The Roots of Policy Incoherence: Domestic Policy, Global Public Goods, and International Development

October 12, 2021

Governments make policy to affect three domains: domestic outcomes; outcomes in foreign countries; and shared global challenges. These policy domains are interconnected, but not necessarily in obvious ways, and few governments make policy across them in a rational manner. This note sets out how the conceptual and analytical incoherence of policy set in developed countries across these three domains undermines their own effectiveness—most notably on international development and shared global challenges. It sets out the incentives that drive this incoherence and the opportunities for greater impact being missed through it, and concludes with a brief consideration of how institutional reform can support better policymaking across these domains.

Most governments treat domestic policy as an isolated domain over which to pursue purely domestic priorities, while international development policy and global public goods (GPG) policy are barely distinguished from each other. Such a conceptualization of the policy space leads to incoherence: one domain undermining another, and opportunities for positive reinforcement left begging. Domestic policy is pursued without much thought to its effects on GPG and development policy, while the latter two are either an afterthought with respect to their impact on domestic policy or worse, treated transactionally. What’s more, action on international development and global public goods is too often considered interchangeable, a conceptual error that undermines both. These errors arise from a failure to adequately consider whether policy actions in different domains are complements, substitutes, or independent of each other. Treating them as independent when they are in fact complements, or treating them as substitutes when they are actually independent, leads to an inefficient policy mix, and ultimately muted impact.

This short note proceeds as follows. The next section introduces the conceptual basis for considering policy actions as complements, substitutes, or independent, and the implications this has for how they should be set. The third section applies this framework to a few examples where policy coherence could be improved by adopting this approach. The fourth section discusses the incentive and institutional barriers that hamper its adoption. Finally, the note concludes with consideration of how institutional reform can support improved, more internally consistent policymaking.

A simple approach to interdependencies in the policy space

A simplified conceptualisation of a government’s set of policy concerns is necessary before we can make judgements over how different policy concerns should interact. In its simplest form, we can argue that there are three kinds of policy. First, there are those that are primarily self-interested, aimed at the betterment of the domestic population (now and in the future). I call these “domestic policy concerns,” aimed at influencing domestic outcomes. Second, there are those that are primarily about improving welfare in other countries. These policies may have second order benefits to the country implementing them, but the primary objective is the betterment of people in other countries. These are grouped under “development policy concerns,” aimed at improving international development outcomes. Third, there are policies that benefit the implementing countries, with positive externalities that accrue to other countries. These are classed as “global public good policy concerns,” aimed at improving outcomes with respect to shared global challenges.

Three—hopefully uncontroversial—assumptions set the conceptual argument up. First, all three policy areas matter. That is, we care about results in all three, not just any one. Second, we care about effectiveness of action or policy in each sphere and overall. There is a minimal level of effectiveness in each sphere we seek to achieve, and a desire to maximise effectiveness in each sphere, and the overall amount of “good” achieved. Third, that these policy spheres interact with each other in stronger and weaker ways. For example, domestic outcomes affect global public good outcomes and development outcomes, both positively and negatively. These relationships may not be symmetrical, in that influences in one direction, for example from domestic outcomes to development outcomes, may be much stronger than the influence of development outcomes on domestic outcomes.

Given these three assumptions, the content of the relationship between different policy domains determines what constitutes a coherent policy set. Since governments care about multiple objectives over multiple domains, it should consider the net effect of its policies in the round when taking any new action. This, in itself, is not in any sense a novel argument: the idea of coherence underlies, for example, CGD’s Commitment to Development Index. Differentiating the precise form of interdependence, however, leads to richer conclusions.

Where policies or outcomes in one domain are fully independent of the policies and outcomes in another, policy coherence is easy: nothing a government does in one domain affects the other. A good example here might be policy concerning football club ownership (where policy varies substantially between countries) and policy on the funding of energy projects in third countries. It’s hard to think of a way in which any policy choice in one could possibly affect the set of possible outcomes of the other, or alter in any way the effectiveness of any chosen policy. No matter how football clubs are governed, any choice of restriction (or lack thereof) on the funding energy projects abroad will operate as intended. Were the policy space entirely populated by such policy sets, each individual policy could be determined separately, changed as seen fit and implemented in any fashion without ever reducing the overall efficiency of government action. And it is indeed the case that most policy actions are functionally independent of most others.[1] Unfortunately in the remaining cases, governments often set policy as if independence holds, when in fact it does not, leading to suboptimal outcomes.

Where outcomes in one domain substitute for those in another domain, the parsimonious policy choice is to take policy action to achieve only one. If achieving A has indistinguishable (or nearly so) effects from achieving B, policy to achieve both is unnecessary and inefficient. For example, if domestic policy governing the practice of making remittances to developing countries were to have indistinguishable effects from international development policy to improve reporting standards for correspondent banks, policy to achieve either one would achieve outcomes relating to remittance flows and either could be pursued without the other to that end.[2]

When outcomes in one domain complement outcomes in another domain, achieving one amplifies action in the other—and failing to achieve one undermines the other—an optimal policy set would increase effort on both. Take, for example, action to vaccinate people against COVID-19 domestically, aimed at reducing infections and adverse health outcomes, and action to increase vaccine availability in developing countries. These are complements because increasing vaccination rates at home reduce health risk from contracting COVID, but also increase the return to vaccinating people in countries with lower vaccination rates, for two reasons. First, reducing the pool of infected and unvaccinated people in whom new COVID variants may emerge increases the value of existing vaccinations given at home. Second, by reducing infections and hospitalisations abroad, the return to resuming normal economic and trade ties increases. Doing only one well mutes the returns; doing both well amplifies them.

Confusing these relationships makes policy less effective. Because the implications for what an optimal policy set looks like vary with the kind of relationship between pairs of policy action, getting them mixed up leads not just to incoherence, but missed opportunities and self-sabotage. The next section briefly sets outs examples of this.

Public policy routinely misidentifies complementarities, substitutes, and independent pairs

Vaccine policy in the developed world treats domestic and international action as substitutes, rather than complements, and thus hampers its own effectiveness. The rush to vaccinate ever-less vulnerable populations in developed countries, to hoard purchased vaccines even when regulatory clearance is unlikely or much-delayed, and to purchase boosters before massively scaling up the availability of vaccines in barely vaccinated developing countries suggests that policymakers treat these objectives as substitutes. In fact, they are complements: the susceptibility of a population to COVID-19 depends on both its level of protection and the amount of the virus in circulation. There comes a point—likely long past in many developed countries—where inching protection of the population up marginally will have much less effect than reducing the global burden of disease. More generally, as your demand for one (domestic vaccination) increases, so should your demand for the other (vaccination in largely unvaccinated places) This will only become more true as countries relax travel restrictions. Indeed travel restrictions are a second (or third, or fourth) best solution to a policy problem in which the first best solution would have been rapid deployment of vaccines globally. By failing to recognize the complementarity of domestic and international policy, both were undermined.

Climate policy is typically seen as a complement or substitute for international development policy, when they are—usually—better understood as independent. The general trend has been for donors to fund a substantial portion of their climate action using official development assistance (around 84 percent, according to official estimates). This implies that climate action substitutes for action on development (or is a complement to other development aid and thus increases its effectiveness). Yet at best, development outcomes and climate outcomes are mainly independent. Developing countries—and particular, African ones—contribute a small proportion of global emissions (Moss, 2020), so action on climate in developing countries will normally either have a small effect on global climate change goals (for most mitigation policy) or primarily benefit only those in developing countries (for most adaptation policy). Funding global public good climate activity is thus mainly good for only development or only climate objectives, and not both (except for a small number of opportunities that may genuinely thread the needle of being good for development and good for global climate objectives, which are the subject of a forthcoming set of papers). Even worse, some policies that seek to limit emissions in developing countries in the name of global climate change objectives may be actively bad for development (Ramachandran, 2021) while also being next to meaningless for climate change goals.

Migration policy and development policy are usually thought of as either independent or complements in the wrong direction. Developed countries have often set migration policy in such a fashion as to minimize inflows of people from developing countries, often while maintaining a large and generous ODA budget. For a long time, these domains were treated as broadly independent, set in unrelated policy silos, while research increasingly suggests that they are positive complements—that migration (and even travel ) to developed countries exerts a positive effect on development in home countries (Koczan et al., 2021; Umana-Dajud, 2019). More recently political rhetoric has suggested that spending on development was an effective way to reduce migration , despite research from a number of angles showing that the exact opposite was true (Angelucci, 2015; Clemens & Mendola, 2020).

Each of these examples demonstrate the same basic conceptual error: failing to correctly identify the nature and direction of the interaction between policy or outcome pairs. That said it would be naïve to see these purely as cognitive failings. They stem from the structure of incentives and institutions that govern and set bounds to policy making. This makes them difficult to resolve, but also provides a framework for how to do so.

The institutional roots of convenient misconception

In each of these cases, conceptual errors are driven primarily by incentives and institutional structures that facilitate convenient misconception. None of the foregoing analysis for any pair of mismatched policies will be new to experts (or even well-informed observers) of them. Understanding why they arise requires a more sophisticated theory than simple misunderstanding or lack of access to information. I suggest three main reasons why incoherent policy is an equilibrium in most developed countries.

First, the electoral returns to designing a coherent policy set are low. Voters tend to have incomplete information, inadequate time and resources to compute information fully, and limited attention. Even if voters have a preference for coherent policy, they may not be able to articulate what coherence looks like.[3] As a result, they tend to focus on those domains of policy that carry most salience for them, with the consequence that policy preferences are formulated as if policy domains are independent—which, as we have discussed, is often untrue. To the extent that voters are unable to rationally assess a politician’s impact on the outcomes they care about, and reward them accordingly, (Wolfers, 2007), and depending on the relative electoral returns of different policy outcomes or domains, politicians maximize their returns by treating them as independent. This is true even in the direct knowledge that the actual outcomes they are pursuing will be undermined by treating them as such.

Second is the mistaken and under-scrutinised belief that all good things are alike. This might be described as the Anna Karenina fallacy, from the oft-quoted first lines of Tolstoy’s novel: “All happy families are alike; each unhappy family is unhappy in their own way.” [4] Governments treat policy and effort dealing with “good deeds” as interchangeable—hence the belief that working on climate change is somehow the same as working on development, and vice versa, and the idea that girls’ education and climate change are interchangeable deep complements when in fact, more education is likely to increase emissions (Devonald et al., 2021).

Third is the political and fiscal convenience of win-wins. The most obvious reason why development aid is used on things that are not particularly good for development is not mistaken analysis, which is a useful rhetorical cover, but fiscal convenience. Many governments begrudge the use of scarce fiscal resources on activities with the primary aim of bettering welfare in developing countries (Dissanayake, 2021). One way they seek to swallow this bitter pill is by using the money to achieve orthogonal objectives, be that climate change (a GPG objective) or domestic objectives (such as increased trade or exports).

These mechanisms suggest a limit to simple information interventions. The next section suggests a few institutional reforms that can support better, more coherent policymaking across domains.

Changing the incentives requires institutional action

Addressing these root causes of policy incoherence requires more than simply better educating politicians and the public. Instead, reforms could and should focus on how to set limits on government action across domains, and how to assess their net effect.

The development of a policy analysis equivalent to the UK’s Office for Budget Responsibility (OBR). One of the reasons why the electorally optimal strategy for policymaking is often not aligned with the optimal policy set defined by outcomes in each policy domain is that there is no independent institution with a mandate to investigate the likely net effect of policy in the way the OBR is tasked with producing accepted economic and fiscal forecasts. Such an institution would have a harder job, needing to marshal expertise across several domains, and often using less data than is available on government spending, but would play an important role. Of course, think tanks and research institutes also aspire to this function, and could inform such a body. But what makes the OBR a strong institution is that the UK government has accepted it as the arbiter of fiscal analysis. It is not the analytical capacity per se but its institutional role that helps.

Absent this, legislation and rules set by multilateral institutions can help mitigate the Anna Karenina fallacy. The OECD’s DAC sets rules governing how ODA is to be counted by most advanced economies—it surely has a role to play in policing the classification of climate spend as ODA. Meanwhile, the UNFCCC should set rules or targets on minimum non-ODA spend on GPGs in the climate space. This could even happen at the national level, as Norway’s mooted move to completely separate ODA from GPG spend suggests.

Increase the use of impact-focused modalities of policy action. CGD has long argued for increased use of cash-on-delivery modes of aid, and pioneered the use of an advance market commitment (AMC) for vaccine delivery in developing countries (AMC Working Group, 2005; Birdsall & Savedoff, 2010).But there is no reason why such methods should be limited to any one policy domain. Indeed, the more complex the interdependencies under consideration, the more we should consider using AMCs, cash-on-delivery, and results-based financing more generally, to incentivize action and policy that genuinely achieves multiple objectives and mitigates the temptation for rhetoric or convenience over impact.[5]

Perfect policy is unachievable (and indeed may not exist); but institutional solutions for the most egregious failures of policy coherence should be possible. Facing multiple global, domestic, and development challenges, governments (especially those with fairly secure governing mandates) should be thinking very carefully about how to get the most from limited resources. A start would be to stop undermining themselves willfully with suboptimal policy sets.


AMC Working Group. (2005). Making Markets for Vaccines: Ideas to Action. /sites/default/files/archive/doc/books/vaccine/MakingMarkets-complete.pdf

Angelucci, M. (2015). Migration and Financial Constraints: Evidence from Mexico. The Review of Economics and Statistics .

Birdsall, N., & Savedoff, W. D. (2010). Cash on Delivery: A New Approach to Foreign Aid. Center for Global Development.

Caplan, B. (2009). Irrational Principals. Review of Austrian Economics.

Clemens, M. A., & Mendola, M. (2020). Migration from Developing Countries: Selection, Income Elasticity, and Simpson’s Paradox. CGD Working Paper. /sites/default/files/migration-developing-countries-selection-income-elasticity-and-simpsons-paradox.pdf

Devonald, M., Hares, S., Jones, N., Moscoviz, L., Rossiter, J., Shaw, P., & Yadete, W. (2021). Fund Girls’ Education. Don’t Greenwash It. /blog/fund-girls-education-dont-greenwash-it

Dissanayake, R. (2021). Turning the UK from a Development Superpower to a Development Minnow in Six Easy Steps. /blog/turning-uk-development-superpower-development-minnow-six-easy-steps

Koczan, Z., Peri, G., Pinat, M., & Rozhkov, D. (2021). The Impact of International Migration on Inclusive Growth: A Review. file:///C:/Users/badrogue/Downloads/wpiea2021088-print-pdf%20(2).pdf

Moss, T. (2020). Infographic: What is Sub-Saharan Africa’s Contribution to Global CO2 Emissions?

Ramachandran, V. (2021). Blanket banks on fossil-fuel funds will entrench poverty.

Tolstoy, L. (2013). Anna Karenina.

Umana-Dajud, C. (2019). Do visas hinder international trade in goods? Journal of Development Economics, 140, 106–126.

Wolfers, J. (2007). Are Voters Rational? Evidence from Gubernational Elections.


[1] “Functionally” independent because there are many cases where there is some theoretical pathway through which one policy may ultimately exert some weak or imperceptible influence on another policy domain. We will ignore such pathways here, though in aggregate they may matter rather a lot for what the optimal policy set is, since there is no practical way to imagine policymaking accounting for them.

[2] This is, of course, a great simplification, as supporting correspondent banks will have spillover benefits on their ability to engage with third countries; meanwhile changing policy at home affects many other countries at once. The point is that the two policies partially achieve the same objective and should be set in reference to each other—reducing, relative to the assumption of independence, the optimal effort on either.

[3] Indeed Bryan Caplan argues against an extensive role for the state on the basis that voters are irrational and unable to effectively hold it to account (Caplan, 2009). I don’t go down that path here, but the basic argument that voters are—understandably—imperfectly informed and rationally outsource the job of deep policy analysis to others is the basis of this argument.

[4] Quoted from (Tolstoy, 2013), Pevear and Volokhonsky’s translation. Both Stefan Dercon and Dan Honig have previously alluded to the “Anna Karenina principle” in their work, though each focuses on the second clause rather than the first – looking at how fragile and conflict affected states are each fragile in a different way.

[5] Subject, as always for such methods, to the practicalities of contracting.

Rights & Permissions

You may use and disseminate CGD’s publications under these conditions.