C2) in the first two terms, and the definition of \(\pi _{L}\) in the last two terms of (C6), we obtain:
which, using (A3), (C4), and the definition of \(\eta \), results into:
Consider now (C1). By multiplying both sides by \(\theta _{HL}\) and rearranging terms, it can be rewritten as:
Using (C1\('\)) into (C7) we obtain:
The first term in (C8) is negative, since \(\theta _{HL}<0\) (see Step II of this appendix) and \(\Psi ^{\prime }>0\) at equilibrium. In the sequel we prove that the sum of the remaining terms is equal to zero at equilibrium. By using repeatedly Eqs. (5\('\)), (10), (A3) as well as the expression for \(g^{\prime }\) and the definitions of \(\eta \) and \(\pi \), tedious algebra shows that this amounts to proving that:
By substituting \(\ p_{HL}+\theta _{H}\) from Eq. (10), collecting terms, and substituting for \(p_{LL}\) from Eq. (C5), this expression can be transformed into:
The term \(\theta _{L}\eta _{L}\left( 2-\eta _{H}\right) /\theta _{H}\left( 1-\eta _{L}\right) \left( 1-\eta _{H}\right) \) can again be substituted from Eq. (10), so that we get:
Now, we deal with the term \(\left( p_{HL}/\theta _{H}\right) +1.\) At equilibrium, given (B2) and (10):
which, substituted into (C9) yields:
which is actually true. Therefore Eq. (C8) collapses to
Rights and permissions
About this article
Cite this article
Benassi, C., Chirco, A. & Colombo, C. Vertical differentiation beyond the uniform distribution. J Econ 126, 221–248 (2019). https://doi.org/10.1007/s00712-018-0631-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00712-018-0631-3
Keywords
- Vertical differentiation
- Duopoly
- Non-uniform distribution
- Subgame perfect equilibrium
- Income share elasticity