I. INTRODUCTION
The COVID-19 pandemic has profoundly changed how we, as human beings, live our lives.Footnote 1 As infection rates around the world continue to surge,Footnote 2 and worrisome variants of the virus emerge,Footnote 3 policymakers struggle to adopt effective policies to curb the progress of the virus.Footnote 4 The measures taken around the world have been diverse, including lockdowns, curfews, travel restrictions, mask mandates, social distancing rules, shutdowns of non-essential businesses, movement tracing, and quarantines.Footnote 5 While some of these strategies seem to be working, at least in some areas,Footnote 6 many have failed,Footnote 7 resulting in major restrictions of freedom and interruptions of economic activitiesFootnote 8 but only minor benefits in slowing the virus’ progression. Notwithstanding the scientific breakthroughs of COVID-19 vaccines,Footnote 9 the (in)efficacy of some of the above mentioned strategies remains a focal point of public debate.Footnote 10
Existing studies on COVID-19 mitigation strategiesFootnote 11 mostly focus on figuring out which individual policies are effective in specific countries or specific spheres of everyday life. For instance, one study found that a lockdown policy was highly effective in China’s city of Wuhan (the virus’ “ground-zero”),Footnote 12 whereas another study found that voluntary social distancing in China was largely ineffective.Footnote 13 In the Netherlands, an “intelligent lockdown” (combining stay-at-home and social-distancing measures) was found to be effective, but only for some people (e.g., those for whom it was feasible to follow the measures).Footnote 14 In Israel, self-quarantine policies were found to be effective, but the efficacy was much greater when individuals were compensated for loss of income during their quarantine.Footnote 15 Another study conducted in Ireland found that a policy calling for social distancing using posters is more likely to be effective when emphasizing how social distancing prevents harm to others.Footnote 16
Such studies provide a glimpse into individual policies in isolation but are unlikely to reveal the full picture—compliance may well depend on more intricate factors than the details of a specific policy. For this reason, other studies have taken a different approach and instead examined how individual beliefs and demographics affect compliance with COVID-19 mitigation policies. For instance, compliance with mitigation policies in the United States was found to mainly depend on the capacity to obey the rules, the opportunity to break the rules, self-control, and intrinsic motivation.Footnote 17 Another (cross-country) study found that the main determinants of compliance with (voluntary) requirements are the beliefs in the efficacy of policies and concerns for one’s own health.Footnote 18 Yet, another study found that the only relevant individual factor affecting compliance was the person’s subjective fear of contracting COVID-19.Footnote 19 A different study of young adults in Switzerland found that social-distancing measures yielded more compliance than hygienic requirements.Footnote 20
These types of studies shed a bit more light on individual behavior but are still somewhat narrowly focused because they implicitly approach compliance as a static concept, where each policy is evaluated independently of others. In other words, many of the existing studies implicitly treat compliance as a result of what economists call a “partial equilibrium:”Footnote 21 an analysis that focuses on one limited set of variables to examine how a given market behaves in a vacuum, that is, without accounting for any interaction effects between markets.Footnote 22
To illustrate, suppose that employers desperately need to hire new employees due to some economic shock.Footnote 23 The direct effect of this shock will be an increase in wages because employers will be willing to offer employees more money for their work. The adjusted wage and number of employed people will constitute a “partial equilibrium.” However, consider what might happen next: as employers pay more money to their employees, the cost of production for goods that those employees produce would increase. Then, producers would ask for a higher price from consumers in the product market (to cover their uptick in costs). Consumers might not be willing to pay a higher price, so they would respond by buying less. Producers would see that consumers are not buying and cut back on production, thereby reducing the demand for workers, which would (at least partially) countervail the initial effect. The final outcome, which also takes into account the feedback loops due to the interaction between markets, is what is known as a “general equilibrium.”Footnote 24
Translating this into the context of COVID-19 requires viewing mitigation policies as the “shock” and the final outcome of changes in behavior as an “equilibrium.”Footnote 25 Much like in the example, one might look at either a partial equilibrium—asking what is the direct consequence of a given policy—or a general equilibrium—asking what is the final outcome after taking into account spillovers between different “markets” (or, in the case of COVID-19, the result of different behaviors in response to different policies).Footnote 26
Such spillovers may take two forms depending on whether the two behaviors are substitutes or complements.Footnote 27 Behaviors that are substitutes serve a similar need and can be mutually exclusive—for instance, going to the cinema and going to the theater are both entertainment activities, but a person seeking to be entertained can, at any given moment, choose only one of those activities.Footnote 28 An important characteristic of substitutes is the inverse relationship between the cost of one and the demand for the other. If a theater is relocated to a further location (i.e., the cost of going to the theater increases), then people are more likely to switch to the cinema, leading to a higher demand for cinema tickets. Conversely, behaviors that are complements provide higher value when they are combined (simultaneously or sequentially). For instance, going to a restaurant and going to the cinema might allow for a “dinner-and-a-movie” combination, providing higher value when jointly consumed.Footnote 29 Complements, respectively, have a direct relationship between the cost of one and the demand for the other: if a restaurant raises its prices, people may opt-out of going out at all, leading to a lower demand for cinema tickets.
Similar to behavior, policies can also be classified as “strategic substitutes” and “strategic complements.”Footnote 30 Policies are strategic substitutes if they offset one another but are strategic complements when they reinforce one another. In the context of COVID-19, the chief goal of mitigation policies is to slow down (or potentially halt) the progress of the virus. Hence, if the combination of two policies yields a synergistic effect,Footnote 31 then they are strategic complements, but if the two policies interfere with one another, then they are strategic substitutes.
In this context, it is important to differentiate between the relationship of two behaviors and the relationship of two COVID-19 mitigation strategies. For instance, hand washing with (i) soap and (ii) water are two complementary behaviors. Policies supplying free water and free soap to the public are then strategic complements, as they incentivize the two complementary behaviors. Conversely, going to (i) restaurants and (ii) cafés are two substitutable behaviors (as one can only go to a café or a restaurant at any given time), whereas two policies that eliminate both of these options (e.g., through a shut-down) are still strategic complements. Eliminating both of these options has a synergistic effect, making it more likely that infections will not spread through indoor gathering.
With this framework in mind, the question we are interested in is which types of COVID-19-related behaviors and mitigation strategies are substitutes, which are complements, and which are simply unrelated.Footnote 32 The reason why this distinction is important is straightforward: any COVID-19 policy is an intervention that attempts to influence behavior, usually through incentives. Thus, to determine what happens if policymakers change the cost (or benefit) of a certain option by adopting a COVID-19 mitigation policy, one needs to have a good grasp of the relevant alternatives that individuals face. Yet, the set of relevant alternatives may well depend on which other concurrent policies are introduced, an aspect that might also differ from one country to another.Footnote 33 As a result, implementing individual policies without considering the connection between behaviors is often bound to fail.
In the following pages, we investigate how COVID-19 mitigation strategies might interact with one another and other (possibly country-specific) factors. This will allow us to explain why some strategies fail due to their neglect of final outcomes in a general equilibrium.
It is important to emphasize that COVID-19 mitigation strategies may also fail for a variety of reasonsFootnote 34 other than the interaction effects we discuss. However, identifying the additional channels that we propose should help policymakers make better decisions, taking into account any and all possible effects that may hinder the effectiveness of a given mitigation strategy.
The rest of the Article is organized as follows: Part II discusses how a general equilibrium approach, mostly focusing on substitution effects, provides a better framework for creating and evaluating effective COVID-19 mitigation strategies. Part III provides several examples of interaction effects among COVID-19 mitigation strategies and other factors, such as social and environmental factors. We discuss our insights and offer some advice for policymaking in Part V. Thereafter, Part VI concludes.
II. COVID-19 MITIGATION STRATEGIES: A GENERAL EQUI LIBRIUM APPROACH
A. What is the Difference between a Partial and a General Equilibrium?
To explain what one should look out for in the context of COVID-19 mitigation strategies, let us first explain in more detail what a general equilibrium is and how it differs from the effects that existing studies typically measure (i.e., those of a partial equilibrium).
In neoclassical (micro-)economic theory, the price of goods is determined by an intersection of “demand”—reflecting consumers’ willingness to pay for the good—and “supply”—reflecting the suppliers’ willingness to (produce and) sell the good.Footnote 35 When supply equals demand, the market reaches an equilibrium: a point where a certain quantity of the good sold for a certain price is acceptable by both consumers and suppliers.Footnote 36
When the market suffers a shock (on either the demand or the supply side), the equilibrium price and quantity may change. There are many reasons why shocks occur, but for our purposes, we restrict attention to the introduction of a new legal policy as the shock of interest. For instance, suppose that consumers are willing to buy 10 widgets for the price of $1 per widget. Then, the government declares a shutdown of widget factories, causing a delay in production. As fewer widgets are being produced, they become scarcer, and supply decreases. If consumers still want to buy a widget, they would now need to pay more. This effect thus translates into a new “partial equilibrium” with a new price and a new quantity. The equilibrium is only “partial” because it is the result of a simple analysis that strictly focuses on the market for widgets, neglecting any and all side effects.
However, side effects are typically present in markets in the form of spillovers to (or from) other related markets. These may include, for example, people switching to a close substitute (i.e., any other good satisfying a similar need as a widget), people reducing their purchases of a complementary product (i.e., any other good that is consumed in combination with a widget), or people reducing their shopping of unrelated products because of income effects.Footnote 37 Additionally, if there are other (possibly unrelated) events or policies that occur at the same time and affect related markets, they might also indirectly shift the equilibrium price and quantity in the market for widgets. Hence, if one accounts for the universe of all possible effects, one can attain a better prediction—a “general equilibrium.”
Scholars of law and economics have extended the idea of a general equilibrium into the analysis of non-market choices.Footnote 38 Namely, instead of considering what happens in the market for a particular good, one needs to assume that different behaviors are associated with a “price” and a “quantity.” Subsequently, individuals conduct a cost-benefit analysis of the different choices in order to choose their profit-maximizing option.Footnote 39 For example, in the decision to drive a car, the “price” might be the cost of gas and the “quantity” might be the frequency of driving.Footnote 40 Intuitively, the choice of how much to drive is also subject to possible demand shocks ( e.g., when the government imposes a tax on gas) and supply shocks (e.g., if the government closes a main road). Also in this context, one can seek out the partial equilibrium or the general equilibrium; for example, a tax on gas leads to less driving in a partial equilibrium (as driving becomes more costly, the demand for driving decreases), but if the tax on gas also affects the production of some other activities associated with driving (e.g., car-parts manufacturing) there are additional side effects.Footnote 41 Following a similar line of thought, we proceed to analyze COVID-19 mitigation strategies.
B. What is the Justification for a Legal Intervention in a General Equilibrium?
From an economic perspective, a general equilibrium can be a desirable phenomenon, as the market forces can bring about an outcome that is not only stable but also efficient. Footnote 42 Namely, if markets are competitive and frictionless,Footnote 43 then the (general) equilibrium outcome is efficientFootnote 44 so there is no clear economic need for legal intervention.Footnote 45 This result reflects Adam Smith’s “invisible hand”—even though people (rationally) maximize their own self-interest, the result is socially beneficial.Footnote 46
Under this prism, COVID-19 mitigation strategies make (economic) sense insofar as the general equilibrium is inefficient, which only happens if some frictions are present. Yet, it is fairly clear that frictions, in the form of negative externalities,Footnote 47 are a strong driver of behavior during a pandemic because (insufficient) precautions to avoid infections have a strong influence on others.Footnote 48 As an example, the quantity of facial masksFootnote 49 sold in equilibrium may be inefficiently low because either sellers selfishly charge an overly high price from consumers (while disregarding the social cost)Footnote 50 or consumers selfishly underutilize (and therefore demand too few) masks. Note that the problem arises through two channels: (i) masks prevent their wearer from infecting others, and (ii) masks reduce the chance that their wearer gets infected and hence impose costs on others (either by infecting them or by taking up scarce medical resources).Footnote 51 Such behavior leads to a “market failure”—people behave rationally and the outcome is, nonetheless, inefficient.Footnote 52 Framed differently, individuals may refuse to bear a personal cost to enhance health by reducing the spread of COVID-19, given that health is a “public good”Footnote 53—a good that everyone can enjoy at the same time (“non-rivalrous”) and that no one can easily exclude others from consuming (“non-excludable”).Footnote 54 Public goods typically suffer from undersupply because people who produce (or consume) them have an incentive to free ride on other people’s effort—here, by not incurring the inconvenience of wearing a facial mask.Footnote 55 Such a market failureFootnote 56 serves as the economic justification for governmental intervention in the form of mitigation strategies.
C. How can COVID-19 Mitigation Strategies Change the Equilibrium?
In the context of COVID-19, policymakers can attempt to influence human behavior using public policies that create shocks to targeted behaviors. Consider a COVID-19 mitigation policy stating that “no person shall be allowed to stay at another person’s private residence.”Footnote 57 The apparent logic of such a policy is that the prohibition will decrease the frequency of face-to-face meetings using a threat of sanctions, mostly in the form of a monetary fine. Whether such a policy will indeed lead to the desired switch can be conveniently captured by standard models of crime deterrence in the spirit of the canonical model of Nobel Prize Laureate Gary Becker.Footnote 58 In such models, rational individuals make choices based on their anticipation of the costs and benefits associated with each option. Thus, they calculate whether the net benefit from some action “A” is higher than the net benefit from the alternative action “B.” Whichever option is expected to yield higher utility is then chosen. In the simplest case, where sitting idly (action “A”) yields a zero utility and violating the law (action “B”) yields a benefit,Footnote 59 individuals violate (i.e., commit a crime) if and only if
$$ b-p\hskip0.5em \ast \hskip0.5em f>0, $$
where b is the benefit from crime (from action “B”), p is the probability of being caught and punished, and f is the size of the penalty (e.g., a monetary fine).Footnote 60
As a more general case, individuals will prefer action B to action A if
$$ E\left({b}_B\right)-E\left({c}_B\right)>E\left({b}_A\right)-E\left({c}_A\right), $$
where
$ {b}_i $
is the benefit from action
$ i\in \left\{A,B\right\} $
,
$ {c}_i $
is the cost of action i, and
$ E\left(\cdot \right) $
denotes an “expected” cost or benefit (there is usually some uncertainty about whether the benefit will be attained or the cost will be incurred).Footnote
61
Relating this inequality to the example above, suppose that action “A” is violating a rule (e.g., by visiting a friend at his private residence), whereas action “B” is staying alone at home. A person considering what to do will need to calculate the costs and benefits of each option: the benefit of visiting a friend, the chance of being caught, the severity of the penalty, the level of boredom from remaining home, the chances of being infected at home as compared to at a friend’s house, and so on. Then, if the net benefit of visiting a friend is higher (lower), the individual will (not) violate the rule.
The key question, then, is what determines the net costs and benefits of each option. One straightforward component is the expected penalty:Footnote 62 individuals should care about the probability of being caught when visiting their friend and the magnitude of the fine that they might incur. For simplicity, we assume that individuals are risk-neutral, i.e., they only maximize their expected utility.Footnote 63 The expected penalty depends also on how the policy is enforced, such as whether the police often enter private homes and whether fines are indeed issued by police officers.Footnote 64 Thus, it is tempting to restrict attention to the expected penalty, which policymakers can easily influence,Footnote 65 and assume that if the cost of a certain action is high enough, it will simply not be chosen.
Yet, this is precisely the problem of over-focusing on a partial equilibrium: the expected penalty affects the absolute net benefit of the targeted action but does not reveal what happens to the relative net benefit. In particular, the implicit assumption here is that action B is benign, that is, it entails no (or a negligible) social harm. Concretely, the policy implicitly assumes that a person who abstains from visiting a friend will instead choose to stay home, but whether this is true depends on (i) whether the policy also indirectly changes the benefit from staying at home and (ii) whether feasible alternatives to staying at home that are less benign exist.Footnote 66
If policymakers were only interested in finding out whether the policy reduces the frequency of visiting friends, these two issues would not matter. However, as the goal of the policy is reducing COVID-19 infections (and not reducing visits per se), what policymakers should care about is the general equilibrium—what happens to the number of infections and deaths. With this in mind, we proceed to consider which types of interaction effects come into play.
III. INTERACTION EFFECTS: SUBSTITUTION AND COMPLEMENTARITY
In this Part, we consider the role of “interaction effects:” effects that cause heterogeneity in behavior (and hence in the efficacy of a policy) depending on some varying factors.Footnote 67 In the context at hand, we consider two kinds of interaction effects: substitution effects and complementarity effects. As mentioned above,Footnote 68 two goods are substitutes if they serve a similar need and are complements if they generate a higher benefit when they are jointly consumed. Similarly, two behaviors may also be either substitutes or complements.
This distinction, which lies at the heart of this Article, determines what happens when a COVID-19 mitigation policy is adopted. If the policy increases the cost of a certain activity, then people will switch from complements to substitutes.Footnote 69 To illustrate the importance of this point, we consider two concrete examples in the next Sections.
A. COVID-19 Mitigation Strategies: Substitution Effects
Let us return to the example of a policy prohibiting staying at other people’s private residences. Suppose that while visiting a friend’s residence is prohibited, visiting a friend’s office is allowed. In this case, it seems plausible that a meeting will still take place, but the meeting’s location will simply switch from the private residence to the office, rendering the policy largely ineffective.Footnote 70 This is a straightforward example of a substitution effect among behaviors—a policy discouraging one harmful behavior might just lead people to switch to another (possibly no less) harmful behavior.Footnote 71
The possibility of substitution effects reflects a common criticism of the Becker model, known as the “marginal deterrence argument:”Footnote 72 changes in the penalty of one crime can cause people to switch to another, possibly worse, crime (or to a higher degree of the same type of crime).Footnote 73 A similar insight also arises in criminology’s “crime displacement theory,”Footnote 74 which suggests that police enforcement efforts in one area may simply cause offenders to switch the location of their criminal activity to a nearby location with less enforcement.Footnote 75
Conversely, suppose instead that it is also prohibited to meet people at the office (and that this is strictly enforced). This would eliminate the substitution effect and may also have a complementary effect, as a prohibition to go to the office can reduce the use of public transportation and incentivize people to stay at home as the viable alternative.
To further illustrate, consider the following numerical example: an individual values the net benefit from staying at home at “2” and the net benefit from visiting a friend at “5.” This individual would clearly choose to visit a friend (which is a costly activity to society in times of COVID-19) unless there is some threat of a fine. Then, suppose the government decides that visiting a friend at home is subject to an expected monetary fine of “4,” anticipating that the individual would prefer staying at home. With these numbers, as summarized in Table 1 below, the policy would work—but this is only because we are looking at the policy in isolation (i.e., at the partial equilibrium).
Table 1. Illustration of a simple choice between two options
Note: This table presents an illustration of payoffs with and without legal intervention (a monetary fine). The numbers represent the utility of an individual, who then chooses the column yielding the highest utility in the row that applies (depending on whether there is intervention). The utility-maximizing option in each row is marked using bold text.
Next, suppose that the individual has the alternative option to visit the friend at the office, yielding a net benefit of “3” (i.e., it is a close substitute for visiting a friend at home). We would then instead get Table 2, showing that, with an intervention, the individual would simply choose to visit the office (rather than stay at home).
Table 2. Illustration of choice between three options
Note: This table extends Table 1 by adding a third column capturing the possibility of visiting a friend at the office. The utility-maximizing option in each row is again marked using bold text.
To avoid the adverse consequences of this substitution effect, the government must adopt a complementary policy of punishing visits to the office. For instance, if a visit to the office entails a punishment of “2,” then we get Table 3, showing the individual chooses the desired, benign behavior.
Table 3. Illustration of choice: complementary policies
Note: This table describes the payoffs of the same choice detailed in Table 2 but under the assumption that there is also a fine for visiting a friend at the office. The utility-maximizing option in each row is again marked using bold text.
Note that the two policies—prohibiting visits to a residence and prohibiting visits to an office—are indeed strategic complements: their interaction creates a synergy that achieves the desired outcome. In fact, in this example, neither policy is effective on its own—if one would only prohibit office visits, this would have no effect (as the meetings at a residence would continue as usual).
B. COVID-19 Mitigation Strategies: Complementarity Effects
As a counter-example, consider the policy of alcohol bans, which was adopted in various forms and in multiple countries during the COVID-19 pandemic.Footnote 76 At first glance, the logical connection between alcohol and COVID-19 is unclear because these policies do not directly reduce infections. A first rationale might be that such a ban reduces the likelihood that some intoxicated individuals would carelessly get too close to others and thereby hinder the ability to maintain social distancing. However, a second explanation might be related to complementarity effects: if people consume alcohol mainly in conjunction with social gatherings, then the ban would also indirectly induce people to stay at home.
Using a numerical example as before, suppose that an individual gets a utility of “5” when he can drink alcohol with his friends, “3” when he meets his friends but does not drink, and “4” when he stays at home. A policy that punishes alcohol consumption with a fine of “2” then yields Table 4:
Table 4. Illustration of complementarity effect
Note: This table presents an illustration of payoffs with and without legal intervention (a monetary fine), for the example of bans on alcohol consumption. As in all previous tables, the numbers represent the utility of an individual, who then chooses the column yielding the highest utility in the row that applies (depending on whether there is intervention). The utility-maximizing option in each row is marked using bold text.
In this example, a fine for consuming alcohol, by itself, is sufficient for causing individuals to switch to the benign activity of staying at home. However, with a slight adjustment, this may not work, either because meeting friends without alcohol would bring a high benefitFootnote 77 or because people would substitute alcohol for another product. For instance, some individuals might simply consume drugs (e.g., marijuana) while gathering with friends instead of drinking alcohol,Footnote 78 rendering the policy ineffective due to a substitution effect.
Notably, the two policies of banning alcohol and prohibiting gatherings are not necessarily strategic complements. For instance, suppose we add another column of “consuming alcohol outdoors alone,” referring to an activity that is still costly from a social perspective (e.g., because intoxication may cause the person to get close to strangers) but not as costly as drinking with friends. Then, slightly changing the numbers from the previous example, Table 5 below illustrates a scenario that combines complementarity and substitution effects.
Table 5. Illustration of effects with costs vs. benefits
Note: This table presents an illustration of payoffs with and without legal intervention (a monetary fine), given four mutually exclusive choices. The first row specifies the social cost. The other rows specify, as in all previous tables, the utility of an individual. The individual again chooses the column yielding the highest utility in the row that applies (depending on whether there is intervention). The utility-maximizing option in each row is marked using bold text.
In this scenario, the government can impose a fine for gatherings, a fine for consuming alcohol, or both. The worst possible outcome is meeting friends and drinking alcohol—yielding a social cost of “20,” followed by the (somewhat) less problematic activity of drinking alone (social cost of “17”), gatherings without alcohol (social cost of “13”), and staying at home as a benign activity. As before, the individual chooses the option with the highest private benefit. Here, if the government adopts only an alcohol ban, the individual still meets friends (without drinking). Conversely, if the government prohibits only gatherings, then the individual prefers drinking alone outside. This means that both policies have some positive impact—and in that sense, if the goal is simply to eliminate the most dangerous activity, then the policies are strategic substitutes. Yet, as the last row of the table demonstrates, when the policies are combined, the individual chooses the optimal action (staying at home). Thus, if the goal is to reduce the social cost as much as possible, the policies are strategic complements.
As another brief example, consider the mitigation strategy of international travel bans, adopted in many countries at different times throughout the pandemic.Footnote 79 If one was exclusively interested in the effect of this policy on the spread of COVID-19 during flights (a partial equilibrium), there might be ways to alleviate this concern through rigid testing and proper ventilation. However, travel bans seem more justifiable from a general equilibrium perspective because of activities that are complements to flights (e.g., dining in different restaurants while on holiday abroad) and may bear a high risk of infection. Thus, by banning flights, one also gets the indirect benefit of reducing the frequency of complementary activities.Footnote 80 At the same time, when unable to travel internationally, individuals may respond by turning to the closest substitute—a domestic vacation (with domestic dining), partially countervailing the benefit.
These examples highlight the challenge that policymakers face: they must account for multiple substitution and complementarity effects.Footnote 81 To evaluate policy options, Neal Katyal has proposed that policymakers conceptually differentiate between substitution and complementarity effects in crime deterrence that yield either increases or decreases in the demand for both the punishable activity and other activities.Footnote 82 This model can be mapped onto most of the effects identified aboveFootnote 83 but includes some additional measurements. For instance, a prohibition of an activity might generally affect preferences, (e.g., cause people to dislike the prohibited activity). For COVID-19 mitigation strategies, however, this seems less relevant—the prohibitions do not generally have a declaratory moral aspect; rather, the goal is usually instrumental (to temporarily reduce the frequency of the activity to prevent infections).
There are additional challenges that arise because the various effects may be heterogeneous, that is, in some areas, the effects will be strong while in others they will be weak (or non-existent). To account for this, one must consider how COVID-19 mitigation policies interact with additional factors. As the range of relevant factors may be vast, we consider only a few examples in the next Part (without claiming to be exhaustive in any way). These examples will help illustrate the type of interactions that policymakers may want to consider.
IV. FURTHER APPLICATIONS
The prohibition of staying at another person’s residence and the ban on selling alcoholFootnote 84 provide straightforward case studies of interactions in terms of either substitution or complementarity effects. However, given that interaction effects are a function of the elusive, immediate environment, they are likely to be multidimensional. In this Part, we broaden the discussion by focusing on three salient examples, some of which are also supported by existing empirical evidence.Footnote 85 In so doing, we shed light on how policymakers can fine-tune effective COVID-19 mitigation strategies in a cost-efficient manner, namely by taking into account important phenomena such as social normsFootnote 86 and environmental factors.
A. Restricting Freedom of Movement in Light of Social Norms: Family Ties
Traditional microeconomic theory assumes that individuals behave as “homo oeconomicus”—homogenous economic units that act selfishly and only strive to maximize their own utility.Footnote 87 This is precisely why the aforementioned public-goods problem emerges: people do not care about infecting others. However, the validity of the homo oeconomicus assumption has been challenged by a long line of experimental findings suggesting that people have “social preferences,”Footnote 88 that is, they also care about what happens to others.Footnote 89 If social preferences were the dominant factor driving behavior, we might not need much intervention at all (as people would take precautions to avoid infecting strangers). However, there may be specific types of social preferences in play that affect behavior but not with respect to strangers. For instance, a well-known theory in social sciences distinguishes between several values, one of which is “benevolence,” or the preservation and enhancement of people with whom one is in frequent contact (e.g., family members).Footnote 90 This differs from “altruism” or “pro-sociality” in that the individual cares about a specific subset of in-group people, rather than human beings in general.Footnote 91 However, the degree to which social preferences translate into behavioral changes may depend on social norms,Footnote 92 which can be defined as informal rules of everyday life that reflect a shared view of what is considered appropriate behavior.Footnote 93 For instance, whether one is expected to visit one’s family is often a social norm that differs starkly across countries. An implication of this social norm is that policymakers should be more concerned about violations of social distancing among the family members in some countries but less so in others, and COVID-19 mitigation strategies should be tailored accordingly. However, social norms need not be detrimental to mitigation efforts. Policymakers could try to leverage social norms and tailor their COVID-19 mitigation strategies to create better incentives.
Consider the said norm of visiting (or, more generally, taking care of) one’s family.Footnote 94 Social support within the family, such as by paying weekly visits to one’s parents, becomes even more significant in times of social isolation due to COVID-19. On the one hand, this may then just lead to the obvious effect—people would be more likely to visit their family, even in violation of social-distancing rules. On the other hand, the opposite may hold: younger generations may have an inherent incentive to protect older generations, such as parents and grandparents (so-called “groups at risk”), from contracting the virus, which then provides a strong incentive to avoid visiting close family members.Footnote 95 A third, more interesting option, however, is a mix of the two: the young may continue to visit their family, but instead take extra care when meeting others, thereby mitigating the risk that they themselves get infected prior to the visit. If this practice occurs, then the possibility of visiting one’s family enhances the incentive to comply with social-distancing measures in everyday life.
Given this, what would then be the effect of a policy that restricts family visits? If properly enforced, the outcome of this policy should be fewer family visits, which can reduce the transmissions within the family. At the same time, such a restriction would crowd out the intrinsic motivation to socially-distance:Footnote 96 if young people cannot meet their family at all, they no longer have to worry about infecting family members.Footnote 97 Such reasoning would likely apply to restrictions of freedom of movement that prevent a person from going out of one’s residence beyond a small perimeter (e.g., 1,000 meters);Footnote 98 if not all family members live less than 1,000 meters apart from each other, the restriction may crowd out a young person’s incentive to comply with social-distancing measures because the person cannot visit—and therefore does not need to protect—their family from contracting COVID-19. That person’s friends (or neighbors), on the other hand, might live nearby, especially in dense cities. Looking at the numerical example specified in Table 6, one can see that visiting one’s family (thereby violating the 1,000 meters restriction) provides higher levels of utility as compared to visiting a friend who lives less than 1,000 meters away. Namely, in this example, we assume that individuals have an intrinsic utility of “3” from keeping their family safe. What happens, then, is that an intervention imposing a monetary fine for violating the 1,000 meters perimeter (to visit one’s family) has two simultaneous effects: (i) it reduces the benefit from visiting the family (via the threat of a fine) and (ii) it increases the benefit from visiting one’s friend by “3” (the intrinsic motivation). The result reflects the same kind of substitution effects that we discussed in detail above—switching from family visits to friend visits—but due to the simultaneous effects, the result is not a switch to the benign activity but to another harmful activity.
Table 6. Illustration of choices in light of strong family ties
Note: This table presents an illustration of payoffs with and without legal intervention (a monetary fine) for the example with a social norm. The individual chooses the column yielding the highest utility in the row that applies (depending on whether there is intervention). The utility-maximizing option in each row is marked using bold text.
As a result, not only is the policy possibly ineffective, but also harmful—it has other side effects: by restricting family visits, the elderly may suffer psychological hardship due to social isolation without a clear reduction in infections. Of course, this does protect the elderly because those infected would be friends and not family members. However, this may prolong the time needed to fight the pandemic, thereby increasing the suffering of everyone.
It should be noted that the relationship between social norms and COVID-19 strategies is a special case of a more general discussion on whether social norms and sanctions are substitutes or complements.Footnote 99 In the example we considered, the social norm of visiting one’s family is a substitute for any policy that tries to reduce COVID-19 infections, but one might also think of a more explicit substitute for COVID-19 mitigation policies, such as an informal rule of social distancing. Additionally, the degree to which such an informal rule influences behavior may also be heterogeneous and may (or may not) be correlated with the social norm of visiting the parents. For example, countries that emphasize family life may also have a strong tendency of physical vicinity to others as a social norm, making social-distancing measures in light of COVID-19 even more necessary.
Another example can be found in the distinction between “loose” and “tight” societies:Footnote 100 loose societies are informal, individualistic, and expressive, whereas tight societies are formal, orderly,Footnote 101 and characterized by “strong norms and low tolerance for deviant behavior.”Footnote 102 This distinction may help predict the strength of the interaction effects, but interacting societal values may yield countervailing results. For instance, tight countries in which the social norm (e.g., visiting the elderly) is strong will demonstrate a stronger interaction effect but may also tend towards strong compliance to formal rules, such as those imposed by COVID-19 mitigation strategies. Recent empirical evidence from late 2020 suggests that tight countries outperformed loose countries in mitigating COVID-19,Footnote 103 an effect attributed by the researchers to the superior ability to enforce rules.Footnote 104 At the same time, some argue that the pandemic itself may “tighten” some countries,Footnote 105 which would then lead to further heterogeneity. To make things more complicated, some have pointed out an interaction between the degree of looseness and the degree of centralization, suggesting that a “one-size-fits-all” strategy for combating COVID-19 is doomed to fail.Footnote 106 Thus, a full consideration requires observing not only which social norms are in play, but also whether these norms interact with other factors.
B. Taking into Account Environmental factors: the Weather
Next, we consider an environmental factor that policymakers need to take into account in introducing effective COVID-19 mitigation strategies: the weather. Evidence on the connection between COVID-19 and climate conditions is generally mixedFootnote 107 and seems to have received insufficient attention in the early days of the pandemic. While the weather may matter for several reasons, we focus here on how it may affect the level and type of social interactions in which people would like to engage.
People typically draw a higher benefit from staying at home in times of bad weather (e.g., in winter) as compared to times of good weather (e.g., in summer). In other words, staying at home is relatively more attractive in times of bad weather because the opportunity cost associated with leaving one’s home is high, whereas staying at home in times of good weather is relatively less attractive because the opportunity cost associated with leaving one’s home is low.Footnote 108 Similar to the underlying reasoning of complements, one may conceptualize staying at home as complementary to bad weather and leaving one’s home as complementary to good weather. Policymakers should anticipate this line of thought when introducing COVID-19 mitigation strategies to render the strategy more cost-efficient. If we consider the numerical example below in Table 7, it becomes evident that, given the higher benefit derived from staying home in winter, one can achieve the same level of deterrence in winter by introducing a lower fine.
Table 7. Illustration of Choices in Summer vs. Winter
Note: This table presents an illustration of payoffs with and without legal intervention (a monetary fine), in summer vs. winter time. The individual chooses the column yielding the highest utility in the row that applies (depending on the weather and presence of intervention). The utility-maximizing option in each row is marked using bold text.
While this example applies to countries with varying seasons (i.e., in moderate climate zones), the same reasoning holds in a one-sided way for countries where the changes between seasons are less pronounced. For instance, the fines in countries located around the equator may have to be higher throughout the year due to consistently better weather conditions (and vice versa for countries in the high north or south). However, even in these one-sided cases, people in countries with similar weather conditions throughout the year might display higher levels of sensitivity to small changes in the weather.
Weather may also have other effects that influence the rate of COVID-19 infections independent of compliance with restrictions. For example, one study found a positive correlation between the average temperature and the rate of infections in Jakarta, Indonesia,Footnote 109 whereas another study found a negative correlation in Turkey.Footnote 110 Note that a negative correlation would reinforce the example (the private benefit from going outside increases in summer and the cost of going outside decreases due to more infections), whereas a positive correlation is a countervailing effect (i.e., the private benefit from going out in summer is higher and the social cost is then higher as well due to more infections). Moreover, the relevance of the weather may also differ across population segments due to differences in living arrangements. For instance, the inability to go outside should matter more for those who reside in buildings without a garden than for those who live in rural areas.Footnote 111
One could also speculate that bad weather may sometimes drive people to meet others indoors whereas good weather may yield meetings outdoors (which is beneficial, as COVID-19 tends to spread more indoors).Footnote 112 But, outdoor meetings may possibly occur with a larger group of people because being outside entails less space constraints. Hence, the exact interaction effect depends on the circumstances, but is, in any case, relevant.Footnote 113
C. TIMING OF COVID-19 MITIGATION STRATEGIES: BACK-TO-BACK LOCKDOWNS
As a final example, one might consider whether the value of the payoffs for (non)compliance with mitigation policies changes as a result of (non)compliance over time. Consider the case of back-to-back lockdowns; some governments declared a lockdown for a given period of time but then extended the lockdown further, with or without tightening the restrictions.Footnote 114 When this happens, the benefit from the activities prohibited during the lockdown may change over time. For instance, suppose that a lockdown restricts the ability to meet other people (family or friends), leading to a feeling of loneliness and isolation. This feeling can intensify over time if one remains isolated during a lockdown.Footnote 115 Hence, the longer one avoids meeting one’s family—the higher the benefit of a family visit.Footnote 116
In other words, a family visit today and a family visit tomorrow are not perfect substitutes, and hence the dilemma of whether to breach the lockdown for visits differs in back-to-back lockdowns. To illustrate, Figure 1 presents a decision tree capturing the choices and payoffs of an individual who needs to decide whether to stay at home (i.e., comply with the lockdown) or visit family.Footnote 117 In the example we use, the individual gains “4” from visiting friends if he is not lonely. However, if the individual did not visit friends during the first lockdown, he becomes lonely so the benefit of visiting a friend during the second lockdown increases by some factor to “4x” (where
$ x>1 $
). For simplicity, the probability of apprehension is the same in both lockdowns and represented by p. The penalty, however, can vary—so that
$ {f}_1 $
and
$ {f}_2 $
are the penalties for visiting in the first and second lockdown, respectively.
Figure 1 Illustration of Back-to-Back Lockdowns
Note: This figure presents an illustration of payoffs in the example of back-to-back lockdowns. Rectangles represent a decision and circles represent a random process. The choices to Visit (V) and No Visit (N) are marked below the lines. The probability (p or 1-p) are marked above the lines. The last nodes are the payoffs, which account for the benefits of visiting and the respective fines.
Let us assume the individual did visit a friend in time 1 (during the first lockdown), so he is not lonely. The benefit from visiting in time 2 is then still “4” and so the individual visits if (and only if)
$ 4>p{f}_2 $
.Footnote
118 This is the same calculation that the individual does in time 1—he visits if
$ 4>p{f}_1 $
. Therefore, if loneliness is not an issue, then policymakers could just set
$ {f}_1=\hskip0.3em {f}_2\hskip0.3em \ge \hskip0.3em \frac{4}{p}, $
i.e., keep the fine constant over time. However, what happens if we examine the scenario where the individual complied with the first lockdown and did not visit (and is therefore lonely)? Then, during the second lockdown, the individual visits if
$ 4x>p{f}_2 $
. This means that policymakers must increase the fine in time 2.
To see this, suppose that
$ p=0.5 $
and the fine in both time 1 and time 2 is
$ {f}_1={f}_2=12 $
. Then, in time 1, there is no visit, as
$ 4<0.5\ast 12=6 $
. But to deter visits in time 2, it is required that
$ 4x<6 $
, i.e., this only works if
$ x<1.5 $
. Thus, if loneliness is strong enough (
$ x>1.5\Big) $
the same policy that worked during the first lockdown will not work in the second lockdown.
Note that the calculation in our example was simple because there was no effect on the choice in time 2 of being punished in time 1. That is, the choice of whether to visit in time 1 is basically independent of the choice in time 2. If we were to complicate the analysis by introducing different fines for repeat offenders (e.g., by having a different
$ {f}_2 $
depending on whether the individual was caught and penalized in time 1), we would get even more intricate results. An increased fine would decrease the payoff from a repeat offense in time 2, and the anticipation of the fine in time 2 could also affect visits in time 1. This result mirrors the debate in law and economics on whether escalating fines are needed for repeat offenders.Footnote
119 In an even more complicated setting, individuals could also have heterogeneous benefits from committing a crime, so the policymakers’ choices depend on the distribution of benefits.
Accounting for these different complications is beyond the scope of this Article, but the bottom line is that COVID-19 mitigation strategies raise, in essence, a similar question as other types of legal sanctions—and are therefore subject to a large variety of interaction effects. Still, our example illustrates that even in a fairly simple setting, policymakers should not only carefully assess the sequencing of COVID-19 mitigation strategies but also consider how to set the sanctions associated with prolonged COVID-19 measures. In the next Part, we also consider some complications in the form of behavioral effects.
V. DISCUSSION
A. GENERAL
Our analysis shines a spotlight on what policymakers ought to do to best leverage substitution and complementarity effects in fighting COVID-19. Simply put, a general equilibrium approach prompts policymakers to target strategic complements and—insofar as possible—avoid policies that are strategic substitutes,Footnote 120 not only to slow down the progression of the virus, but also to deploy COVID-19 mitigation strategies in a cost-efficient manner.
With respect to strategic complements, the foregoing remarks show that a ban on alcohol, for example—without directly targeting the progress of the virus per se—naturally reduces a person’s incentive to participate in social gatherings, which in turn greatly contributes to slowing down the spread of COVID-19. Policymakers should also be aware that, while complements may be found in all spheres of everyday life, they may well vary from one country to another. That is, policymakers should also take into account environmental factors such as social norms and climate conditions. Bad weather, for example, naturally raises the cost associated with visiting a friend, which means that policymakers may reduce enforcement efforts in connection with social-distancing measures without jeopardizing the effectiveness of the measure. Social norms such as weekly visits of one’s parents also serve as a strong natural incentive to comply with social-distancing measures and should not be lowered through restrictions of freedom of movement. That is to say, preventing a person from paying weekly visits to their family forces policymakers to bear additional enforcement costs in light of the person’s higher incentive not to comply with social-distancing measures within the permitted perimeter.
In contrast, adopting COVID-19 mitigation strategies that are strategic substitutes compromises the efforts to slow down the progression of the virus and are wasteful in terms of resources (as the policies undermine one another). As the foregoing examples show, a person may be inclined to spend more time with their colleagues at the office (such as by working overtime) rather than staying at home. Similarly, in the case of back-to-back lockdowns, as visiting one’s family and friends either today or tomorrow are not perfect substitutes, a person might prefer to violate the rule of staying home during the first lockdown in anticipation of the second, subsequent lockdown. Alternatively, that person may also become more likely to violate the rule during the second lockdown in light of the higher benefit from visiting one’s family due to, for example, acute feelings of loneliness. When considering the extension of a lockdown, policymakers should therefore be aware that enforcement costs are likely to increase.
B. A BRIEF DISCUSSION OF BEHAVIORAL LAW AND ECONOMICS
Our analysis has thus far implicitly assumed that people are rational in the sense that they maximize their utility based on an optimally gathered amount of information.Footnote 121 By looking at how people “really” behave (e.g., by using experimental research methods), behavioral law and economicsFootnote 122 challenge the assumptions underlying traditional law and economics—relaxing the aforementioned assumption of homo oeconomicus. Footnote 123
Instead, behavioralists assume that people may systematically diverge from rational decision-making by, for example, following heuristics (instead of conducting a full-fledged cost-benefit analysis) or falling prey to various cognitive biases.Footnote 124 Several existing papers have already identified the potential effects of these divergences on compliance with COVID-19 mitigation policies.Footnote 125 Fully considering these effects is beyond the scope of this Article, but it is nonetheless interesting to briefly consider how these effects relate to what one might expect in a general equilibrium.
A first, rather obvious, insight is that if people miscalculate the benefits or the costs of their choices, some behaviors may stop serving as a substitute (complement) due to that miscalculation. A second, more interesting insight concerns the behavioral tool known as “nudges.”Footnote 126 Nudges are an intervention that create a choice architecture that change people’s behavior in a “predictable way without forbidding any options or significantly changing their economic incentives.”Footnote 127 For instance, in the context of COVID-19, one mitigation strategy is placing markings on the floor to help people comply with social-distancing measures in a crowded area.Footnote 128 These markings do not reveal any new information or affect the payoffs but serve as a reminder that makes the rule more salient. A recent example is the use of nudges to encourage COVID-19 vaccinations; experimental evidence suggests that sending simple text messages reminding people of their vaccination appointment increases the rate of vaccinationsFootnote 129—particularly when the message makes recipients feel like the owner of a specific vaccine dose.Footnote 130
The interesting question for this Article is then how such nudges affect the general equilibrium. A recent study argues that nudges serve as either substitutes or complements to COVID-19 mitigation strategies;Footnote 131 whenever mandates are impractical (e.g., due to constitutional or political constraints), nudges can be adopted as a substitute.Footnote 132 Respectively, nudges may become complements when mandates are ineffectiveFootnote 133 (e.g., by inducing compliance with the (mandated) rule where enforcement would otherwise be difficult).Footnote 134
Notably, the relationship between mandates and nudges seem independent of the target behavior. That is, two policies that are strategic substitutes (or complements), as we defined them above, may each take the form of either a mandate or a nudge. Thus, the added value of nudges is interesting only insofar as they open new channels to restrict substitution effects among behaviors.
One area where nudges may be relevant is in the avoidance of undesirable substitution effects (e.g., people visiting friends at the office instead of their residence). The mandates which we considered above simply change the payoffs of different choices without any side effects that push people specifically toward the benign activity. However, several studies have shown that nudges could be leveraged to emphasize the moral aspect of decisions by highlighting what would be “the right thing to do.”Footnote 135 For example, placing a message at workplaces that highlights the importance of staying at home may drive people toward the benign choice. Another advantage of this nudge, compared to a mandate, is its relatively lower cost (as posting signs is much cheaper than, e.g., deploying police officers).Footnote 136 However, nudges might not only interact with explicit prohibitions but also with each other—or with other country-specific factorsFootnote 137—which renders the analysis far more complex, if not overly cumbersome.
An additional layer of complexity arises when one considers the time-dimension, such as in the example of back-to-back lockdowns analyzed above. As part of the behavioral movement, several studies argue that intertemporal preferences (i.e., how people evaluate the costs and the benefits arriving at different points in time)Footnote 138 are more complex than what standard economic theory would predict. In particular, a phenomenon known as “hyperbolic discounting”Footnote 139 supposes that people heavily and disproportionately discount future payoffs.Footnote 140 Relating to the lockdowns example, this may mean that either the penalty or the degree of loneliness may be heavily discounted, thereby completely changing the calculation. Hyperbolic discounting is also closely related to the issue of self-control.Footnote 141 Namely, as loneliness accumulates, people may find it increasingly difficult to control themselves,Footnote 142 thus opting to visit a family member in contrast to the predictions of standard models. A similar outcome can result due to yet another effect—like “behavioral fatigue” or, more broadly, “ego-depletion.”Footnote 143 In the context at hand, behavioral fatigue causes people to become less sensitive to new mitigation policies, given they feel depleted from complying with earlier policies. While the effect of behavioral fatigue is controversial, even in the context of COVID-19,Footnote 144 it can affect how policies interact with one another. For example, two substitutes that are deployed at different times may have weaker spillovers because the later policy is met with fatigue.
Such arguments can, of course, be applied to many other behavioral effects.Footnote 145 For instance, the phenomenon of “loss aversion”Footnote 146 can affect people’s incentives to comply with the law during COVID-19Footnote 147 differently depending on which action is perceived as a “loss” and which one as a “gain.” For instance, one might perceive the family visit as a recovery from a loss (rather than as a gain) and therefore be more likely to visit even when there is a risk of being punishedFootnote 148 because losses are perceived to be worse than equally-sized gains.Footnote 149 However, what is perceived as a loss might change over time if a pandemic persists and people get used to a “new normal.”Footnote 150 Moreover, loss aversion can affect discounting,Footnote 151 adding further complications.
While such complications certainly introduce complex challenges to policymakers, their potential presence only supports our main claim: narrow attempts to focus on individual policies are unlikely to be sufficient, and policies relying on such analyses are doomed to fail. However, whether or not behavioral effects are meaningful depends on the magnitude of these effects. If effects are negligible, one can simply neglect them and treat them as random errorsFootnote 152 (but still focus on regular substitution and complementarity effects). Insofar that effects are consistent—they are also predictable.Footnote 153 Thus, one would simply add these to the analysis, but the applicable logic of searching for a general equilibrium would not change.
C. FINAL NOTES
Our analysis mainly builds on traditional deterrence theory—that is, cost-benefit analysis in light of expected sanctions. Some recent studies seem to cast doubts as to whether deterrence plays a central role in compliance with COVID-19 mitigation strategies with some studies finding a weak role for deterrenceFootnote 154 and others suggesting that deterrence cannot work as the sole lever of influence.Footnote 155 Nonetheless, there are at least two reasons why our analyses should be robust. First, some of the studies rely on survey evidence,Footnote 156 which can be less appropriate for estimating actual deterrence. In particular, scholars of law and economics are skeptical toward measurements that do not constitute “revealed preferences” (i.e., people’s actual behavior); hypothetical answers in a survey may thus be less reliable than observed behavior.Footnote 157 This is particularly relevant when measuring compliance because people may be reluctant to be truthful by saying they will break the law. Second, and more importantly, these studies are precisely those whose research question is based on a partial equilibrium—as opposed to general equilibrium—analysis, which does not account for whether deterrence translates into more or less infections.
A final thought experiment might provide further insight. Consider what happens if humanity were to adopt a full global lockdown. That is, shut down the entire world for a period of, say, four weeks. Such an idea may be politically infeasible, but one can make the logical argument that if everything was to shut down at once, then all substitutes would disappear and everyone would switch to benign behaviors like staying at home. Some existing models have tried to analyze such an option, yielding mixed conclusions; one study found that local lockdowns would outperform a global lockdown,Footnote 158 whereas another found precisely the opposite.Footnote 159 Our analysis can provide some insight into why this question is difficult to answer: a complete shutdown not only entails substitution effects, but also complementarity effects. Hence, when a lockdown restricts a complement of a desirable behavior (as in the example of social norms, where discouraging family visits decreases the incentive to take precaution), the result may be an overall negative—depending on which of the effects of the lockdown dominate. This too requires diving into the details and potential heterogeneity across countries. Thus, analyses must be more robust to provide more accurate calculations.
VI. CONCLUSION
As COVID-19 continues to wreak havoc around the globe, designing mitigation strategies that work is more important than ever. However, this is easier said than done. Any attempt to deter socially harmful behavior inevitably sets off a chain of events that entails potential substitution and complementarity effects. While some effects may be difficult to predict, an abstract analysis that ignores complications by focusing solely on partial effects is bound to fail. In other words, when designing mitigation strategies, as the saying goes: the devil is in the details. Interaction effects are thus one of many important details policymakers ought to consider. Using various examples, we illustrated why some policies may be ineffective (or even counterproductive) because they shift behavior to other harmful behaviors (substitution effects) while others can “kill two birds with one stone”—discouraging one bad behavior and indirectly reducing another related behavior (complementarity effects). The examples further demonstrate that the strategies need to account for the heterogeneity of factors, such as social norms, the weather, and the timing of lockdowns.
From a practical perspective, one may ask whether policymakers can truly know which interaction effects are relevant. In other words, how should one estimate the size of the interaction effects? Answering this question requires empirical data. Yet, accurate empirical data is seldom available because governments are often reluctant to conduct public policy experiments. One could imagine that the best way to figure out whether a lockdown works in a certain country would be to implement some kind of a field experiment, where area A is under lockdown whereas area B continues as normal. This would, of course, be imperfect if people could substitute activity in area A for activity in area B (resulting in the same substitution effects we detail in this Article), but the main problem is a political one: people in area A would be outraged that they are serving as “Guinea pigs” in a discriminatory way. However, one could make the argument that this is again the same “public-goods problem”Footnote 160 that COVID-19 mitigation strategies aim to overcome in the first place—that is, the refusal to participate because people do not internalize the benefit to society, even though clearer results may spare others’ suffering (by developing more effective lockdowns).
As the pandemic progressed, some scholars did engage in more elaborate research designs by contrasting sets of policies in specific areasFootnote 161 or contexts.Footnote 162 Moreover, some scholars designed ad-hoc field experiments, mostly to test whether COVID-19 is more or less transmissible under different hygiene conditions.Footnote 163 Furthermore, in some cases, variation in policies across different jurisdictions might meet the requirements of a “natural experiment,” so that causal inferences are feasible.Footnote 164 In any case, in the face of scarce experimental data, policymakers would still be better off expanding the analyses of individual policies into a broader perspective that considers what happens in a general equilibrium.
Another challenge from a policymaking perspective is the multitude of authorities that might be involved in setting the policies. For instance, responses such as travel bans might require coordination between transportation and health authorities, possibly with some overlap in responsibility. This may intuitively lead to either over-regulation, if authorities perceive the situation as an implicit competition, where each wants to be conspicuous in policymaking, or under-regulation, if authorities try to free ride on each other’s efforts.Footnote 165 Whether the former or the latter occurs is, again, a question of whether policies under each authority are substitutes (in which case, there is an incentive to free ride, because the policy of one authority partially solves the same problem that its counterpart is trying to solve) or complements (where there is a stronger incentive to cooperate in order to achieve synergy). A special case of this problem can arise when nearby geographical areas serve as potential substitutes due to differences in regulation, leading some people to engage in the same activity that is prohibited in one place by traveling across the border to another area that does not prohibit the activity.Footnote 166
A different challenge arises if regulators can determine which actions are substitutes. For instance, consider the rules governing so-called COVID-19 certificates (“green passes”).Footnote 167 These certificates grant certain liberties (e.g., access to public events) only to vaccinated individuals or unvaccinated individuals that meet narrow exceptions. The economic logic of such certificates seems to be two-fold: (i) they incentivize people to get vaccinated, as doing so yields more liberties, and (ii) they minimize the restriction of liberties of the vaccinated, who are generally at lower risk of transmitting the virus. Given the latter rationale, some governments subsequently exempted vaccinated people from the need to get tested,Footnote 168 thereby determining that tests and vaccines are substitutes from the individuals’ perspective. When the relationship between actions (substitutes or complements) is endogenous (i.e., under the regulator’s control), governments face the additional complication of how to optimize the menu of actions available to individuals. A sub-optimal menu can easily lead to the same problems discussed above. For instance, as vaccinated individuals are not entirely risk-free, tests and vaccines might be preferably set as complements rather than substitutes. It may be preferable to grant rights only to those who are both vaccinated and tested. Deciding whether this is indeed preferable requires a delicate balancing of costs and benefits (and indeed there is some debate on whether COVID-19 certificates are proportional)Footnote 169 but is a necessary part of accounting for interaction effects.
Finally, investing effort in locating the general equilibrium is only efficient if the costs of calculation are not too high compared to the benefit. That is, even if there are spillover effects of some policies due to interactions with other policies, they need to be of sufficiently large importance to justify the cost of investing resources to calculate the different payoffs.Footnote 170 In this context, one should also consider the difficulty of choosing the right measure, as the spread of COVID-19 can be estimated using many different variables (e.g., number of cases, number of deaths, or number of hospitalized patients), which further raises estimation challenges.Footnote 171
A social planner seeking to balance costs and benefits in terms of COVID-19 mitigation strategies must therefore construct a target function and engage in solving an optimization problem.Footnote 172 This is no simple task and one must account not only for the multitude of effects but also uncertainty, which requires some adjustments.Footnote 173 Moreover, even with reliable experimental data on which policies work, scaling up might be difficult.Footnote 174 While mitigation strategies may be less susceptible to funding constraints (which are one typical inhibitor of scaling),Footnote 175 they are still often fragmented and based only on partial dataFootnote 176 unless some form of effective global data-sharing cooperation emerges. Encouragingly, the COVID-19 pandemic has indeed spurred some initiatives for keeping track of infections and comparing policies.Footnote 177 Combining such initiatives with a general equilibrium approach may therefore be a positive step in the right direction. This combination need not restrict attention to standard preferences and can (or even should) combine behavioral aspects (e.g., by adopting a stepwise model along the lines of so-called “Behavioral Interventions to Advance Self-Sufficiency” (“BIAS”) model).Footnote 178 This model is a stepwise process: it begins with identifying where public policies perform sub-optimally and proceeds by analyzing which behavioral explanations can potentially explain the policy’s failure using feedbacks from stakeholders in the field. Thereafter, randomized control trials are used to contrast potential interventions, including nudges,Footnote 179 to develop an evidence-based solution. While behaviorally-informed policies (and nudges in particular) may also face scaling challenges,Footnote 180 we would view such an approach that focuses on a general equilibrium and accounts for behavioral effects as a success.
