In Ecuador, as thousands of protesters shut down streets, set fires and pushed into the National Assembly building, President Lenín Moreno declared a state of emergency and moved his government out of the capital. Protests began Oct. 1 when Moreno announced an end to government subsidies on fuel prices. Seven people have been reported dead and 1,152 arrested. Late Sunday, Moreno and protest leaders struck a deal to end the protests — and again reduce fuel prices.

The decision to repeal fuel subsidies was prompted by targets set within Ecuador’s $4.2 billion IMF arrangement, which require the country to tackle its fiscal deficit. Fuel subsidies were costing the government $1.3 billion annually. Although the International Monetary Fund proposed a “careful and gradual optimization of fuel subsidies” by Thursday, the price of gasoline increased by about 25 percent and the price of diesel doubled. While protests were initially led by transportation workers, they’re now led by Ecuador’s powerful indigenous federation, CONAIE, which has led thousands of indigenous citizens from Ecuador’s Amazon basin to Quito for the national strike.

Why has ending fuel subsidies repeatedly set off protests and revolutions?

The protests raise a common dilemma for political leaders: How can governments reform fuel subsidies — which are both fiscally and environmentally disastrous — without setting off widespread protests? Last year’s yellow vest protests in France began in response to increased fuel prices. Further back, fuel price increases triggered protests that ultimately deposed Indonesian leader Suharto in 1998; set off Burma’s (now Myanmar’s) Saffron Revolution in 2007, and launched the Occupy Nigeria movement in 2012.

In recently published research, I study why poorer citizens reject fuel subsidy reform — even if governments promise to shift funds into programs that could really benefit the poor, like cash transfers and better infrastructure. Moreno, for instance, promised to expand welfare subsidies for the poor. My research suggests citizens reject fuel subsidy reform when they do not trust governments to deliver on those promises. Although I conducted this research in Indonesia, there are important lessons for Ecuador.

Fuel subsidies mean more to citizens than low gasoline prices

Consumer fuel subsidies often start by accident. Many subsidies, including those in Ecuador and Indonesia, began in the 1960s and 1970s when governments launched programs to stabilize prices of basic goods. As fuel prices rise, however, price stabilization programs can balloon into large-scale subsidies. That happens if governments initially set a fixed price — when the consumer pays the same price and the size of the subsidy changes with market prices — instead of a fixed subsidy — when the consumer pays different prices with market fluctuations but the size of the subsidy stays the same.

Citizens know fixed prices mean that whenever gas prices changes, the government has made a deliberate policy choice. That makes fixed price programs particularly difficult to overturn — which stymies global progress on reducing carbon emissions. Citizens in countries that extract and sell oil or gas have come to see low gas prices as among the few or only ways they actually benefit from resource extraction. Ecuador’s indigenous groups have long protested oil drilling in the Amazon basin for destroying their environment. Losing the one visible benefit from that cuts deeper than other austerity reforms.

Citizens are more likely to support subsidy reform when they trust the government

My research tested when citizens in Indonesia were willing to support fuel subsidy reform. When I did my research in 2013, Indonesia was in the midst of a multiyear process of increasing the price of fuel in fits and starts, while giving poorer citizens more targeted social programs, like cash transfers and health fee waivers for poorer households. My theory was that citizens would be less likely to trust new programs if their local government officials were corrupt.

To test this, I first examined Indonesian administrative data on the total benefits that arrive each month to particular villages. Then I compared those records with household survey data from more than 10,000 respondents across 572 villages that economists Abhijit Banerjee, Rema Hanna, Benjamin Olken, Sudarno Sumarto and I collected as part of a large-scale experiment on improving poor households’ access to social programs. These surveys asked what benefits the household actually received.

The difference between how much was supposed to arrive and how much actually arrived gave me an estimate of how much of a village’s total allocated benefits “go missing” each month. This proved a remarkably accurate way to estimate whether poorer citizens actually benefited from the shift from mass fuel subsidies to government cash transfers. A later survey we conducted in the same villages asked whether citizens received the cash transfers linked with the 2013 fuel price hike. Citizens in “corrupt” villages — those with a larger gap between recorded and received benefits — reported they were less likely to receive the cash transfer at all, and more likely to have to “donate” some to governmental officials.

In the first survey, we asked Indonesians to imagine that they were designing the national budget and could scale up three government programs but also had to scale down three programs. Indonesians were much more likely to support scaling down fuel price subsidies (and less likely to support scaling them up) in areas with less corruption — regardless of their economic circumstances or how much they benefited from the subsidy. In villages where corruption levels were near zero, poorer citizens were more than 2.5 times as likely to support rather than to oppose reform.

Lessons for Ecuador

Solving Ecuador’s fiscal crisis and retaining IMF support are key priorities for Moreno. His approach makes sense: Who could oppose replacing regressive, expensive and environmentally unsustainable fuel subsidies with expanded welfare payments for poorer households? But if citizens don’t trust Moreno’s government to deliver, they aren’t likely to leave the streets.

Governments seeking to replace fuel subsidies may wish to focus on delivering what citizens will find credible. That may look like Indonesia’s successful approach: enacting reforms piecemeal rather than all at once and showing that each cut in fuel subsidies actually will be replaced by spending on the poor. Citizens, in short, are more willing to go along with reducing fuel subsidies in areas where the government has earned their trust.

Jordan Kyle (@jkyleindc) is a senior fellow at the Tony Blair Institute for Global Change and holds a PhD in political science from Columbia University.