Conditional valuation for combinations of goods in primates

Abstract

Valuing goods and selecting the one with the highest value forms the basis of adaptive behaviour across species. While it is obvious that the valuation of a given type of goods depends on ownership and availability of that type of goods, the effects of other goods on valuation of the original good are sometimes underappreciated. Yet, goods interact with each other, indicating that the valuation of a given good is conditional on the other goods it is combined with, both in the wild and the laboratory. Here, we introduce conditional valuation in the context of valuing multiple goods and briefly review how human and animal experimentalists can leverage economic tools for the study of interactions among goods. We then review evidence for conditional valuation for combined foods in both human and non-human primates. In the laboratory, non-human primates show increased valuation of certain combinations of foods but decreased valuation of other types of combinations. Thus, similarly to humans, monkeys appear to value combinations of goods in a conditional fashion. Additionally, both humans and monkeys appear to employ similar neural substrates for the valuation of single goods, such as the orbitofrontal cortex. Together, investigations of our evolutionary precursors may provide insights on how we value interacting goods.

This article is part of the theme issue ‘Existence and prevalence of economic behaviours among non-human primates’.

1. Introduction

Humans and non-human animals, including monkeys, rats and starlings, are known to exhibit risk-, delay- and work-sensitive economic preferences implying that common economic principles may be applied across species (see, e.g., [1–3]). For example, monkeys in the laboratory value goods and select the one with the highest subjective value (utility) as suggested by the economic notion of utility maximization [4]. While, for reasons of convenience, laboratory studies often use foods, goods also include tools, activities and services. Subjective value generalizes the notion that the value of two apples is not double the value of a single apple. Instead, subjective value is conditional on a person's wealth level (satiety level for foods; [5,6]), on the framing of the decision problems [7–9] and on the characteristics of the choice set such as the presence and properties of a third option [10–12]. These forms of conditional preferences are also observed in primates [13–16].

Conditional valuation is especially important when goods are not consumed alone but together with other goods (joint consumption). For example, animals eat berries together with leaves and humans may combine bread with butter, coffee with milk, or pasta with tomato sauce. In this sense, conditional valuation is ubiquitous in humans, reflecting the fact that goods are almost never valued independently of each other. Indeed, if they were, consumers would always use their entire budget for the best value-for-money product. Yet, in the supermarket, we do not see shopping carts filled with only oranges, or only carrot juice. Goods are consumed in combination (economists call them ‘bundles’), and valuation is conditional on bundle composition. This fact is so deeply built into modern economics that the idea of ‘value’ makes little sense for individual goods. Rather, one speaks of the ‘utility’ of a bundle of goods, that is, the subjective value of three oranges and a litre of beer compared with the subjective value of five oranges and half a litre of beer.

The ubiquity of economic behaviours reflecting conditional valuation for combinations of foods shown by both non-human primates and humans may imply an evolutionary benefit of those behaviours. One of the most basic assumptions describing human behaviour in economics (see, e.g., [17]) is that preferences are typically ‘convex’, which roughly means that if you are exactly indifferent between two apples and two oranges, you will strictly prefer a bundle containing one apple and one orange to either of the bundles containing only one type of fruit. Economists often speak of a ‘taste for variety’ to provide intuition for this observation. Primates also prefer variety for foods [18,19]. Thus, it seems promising to consider that the interplay between goods and the conditional valuation for combinations of goods can lead us to pursue diversity. Indeed, in the laboratory, tufted capuchin monkeys preferred to spend tokens on diverse rather than on single food types [19]. In the food domain, variety may provide us with a wider range of nutrients and a more balanced nutrition, which may be beneficial for survival. Conversely, recognizing that some goods have similar properties to others (substitution; see below) may increase adaptation to a given environment. Together, conditional valuation for combinations of goods may have evolutionary benefits.

Researchers aiming to understand animal behaviour are only beginning to explore the extent of the interplay between goods and its effects on valuation. The less-is-more effect hints at the possibility that interplay between foods matters for valuation [20–23]. Silberberg and colleagues [23] showed a less-is-more effect where monkeys preferred a single food (for example, banana) to a combination of that same food with an additional food (banana and apple). A possible explanation of this effect is that banana interacts with apple, such that the presence of apple reduces the value of banana. Note that this explanation implies that food type matters and that the less-is-more effect may not generalize to any type of food combinations. The findings from the more systematic and extensive study of Sánchez-Amaro and colleagues [22] supported this idea. In their experiment, the mixed options comprised two of the three foods, pellet, apple and carrot. Primates preferred the combination of pellet and apple when the single option was pellet but were indifferent between single items and the combinations of pellet and carrot or apple and carrot. Thus, the less-is-more-effect did not appear in the combination of pellet and apple. Together, these findings indicate that the food choice behaviour of non-human primates could reflect interplay between goods.

In this paper, we elucidate the approaches used in economics for understanding the interplay between goods, and review relevant studies in both non-human and human primates. These studies, together with indications for similar neural underpinnings for conditional valuation of combinations of goods in the two species, support the existence of evolutionary precursors of conditional valuation in humans. For concreteness, this paper focuses on conditional valuation arising from the interplay between goods and leaves out other interaction effects between goods or choice options (such as decoy effects).

2. Economic approaches for understanding the interplay between goods

Consumption often concerns not only single outcomes or goods but combinations of outcomes or goods. A person in a supermarket with a fixed amount of money must decide what goods, and how much of each good, to purchase. Intuitively, the interplay between goods (the relation between goods) influences valuation and consumption. For example, complementary goods, such as coffee and cake, are more desirable when consumed together than separately. In contrast, substitute goods, such as coffee and black tea, are interchangeable (to some degree). Finally, some goods have no influence on the valuation of other goods. They are called independent goods. For instance, having coffee does not change the satisfaction obtained from an apple.

However, it is not a trivial problem to define the interplay between goods, even though this interplay and how it affects the demand for goods has been the topic of much research in economics (for a historical review, see [24–26]). There are at least two general approaches to capturing the interplay between goods (for detailed discussions, see [27]): (i) Different types of a pair of goods can be classified by looking at the isopreference curves, which are called indifference curves [17]. (ii) When goods have prices, the interplay can be captured with observable demand data, i.e. by the consumption change caused by a price change [17,28]. We briefly describe both of these approaches next.

3. Indifference curve-based approach

Imagine an experiment where decision makers choose between different bundles of two goods, say apples and oranges. For each combination, we can measure which combination is preferred. An indifference curve is simply the locus of all combinations of goods that are equally preferred; hence, it covers the entire space spanned by any combination of the two goods. Curves toward the upper right correspond to stronger preferences (figure 1). The (typical) curvature of the indifference curves depicted in figure 1 reflects a property called ‘convexity of preferences’. Intuitively, this means that the rate of substitution among the goods (how many more units of one good are needed to compensate for the loss of one unit of the other and keep the subject indifferent) is diminishing. The fewer apples we own, the more subjectively valuable apples are to us and hence the less willing we are to trade an apple for an orange. In this example, exchanging two apples for one orange leaves decision makers equally well off (in terms of preferences) if they have only one orange and four apples (from the left dot to the middle dot in figure 1) but not if they have two oranges and two apples (the combination of zero apples and three oranges lies below the black indifference curve).

Figure 1.

An example of two indifference curves. Each indifference curve connects all (equally-desirable) combinations of goods. Thus, the three red dots (one orange and four apples, two oranges and two apples, and four oranges and one apple) represent three equally desirable combinations. Bold blue lines reflect the slopes of the indifference curve and capture diminishing marginal rate of substitution. (Online version in colour.)

Let us now consider indifference curves in the presence of interplay between goods, particularly (perfect) complementarity or substitutability. Perfect complementarity is an extreme case. Left and right shoes of the same style and size are an example of perfectly complementary goods; wearing only one shoe is of no use for individuals with two feet (figure 2a). More substitutable goods have less convex (i.e. flatter) indifference curves. In the other extreme, perfect substitutes are goods that are traded at a constant exchange rate (not necessarily on a one-to-one ratio), regardless of the amounts of goods on offer, and hence indifference curves are flat; black and blue pens are a classic example for perfect substitutes when colour does not matter (figure 2b). If two goods combine unfavourably, the indifference curves could fail to be convex (figure 2c). For example, combined consumption might reduce the subjective value of an olive and the subjective value of a piece of chocolate. Thus, for combinations to reach the same subjective value as separate consumption requires additional olives and chocolates. If one of the goods in the combination is actually a ‘bad’ (more is worse, for example pain), the indifference curve could have a positive slope (figure 2d). The more of this negatively valued good people have to endure, the more of another good they would require as compensation. Accordingly, determining complementarity and substitutability requires careful variation of quantities of goods.

Figure 2.

Indifference curves vary for different types of goods. (a) Example indifference curves for perfect complements (left and right shoes of the same style and size); wearing only one shoe is of no use for individuals with two feet. (b) Example indifference curves for perfect substitutes (blue pens and black pens when colour does not matter); these goods are traded at a constant exchange rate, regardless of the amounts of goods on offer. (c) Example of non-convex indifference curve. Combined consumption reduces the utility of olives and chocolate. Thus, for combinations to reach the same utility as separate consumption requires additional olives and chocolates. (d) Example indifference curve with positive slope. The more of a negatively valued good people have to endure, the more of a positively valued good they require as compensation.

However, it is difficult to quantify the degree of interplay between goods by simply observing an indifference curve. For instance, preferences can (often) be represented by real-valued utility functions, so that an indifference curve in the two-dimensional examples above becomes the set of bundles (x,y) such that u(x,y) = K for a given utility level K, that is, an iso-utility level curve. Unfortunately, the utility function itself is never unique, and the units in which it is measured are meaningless, since any monotonically increasing transformation of a utility function will have the exact same indifference curves. As a consequence, the functional form of a utility function tells us very little about complementarity and substitutability. For instance, the utility function u(x,y) = lnx + lny, defined for strictly positive quantities, might appear to have a separable form, yielding the (incorrect) intuition that goods are somehow independent. After all, the increase in utility arising from one additional unit of one good is independent of the quantity of the other good. Yet, this function generates exactly the same indifference curves as the alternative utility function v(x,y) = xy, and for this function the increase in utility arising from one additional unit of one good very much depends on the quantity of the other good. This is because the degree of substitutability or complementarity is not a statement about utility at all, and ‘utility levels’ are meaningless. Rather, in order to measure the degree of substitutability, one needs to rely on the marginal rate of substitution (MRS), which is essentially the slope of an indifference curve. Given a utility function and an indifference curve u(x,y) = K, assuming differentiability the MRS can be computed through the implicit function theorem by solving y as a function of x and computing the (negative of the) derivative of y with respect to x along the fixed indifference curve, from the implicit equation u(x,y(x)) = K. That is, the MRS is the (absolute value of the) slope of the functional relation determining the necessary trade-off between y and x in order to keep utility constant. Even though it can be computed given any utility function representing the preferences, it is actually independent of the particular utility function used. Of course, except in the case of perfect substitutes, the MRS is itself a function and not a constant. Note that the typical convex shape implies a diminishing MRS along an indifference curve (bold blue lines in figure 1).

Laboratory studies have examined the properties of indifference curves in humans [29–32]. However, we are aware of only two laboratory studies investigating combinations of goods in non-human primates (behaviour: [4]; neural recording: [33]). In the behavioural study, Pastor-Bernier and colleagues [4] observed choices made between options composed of two different liquids with varying quantities by two rhesus monkeys. They tested multiple combinations drawn from different liquids (grape juice, strawberry juice, water, blackcurrant juice, apple squash, lemon juice, liquid yogurt, saline, and combinations with monosodium glutamate and inosine monophosphate) and documented indifference curves with different curvatures. In each trial, the monkeys chose from a reference bundle with fixed quantities of liquid A and liquid B and a variable bundle with a small unit of one of the liquids (0.1 ml) and a varied amount of the other liquid. Changing the amount in variable bundles trial by trial and observing choices allowed the indifference curves to be estimated (i.e. fitted with a Weibull function). The majority of indifference curves was slightly convex and exhibited diminishing marginal rates of substitution. Interestingly, the indifference curve for blackcurrant and grape juice resembled that of perfect substitute goods (similar to figure 2b). The monkeys only cared about the total amount of juice they could get, irrespective of type. In contrast, when monkeys were offered combinations of blackcurrant juice and apple squash, the indifference curves were concave, indicating an unfavourable combination (similar to figure 2c). Combinations of blackcurrant juice and lemon juice resulted in indifference curves with positive slopes (such as figure 2d), indicating that the monkeys required compensation with more blackcurrant juice for consuming lemon juice. This study illustrates that the valuation processes of non-human primates are rather sensitive to the combination of different goods and observed behaviour can be captured well with standard examples of (families of) indifference curves.

However, there are several limitations and challenges when studying conditional valuation for combinations of goods in primates. First, in reality, consumption normally depends on which resources are available. Therefore, it is hard to investigate if primates show preferences for particular combinations over other combinations in the field. Second, even though conditional valuation for combinations of goods is not restricted to foods, previous studies only used foods for investigating combinations of goods in non-human primates primarily because that is convenient for experimental testing. It would be interesting for future studies to investigate the interplay between goods outside the food domain, for example in the context of biological markets in non-human species [34]. In this line of research, it has been shown that grooming could be considered as a form of service and be traded for sexual activity in macaques [35], but the marginal rate of substitution between the two activities remained unclear. Another possible future direction for market-like scenarios might be to study how the consumption of goods is affected by the price of other goods (as we will discuss below).

4. Demand-based approach

The second approach characterizes the interplay between goods at the aggregate level through the changes in demand in response to changes in price [36]. Specifically, complements are goods where the demand for one good decreases as the price of the other one increases $(\mathrm{\Delta }y/\mathrm{\Delta }{p}_x<0).$ For example, when the price of cake increases, people consume not only less cake but also less coffee. Substitutes are goods where the demand for one good increases as the price of the other good increases $(\mathrm{\Delta }y/\mathrm{\Delta }{p}_x>0).$ For example, when the price of coffee increases, people consume not only less coffee but also more black tea. We note that here we assume that everything else remains the same (i.e. no income effects), and more elaborated treatments are needed in more general situations [27].

The notion that food consumption (a proxy of demand) in primates depends on how much effort is required to obtain the food (a proxy of price) can be observed in foraging behaviours in the wild. For example, when the supply of a preferred food is low, primates show substitute consumption. Black spider monkeys eat more leaves and flowers when ripe fruits are scarce [37]. These findings are compatible with the notion that food consumption for primates depends on which resources are available. In the laboratory, Chen and colleagues [38] trained capuchins to use tokens for exchanging foods and tested capuchins' preferences for food combinations before (a token for one piece of apple) and after the price change (a token for two pieces of apple). They showed that two out of three capuchins consumed more apples when the price of apple was relatively cheaper (the price of grapes was relatively more expensive), indicating that apple and grapes are substitutes. Regarding complementarity, black spider monkeys have been observed to feed on 38 different food plants on a single day [37]. Moreover, even though the typical diet of primates is composed largely of plants, they complement it also with some animal matter [39]. This variety could serve nutritional purposes or reflect an evolutionarily-grounded experience favouring foraging for complements: never put all your eggs in one basket.

Together, both non-human primate and human studies show cases of conditional valuation for combinations of goods. However, most previous studies investigated the valuation of combinations of goods with relatively few types of goods and limited quantities. It is important to use more diverse options (e.g. a wider range of foods and activities) and multiple amounts to study how the change of the quantity of one good affects the subjective value of the other good, characterize the slopes of indifference curves and more comprehensively understand the effect of interplay between goods on valuation in the future studies. A possible laboratory study could leverage the token training paradigm further. Potentially, monkeys could be trained to learn the price ratios of a wide range of foods and be tested on how a relative price change affects the relative quantities of food consumed.

5. Neural basis of conditional valuation

Economic theory explains how the combination of different goods can result in valuations that differ from the summed value of the constituents and can be inferred from observable choice behaviour reflected in indifference curves, or changes in demand in response to price changes. However, what is still missing is an understanding of the mechanisms underlying conditional valuations. Given that valuing goods forms the basis of decision-making in both monkeys and humans and must be implemented by the brain, it is likely that studying primates behaviourally and neurally will provide cross-species information. Particularly, the orbitofrontal cortex (OFC) appears to play a critical role in processing subjective value and value-based decisions [40–43]. For example, Papageorgiou and colleagues [21] showed that the less-is-more effect is less evident after OFC lesions in primates. Moreover, satiety effects on taste responses in OFC have been reported in both human and non-human primates (reviewed in [44]). However, only a few studies have investigated the neural representation of combined goods in human and non-human-primates [33,45,46]. In non-human primates, Pastor-Bernier and colleagues [33] demonstrated that the activities of OFC neurons for combinations of juices lying on the same indifference curve are identical, suggesting that indifference curves are statistically related to OFC firing.

6. Conclusion

In summary, it has been shown that not only human but also non-human primates show conditional preferences in various contexts [13–15], including for combinations of foods [4,22,33]. These studies suggest that the fundamental cognitive abilities needed for conditional valuation may be shared between human and non-human primates. To the extent that choice is based on value, it must involve a valuation process where value is assigned to (the representations of) goods. Thus, from an evolutionary perspective, the mechanism of conditional valuation may not be a recent phenomenon tied to the structure of markets in modern society but may have arisen from the repeated experience that the value of combinations of goods differs from that of the summed subjective value of separately consuming the constituents.

Data accessibility

This article has no additional data.

Authors' contributions

All authors wrote the article.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by a Fellowship from the Research Talent Development Fund of the University of Zurich (H.-K.C.) and grant no. 100014_165884 from the Swiss National Science Foundation (P.N.T.)