“Effective altruism,” the philanthropic movement founded on Peter Singer’s ideas, applies a consequentialist philosophy to the problem of global poverty, taking the position that the impact of our actions should guide our giving. To do “the most good we can,” so-called effective altruists reason, we ought to route our philanthropy through effective organizations that spend money efficiently and make the greatest possible impact.

Unfortunately, as Angus Deaton reminds us, identifying a highly effective organization is far from simple. The “aid curse” literature has convincingly shown that pumping aid money into corrupt governments rarely achieves redistributive goals: aid often doesn’t reach its intended recipients, and the money can fund harmful, kleptocratic activities that hurt the poor. In light of this, the philanthropic community has turned its attention toward non-governmental organizations, or NGOs, whose reputation as charitable and altruism-driven has made them the development darlings of the twenty-first century.

However, NGOs vary widely in how well they work. Some are highly efficient and run by well-intentioned leaders, while others (sometimes termed PONGOs, “politicians’ own NGOs”) are little more than moneymaking, reputation-laundering schemes. Even assuming that effective altruists are able to weed out corrupt, vacuous charities, the puzzle of how to identify effective organizations—charities likely to have the desired impact on poverty—remains thorny.

In recent years, one approach has emerged as the new gold standard in development economics: the randomized controlled trial (RCT) field experiment. RCTs are designed to make the complex social world as much like a scientific laboratory as possible in order to isolate the effect of a particular intervention. The focus on impact makes RCTs appealing to effective altruists. Indeed, organizations like GiveWell and Giving What We Can rely heavily on these studies as the “best available evidence” to recommend top charities.

As in medical studies, RCT researchers randomly assign subjects to treatment and control groups to ensure that the two groups are roughly identical prior to the experiment. Then they administer the intervention—mosquito bed nets, de-worming pills, curriculum interventions, eye surgeries—only to those in the treatment group. Any differences in outcomes (malaria rates, parasite infection incidence, literacy levels, vision) between the treatment and control groups are attributed to the intervention. The clean research design makes researchers confident they have correctly identified whether a program has had the intended impact.

The puzzle of how to identify effective organizations—charities likely to have the desired impact on poverty—remains thorny.

However, this approach to assessment has a serious downside: RCTs only capture a narrow view of impact. While they are good at measuring the proximate effects of a program on its immediate target subjects, RCTs are bad at detecting any unintended effects of a program, especially those effects that fall outside the population or timeframe that the organization or researchers had in mind. For example, an RCT might determine whether a bed net distribution program lowered the incidence of malaria among its target population. But it would be less likely to capture whether the program unintentionally demobilized political pressures on the government to build a more effective malaria eradication program, one that would ultimately affect more people. RCTs thus potentially miss broader insights and side effects of a program beyond its target population.

Effective altruists are committed to evidence-based selection of charities, but in interpreting RCTs as the “best available evidence” they have prioritized certainty, narrowing the scope of impact they consider in identifying top charities.

This choice has built political and institutional blind spots into the way the effective altruism movement redistributes money. All charities exist within a broader ecosystem of service providers that includes the welfare state. NGOs that distribute bed nets and provide vaccinations operate alongside an array of public health programs run by the state. NGO-run schools operate up the road from government-run schools. While these state-run programs may be performing poorly and lack resources, they are still the core provider for the majority of the poor in many developing countries. For example, despite the emphasis on privatization of the Indian education system in recent years, well over half of the rural poor are in government schools. As one education NGO leader in India remarked to me, none of us have the illusion that we can replace the state. And evidence from the literature on NGOs suggests that their much-lauded inclination to target the needy is mediated by their own organizational interests, like demonstrating impact to funders and establishing themselves in places where NGO workers want to live. Often, NGOs end up serving a group we might call the “middle poor,” rather than those most in need. Instead, the very poor must often rely on the state to deliver services, inadequate as they may be.

The quality of the state’s social service provision thus critically shapes welfare outcomes for many of the poorest people in the world. Yet it seems that once effective altruists have—for good reasons—ruled out governments as eligible recipients of effective aid, their attention to the state drops off entirely. The RCTs used to evaluate the impact of effective altruism’s favorite charities typically examine outcomes only among individuals targeted by the charity’s programs. State-run public service institutions running in the background are not the intended target of most of these charities, so they are rarely featured in RCTs designed to measure the effectiveness of a program. Thus unintended institutional effects on government welfare programs are seldom incorporated into effective altruists’ calculations about worthwhile charities to fund.

Yet any scholar of the political economy of development would be skeptical of the assumption that the welfare state in poor countries would remain unaffected by a sizeable influx of resources into a parallel set of institutions. My research in India suggests that in the best cases, the presence of NGOs can lead to learning across sectors, where nearby government actors may use NGO programs as a demonstration model for their own service delivery strategies. This, of course, assumes that these government actors are interested in learning—if corruption is rampant and political will for reform is weak, learning is unlikely to diffuse into the government system or translate into widespread improvement in state services.

In the worst case, the presence of NGOs induces exit from the state sector. When relatively efficient, well-functioning NGOs enter a health or education market, for example, citizens in that market who are paying attention are likely to switch from government services to NGO services. The result is a disengagement of the most mobilized, discerning poor citizens from the state. These are the citizens most likely to have played a previous role in monitoring the quality of state services and advocating for improvements. Once they exit, the pressure on the government to maintain and improve services eases, and the quality of government provision is likely to fall.

This dynamic, sometimes called skimming, has unfortunate consequences for those most in need of services. Even when NGOs intend to target the poorest and most marginalized, their services tend to be utilized by the slightly less poor, who are particularly good at seeking out the best services. These are the people who represent the most high-value citizens from the perspective of pro-poor state development—they share the general interests of the most marginalized, but they have more political voice and thus provide local public goods by advocating for better services from the government. When they exit the state system, it has distributional consequences for those who receive state services – those who are already the worst-off are left behind in the state system, service quality drops, and their interests are harmed.

In rural villages in Rajasthan that are home to both NGO schools and government schools, this story of exit is playing out in the short run, and there are signs of a longer-term shift that does not bode well for the future of the local government education system. In one village, many families have shifted their children out of local government schools and into the NGO school that opened several years ago. In the worst government schools, this appears to have resulted in a further drop in quality, as teachers are relieved of the pressure they once felt under the watchful eye of education-minded parents.

One villager I interviewed—an esteemed local elder, and one of the best advocates for education in the village—told me that before the NGO school opened, his grandchildren attended a nearby government school, and he used to conduct daily monitoring visits there. But when the NGO school opened, he was one of the first to shift his grandchildren out of the government school in favor of higher-quality facilities and more accountable teachers at the NGO school. Now he spends hours at the NGO school every week, checking to see that the teachers are in their classrooms on time, ensuring that the water system is functioning properly, and scrutinizing the quality of the midday meal. He no longer has the time, or motivation, to visit the government school.

In the worst case, the presence of NGOs induces exit from the state sector.

The shift in this elder’s critical attention away from the state was not an isolated case. Nearly all villagers I interviewed reported that in the future they would be more likely to direct their advocacy for better education services toward the NGO than toward the government. When citizens exit the state system, they are much less likely to use their political voice to push for improvements in that system. Those who are left behind tend to be poorer and less mobilized. When service quality falls in the absence of vigilant oversight, the most disadvantaged fall further still.

In other words, the work of these charities has the power to reshape the dynamics of citizenship among the poor. By providing services that sometimes prompt a shift in the attention of critical, mobilized citizens away from the state and toward the NGO sector, these charities may interrupt channels of democratic accountability. These channels linking the poor to the state, shaky as they are, may still be necessary for the maintenance of and longer-term improvements in state welfare systems that serve the very poor.

The critical point here is that there is a political dimension to poverty that is ignored in the standard process effective altruists use to target their giving. Charities funded by the movement operate in an inescapably political environment, one in which citizens make demands on their governments to deliver promised services and meet their basic needs. These political channels that link the poor to the state are often precarious and partially functioning. Nevertheless, little else prevents governments from divesting from basic service delivery for poor segments of society.

The effective altruism movement is not alone in its inattention to these dynamics. Observers have commented on the “anti-politics” of the development industry, arguing that development actors tend to seek technocratic solutions to poverty that skirt the touchy issues of politics but leave untouched the power structures that create and maintain systems of poverty.

Broadening the scope of what effective altruists deem the “best available evidence” is a good start. But from a consequentialist standpoint, it is not enough for effective altruists to simply tweak their approach to RCT design. They must contend with the fact that the state remains the primary provider of basic social welfare for most poor citizens in most poor countries, and that pumping money into a parallel set of providers—even good ones—without a plan for reaching the coverage or scale of a state may do serious harm to the poor who are left in the state system.

The politics of state service provision for the poor is a messy business. The simplest thing for far-away philanthropists is to simply sidestep politics and fund programs that appear, on the surface, to be positioned outside the political arena. But it is clear that this neat separation between civil society and the state is a fiction. If the movement is to meet its own consequentialist standards, its leaders and philosophers must make room for the state and for the politics of service provision among its calculations for identifying recipients for its aid.

• • •

At the center of the politics of poverty is the issue of state accountability. Here, a different type of organization, advocacy groups, are working to hold their governments accountable on issues of corruption and service delivery. In settings where the poor mostly receive services from the state, but where government corruption constrains the quality of state service delivery, advocacy organizations may be in a better position than service provision NGOs to broadly improve services delivered to the poor, though this type of welfare-improvement work involves a longer timeframe. Because advocacy directly targets the root problem underlying inadequate services—the problem of state accountability—successful advocacy can have profound welfare payoffs.

But as philanthropic funds have flowed into direct service provision NGOs, advocacy groups often find themselves unable to finance their operations. The costs of running effective advocacy organizations are increasing with the professionalization of politics, and the supply of donors willing to fund contentious politics have not kept pace in developing countries, where domestic sources of funding for advocacy are often hard to find. And governments that maintain a tighter hold on civil society sometimes crack down on advocacy organizations that receive funding from abroad, claiming foreign government interference in domestic political affairs even when NGOs receive a majority of their funds from private citizens. This occurs even in democratic settings—in a recent move widely critiqued as an attempt to silence dissent, India’s Modi government froze the bank accounts of almost 9000 NGOs, including Greenpeace, accusing them of violating rules governing international financial transactions.

International private philanthropists—like effective altruists—may be the best hope for the financial sustenance of these watchdog groups, since their donations are more likely than official aid to be permitted to fund pressure politics. But the work of advocacy is difficult and unpredictable, and its effectiveness can be hard to measure. On the margin, each advocacy “intervention” is less likely to lead to immediate, measurable improvements than, say, a nutrition supplement or a malaria vaccination. Thus this type of organization is unlikely to meet the criteria effective altruists currently use in selecting charities.

Effective altruism without a political blind spot would revisit its funding philosophy with a cautious eye toward state substitution. It would narrow its scope of direct service funding to non-substituting types of service delivery, and redirect substantial funding toward advocacy organizations that specifically work toward strengthening the accountability of the state sector. There may be some specific types of services—eye surgeries, for example (an effective altruism favorite)—that service delivery NGOs are especially efficient at delivering, and that can be delivered in a targeted way without building a parallel apparatus that decouples citizens from the government. In addition to identifying and funding this low-hanging fruit, effective altruists should invest in initiatives that strengthen, rather than weaken, democratic channels of accountability that demand improvements in the government’s systems of public health, sanitation, and basic welfare services.

Effective altruists may object that they know little about the political contexts in which these battles play out, and interfering without an RCT is too risky. This is an important concern. But at some point, once bed nets and eye surgeries are no longer the most pressing need, the movement will be faced with the thornier problems of development, which reside in the political arena. Engaging with the complexity of those problems now, and beginning to develop best practices for funding organizations that work to improve the quality of state services, will give effective altruists a head start on doing the most good they can—for those most in need—in the long run.

Editors’s Note: Read Director of Research at Giving What We Can Hauke Hillebrandt’s response to Emily Clough.