Abstract
We apply widely accepted arguments regarding the inferential properties of scientific experiments to analyse markets and consumer decision-making. Market competition is depicted as a means of generating counterfactual data regarding product characteristics resembling experiments’ use of treatment and control conditions to facilitate causal inference. Much like scientific experimentation, competition creates counterfactual data that allows consumers to make causal inferences regarding rival products’ treatment effects upon their welfare. Absent the multiplicity of differentiated goods that competitive markets produce, consumers cannot engage in causal inference and demand curves may become inaccurate specifications of marginal social benefit. This application of the potential outcomes framework to microeconomics offers an account of market competition that has implications for understanding welfare problems created by uncompetitive and monopolistic markets, industrial organisation, and the underlying justifications for anti-trust policy.
Introduction
Neoclassical microeconomics typically treats Pareto efficiency as the evaluative standard for welfare analysis. Assuming consumers are fully informed and perfect competition prevails, the first theorem of welfare economics states that every market clears at the equilibrium price and marginal social benefit for consumers equals marginal cost for producers. Under these assumptions, markets satisfy the conditions of Pareto efficiency and, as a corollary, the first theorem serves as a benchmark to assess departures from welfare efficiency.
This depiction of market efficiency provides the primary normative justification for competition's desirability over uncompetitive or monopolistic conditions. According to standard neoclassical analysis, the deadweight social welfare losses associated with the elevated monopolistic prices resulting from monopolistic firms’ pricing power constitutes the ‘best-known’ inefficiency problem that monopoly creates and informs much of contemporary anti-trust analysis legislation (Tirole, 1988: 65). While some criticise the emphasis on consumer welfare framing the neoclassical depiction of monopoly (Khan, 2017), no alternative theory of monopoly departs from the basic neoclassical framework.
This article offers a new approach to understanding market competition and the problems created by monopoly power. Our principal contribution is to apply widely accepted arguments regarding the inferential properties of scientific experiments, arguments that have become the ‘gold standard’ of empirical analysis, to develop a new theory of economic choice under competitive and uncompetitive conditions (Varian, 2016: 7311). Our justifications are straightforward: the maximisation assumptions of neoclassical analysis depend upon consumers and firms making causal inferences regarding economic choices’ consequences relative to the counterfactual consequences of decisions that were foregone. However, the critical question of how markets, understood as a specific institution for regulating economic competition, facilitate choice maximisation has not been explained (Binmore, 1987: 180; Dosi and Nelson, 1994: 157).
Our approach applies ideas from the potential outcomes framework for causal analysis to help understand the interaction between consumer choice and market competition (Holland, 1986; Imbens and Rubin, 2015; Rubin, 1974). Although typically confined to methodological arguments regarding research design, the potential outcomes framework offers a general approach to understanding causality in all areas of human decision making, including microeconomic choice environments. We argue a novel way of conceptualising market competition can be created by examining how competitive and uncompetitive markets facilitate or restrict economic actors’ capacity for engaging in causal inference, and do so in ways that have important implications for welfare analysis.
Instead of attributing competition's efficiency properties to actors’ desire to engage in maximising behaviour, we argue the potential outcomes framework offers an account of markets’ welfare properties grounded in widely accepted statistical arguments regarding the nature of causal inference. Specifically, we argue consumers’ ability to identify a purchasing decision's causal effect upon their welfare is derived from the counterfactual data generated by competing firms’ differentiated products. However, consumers’ ability to accurately infer a consumption decision's causal effect upon their welfare depends upon the degree of competition that exists in a market, and the corresponding quantity of product counterfactuals consumers can observe and compare.
We recognise that individual firms can produce differentiated products or product lines that offer consumers a range of product counterfactuals. However, we assume that a greater degree of product differentiation is likely to occur under competitive conditions when multiple firms produce product lines and differentiated products. Hence while a single firm may produce, for example, several kinds of cellphones, if multiple firms produce product lines that are differentiated from those rival firms are producing, consumers can observe a greater quantity of product counterfactuals relative to the quantity that would exist under less competitive conditions. We argue that in situations where a greater quantity of differentiated products exist, as is typically found in competitive markets, consumers can more accurately infer how they value product characteristics. While there is an extensive literature on consumer learning, social influence upon preferences, and how market experiences and learning may alter preferences, we argue competition allows consumers to better understand their preference through the causal inferences consumers can make regarding products’ welfare effects (Dold, 2018; Dold and Lewis, 2022).
However, in so far as monopoly power attenuates or eliminates the quantity of counterfactual product data consumers can observe, the demand curves that emerge in monopolistic markets are less accurate approximations of consumers’ marginal benefit relative to the causally identified demand curves that competitive markets generate. Hence in fully undifferentiated monopolised markets consumers’ demand curves are potentially inaccurate measures of purchasing decisions’ marginal benefits because instead of observing their satisfaction with rival firms’ products, consumers can only observe their satisfaction with the single product the monopoly produced.
One advantage of this inference-based account of competition is that it identifies welfare problems created by non-competitive and monopolistic markets that exist independently of firms’ pricing power. Instead of emphasising welfare losses generated by elevated monopolistic prices, we argue that demand curves’ accuracy in representing marginal social benefit (MSB) is endogenous to the degree of competition (and thus differentiation) prevailing in a market. 1 Hence as competition causes the number of differentiated products in a market to increase, consumers’ ability to make causal inferences regarding their valuation of products improves in tandem with the number of product comparisons they can perform. Yet without recognising competition's inferential implications for the accuracy of consumer demand curves, welfare analysis of monopolistic markets is restricted to examining the deadweight social welfare losses caused by elevated monopolistic prices, a focus that may understate the harmful welfare implications of monopoly power. However, this point only becomes apparent after recognising competition's implications for consumers’ capacity for engaging in causal inference regarding products’ welfare implications.
In an effort to explain the implications of markets’ inferential properties, the rest of this article is organised as follows. The second section presents ideas from the potential outcomes framework and explains how this approach to causal inference can be applied to microeconomic decision making. Using ideas from the potential outcomes framework, we argue consumers require competitive markets to engage in causal inference regarding products’ welfare effects, and that the accuracy of consumer demand curves as representations of MSB is endogenous to the number of differentiated products populating a market.
The third section examines how monopoly creates inferential problems with important implications for the accuracy of consumers’ demand curves. We argue a basic concept in the potential outcomes framework that Holland (1986: 297) has called ‘the fundamental problem of causal inference’, a problem describing the difficulty in observing a treatment's causal effect on individual experimental subjects due to their unobservable counterfactual behaviour, is reproduced in economic reality whenever a monopolistic firm excludes competitors from the market. Specifically, the singularity of monopolistic firms’ products makes it difficult for consumers to observe the counterfactual welfare effects of the products that would have been produced under competitive conditions, and the demand curves that emerge from consumers’ purchasing decisions in monopolised markets are correspondingly less accurate representations of MSB relative to those that emerge when competition prevails. Importantly, these welfare problems persist independently of whether monopolistic firms exercise pricing power or not, and hence constitute a fundamental source of monopolistic inefficiency.
The fourth section discusses how our model of consumer behaviour provides a justification for anti-trust policy that does not require analysis of monopolistic firms’ pricing or production behaviour, but which is instead derived from the inferential problems monopoly power creates for consumer demand curves.
We conclude by examining the argument's implications for competitive and non-competitive allocation systems more broadly, and discuss how widely accepted methodological arguments regarding causal inference provide novel insights regarding market and non-market behaviour. If experiments are a unique means of identifying causal relationships, the same logic that accounts for their advantageous methodological properties may also help explain why competitive markets allow consumers to engage in forms of causal inference regarding goods’ welfare implications that are not possible in uncompetitive or monopolistic markets.
Competition and causal inference
The welfare problems monopoly creates are typically studied by examining the production side of monopoly power. Hence the standard depiction of monopolised markets suggests monopolists reduce the quantity of goods they produce to maximise profits, and welfare-reducing deadweight losses are caused by elevated monopolistic prices. The question we consider is whether a new approach to examining monopoly power can identify welfare problems that exist independently of monopolistic firms’ pricing decisions.
It is important to note that our framework does not make reference to perfectly competitive markets where products are undifferentiated and firms lack pricing power. We instead assume markets are distinguished by some degree of monopolistic competition whereby market entry and exit is possible, firms enjoy some amount of pricing power, and products are, to a greater or lesser degree, differentiated on the basis of their characteristics (Chamberlin, 1933). Our argument is thus one of degree, emphasising the nature of the inferences that consumers can make when they can observe greater or lesser degrees of product differentiation, and is less applicable to markets where goods are largely undifferentiated, as is the case for example with energy or many agricultural products. While Dixit and Stiglitz (1977) emphasise the importance of consumers’ taste for the product variety that monopolistic competition provides, our approach argues product variety is also valuable in helping consumers understand and update their own preferences for specific products characteristics, whereas Dixit and Stiglitz hold preferences to be exogenous and fixed and emphasise consumers’ desire for product variety in general. Similarly, we emphasise consumer assessments of products’ differentiated characteristics that are unrelated to the producer's spatial location, as is the focus of those following the classic arguments of Hotelling (1929).
Instead of emphasising the welfare consequences of elevated monopolistic prices, we examine how monopoly creates problems for claiming the accuracy of the demand curve for a product is representative of MSB. We argue monopoly power creates inferential problems for consumers that have not been recognised, and which may cause welfare inefficiencies that persist regardless of whether monopolistic firms exercise pricing power or engage in marginal cost pricing. Before discussing this problem, it is necessary to present our approach, and explain the potential outcomes framework's implications for consumer choice.
The potential outcomes framework is a specific approach to understanding causality in scientific experiments emphasising that there are always more potential outcomes to an experiment than there are observed outcomes, as the act of assigning individuals to experiment's treatment or control conditions prevents observation of individuals’ counterfactual behaviour had they not been assigned to the condition they were assigned to.
For example, if an experiment seeks to measure a new curriculum's effect upon student test scores, individual students can only be exposed to the treatment, the new curriculum, or the control condition, the old curriculum. As students can only be assigned to a single condition, it is impossible to observe the test scores individual students would have earned absent their assignment to either treatment or control. The potential outcomes framework argues this feature of experiments makes it impossible to observe the treatment's (i.e. the new curriculum) causal effect on individual students’ test scores, as one cannot observe the counterfactual test scores the student would have earned had they not been assigned to the new curriculum. The impossibility of observing a treatment's causal effect upon individual experimental subjects has been described by Holland (1986: 947) as ‘the fundamental problem of causal inference’, as it constitutes a general problem confronting efforts to identify causality in scientific experimentation. 2
Experiments overcome this problem by using an untreated control group to reproduce the counterfactual world that would have existed absent assignment to the treatment condition. If treatment and control conditions are randomly assigned, the difference between treatment and control groups reveals a treatment's average causal effect, as the untreated control group overcomes the challenges the fundamental problem of causal inference poses for observing a treatment's effect upon individual subjects.
Although these arguments are largely uncontroversial in the context of econometrics, they articulate an approach to understanding causality that applies to all realms of human decision making. Indeed, the potential outcomes framework has implications for understanding any economic or political choice environment where actors choose among multiple options that are possible at any given moment in time (DeMartino, 2021; Manski, 2004). Just as the fundamental problem of causal inference confronts scientists’ efforts to observe the causal effects of experimental treatments, economic actors confront similar inferential problems whenever they cannot observe the full range of counterfactual decisions they have foregone.
We assume all economic actors confront basic inferential questions whenever they are attempting to make maximising decisions. For example, consumers seek to purchase products that will generate greater satisfaction than the alternative products available for sale, and each product constitutes a potential economic outcome for consumer welfare in a manner that is analogous with the potential outcomes associated with treatment and control conditions of scientific experiments. Similarly, in the context of firms’ production decisions, profit-maximising firms are engaged in making predictions regarding specific production decisions’ causal effects upon their profitability relative to the alternative production decisions they could have chosen.
While many recognise experiments are a unique means of identifying causal relationships in empirical analysis, the inferential problems experiments mitigate have direct analogues in economic decision making. Specifically, while the fundamental problem of causal inference confronts all forms of human action, the challenges this problem poses for the rationality of economic choice is mitigated or exacerbated by the degree of competition that exists in a market. 3 We argue the problems the fundamental problem of causal inference poses for the rationality of economic choice are maximised in contexts where opportunities for observing counterfactual data are restricted or eliminated, as they are in undifferentiated monopolistic or uncompetitive markets, but are mitigated when market competition ensures large numbers of differentiated products are simultaneously available for sale. 4
Competition mitigates the inferential problems confronting economic choice because the multiplicity of differentiated goods populating competitive markets more easily allows consumers to observe multiple rival products’ treatment effects upon their welfare. Although consumers cannot observe the counterfactual welfare they would have experienced had they consumed a good they did not consume, under competitive conditions they can observe information about other products prior to making their purchasing decisions, as the simultaneous existence of rival firms’ products allows consumers to observe differentiated products’ characteristics. Thus, while we recognise that the fundamental problem of causal inference can never be truly overcome, as it always remains unclear whether a consumer would have, for example, preferred living in a house that they did not purchase in an elapsed period of time, competition mitigates this inferential problem by revealing the potential welfare outcomes of consuming rival goods that are typically available under competitive conditions.
Conversely, while competition can never fully overcome the problems posed by unobserved product counterfactuals, the harmful effects of this problem are maximised by the limited forms of counterfactual product data that exist in monopolistic markets with a single product. As consumers’ capacity to make causal inferences about products is contingent upon the existence of competing products which are employed as counterfactual controls, it is only through product comparisons that consumers can assess which they prefer. This argument applies the basic logic of the potential outcomes framework and its emphasis on the inferential problems created by our inability to observe all possible counterfactual outcomes to the context of economic decision making.
We emphasise that we are assuming that a greater quantity of differentiated products populate competitive markets relative to monopolistic markets. This assumption can be contested by the standard assumption that under perfect competition there is no product differentiation, and by the fact that monopolistic firms may produce differentiated product lines that allow consumers to engage in causal inference among them. Hence, for example, Apple produces several types of iPhone, allowing consumers to causally infer which they prefer relative to the counterfactual product alternatives that Apple also produces.
Instead of denying such cases, we suggest that competition merely ensures consumers can observe a greater quantity of product counterfactuals, and that the inferential problems we identify become less concerning in situations where monopolistic firms produce a large variety of differentiated products. Hence while Apple may, for example, sell three different types of iPhones, in the presence of Samsung's competing product line another three cellphones exist which are differentiated from the three iPhones Apple sells. The presence of a greater quantity of product counterfactuals allows consumers to more precisely assess the welfare effects of existing products, as in this example there are twice as many product counterfactuals relative to a situation where Samsung did not exist.
It is important to distinguish our argument from empirical questions regarding whether a greater number of products may create biases or a reluctance to make purchasing decisions based on choice overload (see Chernev et al., 2015). Rather, we emphasise how consumer product assessments are contingent upon products exhibiting differentiated characteristics, as this condition allows consumers to observe a more complete set of counterfactual treatment effects regarding the consequences of rival purchasing decisions.
However the severity of the inferential problems we examine are influenced by the nature of the product in question and the corresponding number of evaluative dimensions that consumers are assessing. Indeed, there may be instances where the product is only evaluated on the basis of a single evaluative dimension, as is the case in goods such as electricity where the nature of the product makes it difficult to differentiate. We suggest that in such cases consumers face relatively less severe inferential problems than in cases where products exhibit a greater number of evaluative dimensions. Hence our argument suggests that competition increases consumers’ opportunities for assessing a larger quantity of product counterfactuals relative to a situation where the absence of competition restricts the amount of product variety that could exist were competition to prevail.
We argue this oversight is not a minor problem; rather appreciating these inferential issues can help generate insight into markets’ welfare properties. By implication, this approach also illustrates certain problems created by monopoly power, but our identification of the problems monopoly power creates is based upon an understanding of how competition mitigates the manifestation of the fundamental problem of causal inference in economic reality, and is not derived from assumptions regarding whether monopolistic firms exercise pricing power in their desire to maximise profits.
Although we assume the multiplicity of product counterfactuals is the underlying condition that allows consumers to engage in causal inference, monopoly power limits the quantity of product counterfactuals consumers can observe. As fully monopolised markets feature a single firm producing a single product, monopoly power prevents consumers from making the product comparisons that are possible under competitive conditions.
While consumers can decide how much they are willing to pay for a monopolist's product, their inability to observe competing firms’ products makes it difficult to know whether they have maximised their welfare. This problem is mitigated by the causally inferred purchasing decisions that are possible when competition ensures consumers can observe rival firms’ products. Absent the inferences consumers can make when competition provides them with product counterfactuals, monopoly power creates inferential problems undermining the assumption that consumer demand curves accurately represent marginal social benefit.
Competition, demand curves and welfare
This section recasts consumer theory and welfare analysis to incorporate insights derived from statistical studies of causal inference. In standard consumer theory, preferences over goods can be presented using indifference curves, the slope of which reveals consumers’ marginal rate of substitution (MRS). The optimal choice of bundle dictates that the MRS of good i for good j is equal to the price of i relative to the price of j. Following from this, demand curves can be derived for each good by identifying the change in quantity demanded given a change in its price holding all other prices and income constant.
A major claim is that given demand curves are derived from indifference curves, the slopes of which measure the MRS, we can argue that uncompensated demand curves (including income effects) are approximately equal to compensated demand curves (supposing income effects are negligible). This then allows for individual demand curves to be viewed as marginal benefit curves and the summation of all individual demand curves for a good to be viewed as MSB. The equivalence of the demand curve for a good with MSB is important in studying the social welfare that is generated by the good in contrasting market structures such as monopoly or perfect competition.
We reconsider these steps, incorporating the idea that consumers engage in causal inference when presented with goods differentiated characteristics (e.g. a cellphone's battery life, screen size, aesthetics, etc.). While we are sympathetic to Lancaster (1966) in arguing that consumers consume goods’ characteristics rather than the goods themselves, we use the existence of differentiated product characteristics as the basis for consumers’ ability to engage in causal inference. Consumers’ ability to sample rival products of the same category, for example, multiple firms’ cellphones, is akin to a within-group experiment where the control is the consumption of a good with no differentiated alternatives and the treatment is the consumption of the same good after the opportunity to experience a differentiated alternative or alternatives of the good (Charness et al., 2012).
The existence of differentiated product alternatives, may in turn, influence how consumers assess the trade-offs inherent to the derivation of the MRS. We suggest that consumers that can observe counterfactual product alternatives will be better able to assess the trade-offs (as measured by MRS) they are willing to make to consume individual products, as their assessments of the MRS will be based upon their comparison of individual products’ characteristics relative to the alternative products available for sale.
This depiction of the consumer choice environment calls into question consumers’ ability to measure products’ welfare trade-offs when the treatments offered by the existence of differentiated alternative products are not observable. Indeed, can consumers’ willingness to give up a certain amount of good y to receive a unit of x be considered a good measurement of marginal benefit, if consumers are unable to make a causal inference regarding the goods’ implications for their welfare relative to other goods that are available for sale as they are in competitive markets? Following from this, if MRS as an accurate measure of MB is open to question due to the way monopoly limits the product data consumers observe, then by extension so is MSB. This would further imply that the use of demand curves for the purpose of welfare measurement for undifferentiated goods becomes questionable absent the product counterfactuals that competitive markets supply and the corresponding way competition allows consumers to engage in forms of causal inference that are possible when they can observe product counterfactuals.
This implies that for competitive environments populated by multiple firms producing differentiated products, consumer demand curves can be more plausibly viewed as representing MSB. This claim is harder to make the fewer the differentiated products available for the consumer to sample and thus use as a treatment. In the limit, a perfectly monopolised market where a single firm produces a single undifferentiated product, the consumer lacks a treatment condition necessary to observe a causal effect. In essence, monopoly power recreates the fundamental problem of causal inference in economic reality, as consumers, like individual experimental subjects, cannot observe the full range of potential outcomes that they could have consumed in a market where competition provided them with a greater quantity of differentiated products.
The welfare implications from this argument are straightforward. In standard analysis, social welfare is compared under monopolised markets and perfect competition. Social welfare loss is calculated as the deadweight loss caused by units of output where MSB is greater than marginal social cost (MSC) not being consumed. Importantly, in comparing monopolistic and perfectly competitive markets, the same demand curve is employed as representative of MSB. Now the use of the same demand curve may be justified under our analysis for the extreme cases of monopoly and perfect competition. In both cases goods are undifferentiated. In monopoly there is only one good and in perfect competition the many goods are homogenous. However, in both cases there has been no opportunity for consumers to engage in causal inference about products’ treatment effects for their welfare, rendering the identity of the demand curve with MSB unclear in ways that may affect the subsequent welfare implications of monopoly power.
If we relax the assumption of perfect competition to allow for competition with significant product differentiation, employing the same demand curve (and MSB) for both monopolistic and competitive markets becomes questionable, as competition facilitates consumers’ ability to engage in causal inference regarding products’ treatment effects in ways that are not possible in monopolised markets. As competitive markets exhibit demand curves that are causally inferred, while in monopolised markets they are not, we have greater confidence in equating consumer demand curves with MSB relative to monopolised markets.
Note that the argument we present here differs from existing theories that would question the equation of demand curves with MSB. We have mentioned the debate regarding the use of uncompensated demand curves as representing compensated demand curves. There seems to be general agreement that this is a reasonable assumption for most goods, since most goods only make up a small proportion of overall spending and thus income effects are small.
More significant challenges come from the paternalist and behaviouralist critiques. In the paternalist critique of welfare analysis for certain types of goods, for example health and education, the consumer's view of the trade-offs for these goods versus other goods should not be trusted and as a result, demand curves derived from consumer preferences will not reflect the consumer's true marginal benefit of consuming merit goods. In this case, paternalistic interventions can be justified so that economic agents are not free to choose to under consume merit goods. The second attack on standard consumer theory comes from behavioural economics. In multiple ways, this literature has demonstrated that preferences may be highly unstable and exhibit intransitivity (Angner, 2012). There is an overlap between behavioural economics and paternalism in the form of nudge theory that is often described as ‘libertarian paternalism’ (Thaler and Sunstein, 2008), yet none of these critiques of standard consumer theory clearly addresses how they are influenced by markets’ degree of competitiveness.
The analysis developed here is distinguished by its focus on the inferential implications of market competition, and by its identification of a fundamental issue that exists independently of the concerns raised by paternalism or behavioural economics; specifically, we question any approach that simply assumes consumer demand curves are accurate depictions of MSB independently of the degree of competition that exists in a market.
Hence we do not assume there is a better judge of preferences than the consumer (although there might be) and while we agree with behavioural economists that the consumer preferences may display intransitivity, the degree of market competition is not the source of intransitivity in behavioural economics. We argue competitive markets provide a quasi-experimental choice environment that produces counterfactual data consumers can use to assess goods’ treatment effects upon their preferences. Importantly, we argue the more competitive the market, the more accurately consumers can infer their satisfaction from consuming individual goods. This property is not a consequence of consumers desire to maximise their preferences, as the assumption that consumers pursue preference maximisation is held constant regardless of the degree of competition that exists across markets. Rather, we argue competitive markets allow consumers to engage in causal inference, allowing them to identify products’ treatment effects upon their welfare.
We assume that there exists a ‘true’ well-being function that maps consumption goods into welfare, but consumers are uncertain about its contours. Instead of suggesting that consumers possess exogenous preferences that allow them to understand how goods’ characteristics influence their welfare, we argue that it is only the presence of differentiated products that allows such uncertainty to be reduced or eliminated. Our discussion below compares the decisions consumers make in the presence of markets with and without product differentiation, and examine the welfare effects of decision based upon consumers’ opportunities for engaging in causal inference.
We now provide a more formal discussion of these arguments and consider the standard model of consumer utility maximisation whereby an individual is maximising a utility function containing three goods, the first,
Now suppose a hypothetical experiment whereby all consumers can be exposed to a differentiated version of good x,
The logic presented here has implications for understanding the nature of the welfare losses monopoly imposes. First, let us consider a competitive market economy where all goods available for consumption are differentiated. According to our argument's emphasis on consumers’ capacity to causally infer products’ treatment effects for their welfare, under competitive conditions consumers’ MRS provides relatively more precise measurement of consumer satisfaction than they would under uncompetitive conditions. This is because the slopes of consumers’ indifference curves reflect causally inferred judgements of rival goods’ treatment effects upon consumer welfare, and as a result, consumers’ demand curves more accurately reflect each good's MSB.
Our depiction of the original problem, in which the utility function contained an undifferentiated product, is assumed to be more likely to occur in an uncompetitive market due to the lack of competing firms and consumers’ corresponding inability to make counterfactual comparisons of rival products’ welfare implications. In effect
Proponents of rational choice approaches to consumer theory may respond by stating that a preference relation can be represented by a utility function only if it is rational, that is complete and transitive (Mas-Collell et al., 1995: 9). This implies that the welfare consequences of all product bundles are known prior to entering a market. This view of consumer behaviour assumes all product comparisons have already taken place prior to consumers’ actual exposure to products. Clearly this is not intended to describe reality and such a strong assumption can be defended on an ‘as if’ foundation (e.g. Friedman, 1953). 5
It is important to emphasise that we do not critique this view of consumer decision making for its lack of realism. Rather, our critique is that the maximisation assumption fails to explain how actors make maximising decisions, and ignores how consumers’ capacity to make maximising decisions is itself a function of the degree of competition that exists in a given market. Furthermore, simply assuming actors can successfully maximise their preferences ignores the role of causal inference in human decision making, an assumption that has become increasingly questionable given the centrality of causality in scientific analysis and the concern with causal inference currently emphasised by econometrics and the social sciences more generally (Angrist and Pischke, 2008, 2014; Heckman, 2005). Indeed, once one accepts that experimentation is a valuable tool for scientific inquiry, there is little reason to ignore the implications of causal questions for the analysis of individual choice, both in the construction of theoretical models and in real-world economic decision making.
Note however that emphasising the role of causal inference in microeconomic decision making may have varied implications based upon the nature of market structure. For example, one implication of our argument is that the rational choice depiction of preferences is more defensible the more competitive a market becomes because consumers require actual (as opposed to imaginary) exposure to differentiated products to be able to infer their satisfaction with products and calculate trade-offs. However, we do not assume preferences are exogenous and fixed, but rather their accuracy is dependent upon the inferences that are possible in the presence of multiple product counterfactuals.
Hence in cases where exposure to differentiated products is limited, the extreme case being an undifferentiated monopolised market, indifference curves will less accurately measure consumer satisfaction, and the claim that demand curves for monopolistically supplied goods are accurate representations of MSB is on weaker foundations than for competitively supplied goods. This indicates that the validity of the rational choice description of consumer preferences and demand curves is placed on a much sounder scientific footing by recognising how causal inference informs preferences that result from product experimentation. Maximising outcomes are thus a consequence of consumers’ ability to engage in causal inference regarding the differentiated products competitive markets produce, and are not merely a result of consumer preferences that exist independently of the degree of competition a market exhibits.
Although the equation of demand with MSB for uncompetitive goods is not innocent, it is defensible for competitively supplied goods assuming no behaviouralist or paternalist concerns as previously discussed. Our argument merely suggests this assumption overlooks a key property of the market itself, namely, competition is a means of allowing consumers to observe products’ treatment effects, and the market's efficiency properties emerge from the forms of causal identification that are possible when competition offers consumers a large quantity of product counterfactuals to observe and experience.
The argument can be summarised using a modified version of the potential outcomes notation provided in Rubin (2005). In Figure 1 below, consumers are listed as ‘units’ from 1 to N. The treatment Y(1) is the derivation of the demand curve for individual i for good
Consumer choice under potential outcomes.
In the standard rational choice approach
Anti-trust implications
Our approach to analysing consumer choice under potential outcomes has implications for understanding the underlying justification for anti-trust regulation. We have argued that when competitive markets are populated by increasingly differentiated products, consumer demand curves become more accurate depictions of MSB relative to the demand curves in non-competitive settings. This property of the market is derived from the forms of counterfactual data competition generates, and consumers’ corresponding capacity to engage in forms of causal inference that are not possible absent the simultaneous existence of multiple firms’ differentiated products.
Competition's beneficial inferential properties are attenuated, or eliminated entirely, absent the existence of competitive conditions. In the limit, under monopolistic conditions where one firm produces a single product, monopoly power maximises the inferential problems the fundamental problem of causal inference poses for rational economic action, as consumers cannot engage in the forms of causal inference that are possible under competitive conditions. This inferential problem has implications for the demand curves that emerge in monopolistic markets, as the demand curve for a monopolistically supplied good is a less accurate depiction of MSB than when the same good is competitively supplied. 6
While one implication of this problem involves the difficulty in accurately measuring the size of the deadweight loss monopoly imposes, a more striking implication involves monopoly's welfare implications even if a monopolistic firm does not exercise pricing power and instead employs marginal cost pricing. Although the traditional neoclassical framework states marginal cost pricing implies static efficiency and elimination of harmful social welfare costs, if a monopoly prevents consumers from making causal inference about products’ welfare effects, consumers’ demand curves in monopolistic markets may inaccurately reflect MSB relative to the causally identified, and hence more accurate, demand curves that would exist were the good competitively supplied. This insight is overlooked if one retains the simple focus on monopolistic pricing power that is the principal insight of the neoclassical framework.
Hence even if monopolistic firms engage in marginal cost pricing, the point at which the demand curve cuts the marginal cost curve may be inefficient due to consumers’ inability to observe their true MSB. As this problem exists independently of monopolistic firms’ pricing decisions, social welfare losses may be incurred as a consequence of the inferential problems monopoly creates regarding the welfare consequences of unobserved production decisions that consumers would observe if competition prevailed.
Note that in the textbook approach to anti-trust (e.g. Motta, 2004; Vascusi et al., 2018), the debate over anti-trust policy, such as the Microsoft case and Google recently, is less focused on concerns regarding mark-up pricing and instead emphasises the potential for destroying competition and innovation. While Khan (2017) makes a similar argument regarding Amazon, we share the view that solely focusing on elevated monopolistic prices is too narrow, but unlike Khan we do not discard concerns with consumer welfare in the analysis of monopoly power. Rather, our approach indicates that an overlooked justification for product variety is derived from the inferential value of the product counterfactuals that prevail under competitive conditions, and the increased accuracy of the demand curves that emerge from causally identified consumer purchasing decisions that such counterfactuals make possible.
The primary anti-trust implication of our analysis is that even if monopolistic firms engage in marginal cost pricing, either doing so on their own initiative to discourage potential competitors from market entry or through effective public regulation, the absence of competition may still impose welfare losses due to the inferential problems monopoly power creates for the accuracy of consumer demand curves as representative of MSB. This indicates that anti-trust policy should not only focus on pricing considerations, but should also examine whether there are opportunities to enhance consumers’ ability to observe differentiated products, as is normally possible in competitive markets.
However, recognising the nature of this problem requires understanding competitive markets’ inferential properties, the depiction of consumer decision making (summarised in Figure 1) derived from our application of the potential outcomes framework, and the way deviation from competitive conditions undermines our confidence in the accuracy of consumer demand curves as reflective of MSB. Absent understanding of competitive market's inferential properties, neoclassical consumer preference assumptions overlook the specific problem our approach identifies, and in doing so, ignore a basic source of inefficiency that monopoly power creates.
Conclusion
The causal inference revolution in econometrics has generated a powerful set of arguments regarding the beneficial properties of scientific experimentation. The econometric techniques introduced by the ‘credibility revolution’ emphasise the centrality of causality in scientific inquiry, and have been widely adopted across the social sciences. However, the insights from such arguments are not limited to research design and econometrics; all economic decision-makers confront inferential questions whenever attempting to make maximising decisions. While many recognise that assessing public policy decisions involves questions regarding different policy interventions’ causal effects, what has not been recognised is how market choice environments offer opportunities for mitigating various inferential problems depending upon the degree of competition that prevails (Heckman et al., 1997; LaLonde, 1986; Manski, 1993, 2019).
This article has illustrated how applying ideas from the potential outcomes framework helps clarify how consumer decision-making confronts the causal issues central to scientific experimentation, and how applying ideas regarding causal inference help illustrate a new way of thinking about competition's beneficial attributes, and by extension, the problems monopoly power creates. There are several advantages of this approach, as it identifies a number of problems that the standard theory of monopoly overlooks.
Instead of emphasising the welfare losses imposed by monopolistic pricing power, the inferential problems created by unobservable product counterfactuals in monopolistic markets indicate consumers’ demand curves in monopolised markets cannot be treated as identical to the demand curves that emerge under competitive conditions. For if a monopoly prevents consumers from performing the product comparisons that would be possible were competition to prevail, consumer demand curves may be inaccurate representations of MSB. Note that this indicates that even if regulatory oversight compelled monopolistic firms to engage in marginal cost pricing, consumers might still suffer welfare reductions due to their inability to observe the welfare implications of differentiated products that would exist absent monopoly power.
Although some recommend rejecting the centrality of consumer welfare in the analysis of monopoly (e.g. Khan, 2017), our application of the potential outcomes framework offers a new way of thinking about the problems monopoly creates without abandoning the emphasis on consumer welfare. This shift in emphasis may seem minor, especially since we reach similar conclusions regarding competition and market efficiency as the neoclassical approach. However, we diverge from the neoclassical framework in our explanation for how markets reach efficient outcomes. The neoclassical account is derived from maximising assumptions that are axiomatic and treated as basic properties of consumers that exist independently of market conditions; our potential outcomes approach offers an explanation for how markets facilitate forms of causal inference that are attenuated or eliminated under non-competitive conditions.
As methodological issues involving causal inference have been widely adopted by econometricians, we see few grounds for ignoring these statistical arguments’ implications for microeconomic behaviour as well (Manski, 1993). Indeed, there may be fruitful parallels between seeing competition as a means of revealing information about potential economic outcomes that parallel the emphasis placed upon casual understanding as a fundamental characteristic of scientific inquiry (Salmon, 1984). In so far as competition prevails, the market's efficiency may be derived from competition's capacity for adjudicating between the accuracy of different economic actors’ hypotheses regarding the most efficient actions in a manner paralleling the superior inferential properties that scientific experimentation exhibits over rival forms of causal understanding.
By applying ideas regarding the inferential properties of scientific experiments to microeconomic decision making, this article has argued competitive markets reveal data regarding counterfactual economic outcomes in a manner analogous to scientific experiments. A key property of competitive markets lies in the greater quantity of potential economic outcomes revealed when a large number of firms populate markets relative to the quantity that are revealed in uncompetitive or monopolistic markets. The inferences consumers can draw in competitive choice environments exhibit, and the resulting demand curves that emerge in competitive conditions, may offer normative justification for competition's desirability. However, this justification is not a consequence of the properties of consumer preference functions, but is due to the way competition facilitates forms of inference that are not possible under uncompetitive conditions. This indicates that understanding market competition from the perspective of the potential outcomes framework has several additional implications.
First, emphasising how competitive markets may be understood as a means of facilitating economic actors’ capacity to engage in causal inference regarding products’ welfare effects may have novel implications for understanding the forms of knowledge communicated by the market price mechanism (Boettke, 2002; Hayek, 1945). Instead of suggesting markets aggregate decentralised or tacit knowledge, markets are depicted here as a means of facilitating a type of judgment, causal inferences regarding products’ treatment effects, an argument that can be applied to explain consumers’ and firms’ capacity for understanding the consequences of consumption and production decisions upon welfare and profits.
Second, our argument may also have implications for industrial organisation. For example, the inferential problems we have identified may be attenuated if monopolistic or oligopolistic firms produce multiple product lines that allow consumers to compare different versions of the products the monopolist produces (e.g. Apple's multiple categories of iPhones), and it may remain an empirical question as to whether competitive markets produce a greater quantity of counterfactual data than uncompetitive markets. This point may have important anti-trust implications, as it emphasises the importance of certain market characteristics, namely the quantity and nature of the counterfactual data that consumers can observe regarding differentiated products, that are not associated with firms’ pricing power.
For example, if Google were to monopolise control over the search engine market, it would be impossible to determine consumers’ demand for an alternative search engine that might have characteristics that were different from Google's. In the presence of competition, where Google might, for example, face a rival firm's AI driven search engine, consumers might infer that they prefer the non-Google AI search engine, and in doing so, revise their assessment of Google as well. However, before market experimentation occurs there is simple uncertainty regarding the search engine characteristics consumers prefer, and this uncertainty is only resolved once consumers experiment and compare products against each other. The key feature we emphasise is the quantity of counterfactuals consumers can observe, and the way consumer demand curves are altered by the inferences they can make in the presence of the greater quantity of products competitive settings exhibit.
To summarise, this article argues market competition may be understood using ideas from the potential outcomes framework, and has illustrated how this approach offers a novel understanding of consumer demand curves under monopolistic conditions. The utility of this approach lies in its application of widely accepted statistical arguments regarding causal inference to microeconomic decision making, the novel hypotheses it offers regarding demand curves and the inefficiencies created by monopoly power, and the potential for extending this approach's insights to other domains, such as public good provision. Indeed, in so far as modern states are monopolistic producers of regulatory decisions and public goods for their societies, the inferential problems caused by unobserved counterfactuals applie to political choice environments as well (DeCanio, 2021). Furthermore, although we have focused on causal inference and consumer satisfaction with products, our argument is robust to the ends actors pursue, and is also applicable to situations where actors are concerned with ends other than welfare, such as different varieties of freedom (Sen, 1991, 1993).
Although issues surrounding causal inference are widely recognised as fundamental for empirical analysis, these insights involve basic properties of causality that have implications for understanding a range of decision environments, economic and otherwise. By focusing on the counterfactual data populating different competitive realms, and the way different choice environments offer agents different opportunities to observe decisions’ causal effects, ideas from the potential outcomes framework may be fruitfully applied to understand various decision environments in ways that are currently not recognised by existing approaches.
Footnotes
Acknowledgements
The authors thank Stephen DeCanio, Marco Giani, Shaun Hargreaves Heap, Maia King, Pavel Kuchar, Bouke Klein Teeselink, Gabriel Leon and Alberto Vesperoni for helpful comments and suggestions. A previous version of this paper was presented to the Politics, Philosophy, and Economics research group at King's College London and the European Public Choice Society meeting in Vienna in 2024. Special thanks to three anonymous reviewers and David Wiens for helpful comments and suggestions.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The funding support was from the John Templeton Foundation's grant “The Political Economy of Knowledge and Ignorance” (grant number 61823).
