How today's cost-of-living crisis fuels political turmoil
By Sam Wilkin, Director of Political Risk Analytics, WTW
In 2022, as many as three million person-days were devoted to cost-of-living protests in 122 countries and territories*. In the first quarter of 2023 alone, that figure increased by almost half again. March 2023, the latest full month for which data are available, saw the peak of such protests thus far, with the 10 countries hardest hit including Argentina, Costa Rica, Portugal, Germany, France and Greece.
As food prices soared following the escalation of conflict in Ukraine, many commentators predicted that “food riots” would strike the emerging world. That didn’t happen. In this report, we dispel some
myths about social unrest, explain the cost-of-living protests that have occurred thus far – including why wealthy countries have been hit so hard – and look at the countries most at risk in the year ahead.
Our research indicates that protest intensity in 2022 and the first quarter of 2023 correlate most strongly with rising energy prices, higher incomes, greater democratic freedoms, slower real wage growth, and the proportion of wages set by collective bargaining. It appears that wage protests – partly because wages set by collective bargaining have responded slowly to rising inflation – and protests regarding government policies on energy, are among the most important drivers of cost-of-living unrest globally.
On the one hand, this analysis suggests that cost-of-living protests are likely to be most intense in relatively wealthy countries. These wealthy countries are less likely to be severely destabilized by such protests, in contrast to the events of the Arab Spring or the fuel price protests that stuck Sri Lanka in 2022.
* WTW analysis of data from ACLED. Note that in this context “person-days” does not refer to a 24-hour period, but rather the participation of a person at a given protest or riot on a given calendar day. For further details see the technical appendix.
A global wave: cost-of-living protests mapped for Q1 2023, and a snapshot of active protests in April 2023
Source: WTW analysis of data from ACLED. Protest descriptions are quoted from CEIP. Size of the bubble represents cumulative number of person-days of protest at each location. “Cost of living protests” are identified via text search of event descriptions. For further details see the technical appendix.
On the other hand, the finding that intense cost-of-living protests are most likely in rich countries suggests that the trend towards escalating property damage and business interruption from social unrest is likely to continue, because greater exposure values are concentrated in these countries.
Moreover, this finding suggests that countries where collective bargaining is common, including many countries in Western Europe, are likely to be continually impacted by intense and economically-costly protests if inflation remains high.
Our report concludes with a set of risk indices. We find that countries including Argentina, France, Spain, and Italy appear most vulnerable to wage protests (which would continue the patterns of early 2023). That said, emerging market countries including Sierra Leone, Ghana and Zimbabwe also rate as potentially high risks, suggesting that in 2023 wage protests may become increasingly intense in the emerging world. We find that Lebanon, Egypt, Sri Lanka and Haiti may be among the most vulnerable to the less-likely but more
catastrophic scenario of policy protests that destabilize governments.
Since the WTW Political Risk Index was founded more than a decade ago, its strapline has been, “analyzing patterns in the world’s most vulnerable countries.” Recent editions of the Index have looked at new areas of political risk vulnerability, including climate, the impact of social media on politics, and how geopolitical divides are destabilizing countries caught between rival superpowers.
Amidst such new developments, there has been one constant: “the world’s most vulnerable countries” have been poor countries. That tends to be the pattern with political risk. The institutions that are important for sustained economic growth, such as checks and balances and the rule of law, also tend to reduce the likelihood of violent conflict and limit business risks such as arbitrary changes of regulation.
But today’s cost-of-living crisis has been an exception. Rising prices, particularly of food and energy, have been associated with turmoil in rich, poor and developing countries alike. In 2022, violent unrest over fuel prices struck Haiti, the poorest country in Latin America. Also in 2022, rioters toppled the government of Sri Lanka – until then a fast-growing emerging market – and the country defaulted on its sovereign debts.
In the first months of 2023, however, intense cost-of-living protests have struck hardest in the rich world, and in better-off emerging markets. In January through April, as many as 1.5 million person-days were expended on cost-of-living protests in the 10 countries hardest hit, including Germany, France, Portugal, Switzerland, and Greece - with the UK at eleventh place and Spain thirteenth. (See previous page).
Perhaps we should not be surprised by these new patterns of vulnerability. In our just-published political risk survey, a top 10 risk for 2023 was “social instability in rich countries.” Concerns about political risks associated with polarization and populism in Western Europe and North America are a few years old now. As a European automotive executive we interviewed for the survey put it, “the difference that we in the past used to look at between [political] risks in the OECD world and the non-OECD world is much less than it used to be.”
In the pages that follow, we will dispel some myths about social unrest, attempt to explain the cost-of-living upheavals that have occurred thus far, and assess the risks for 2023 and beyond.
We draw on two databases for this analysis. The first is the Armed Conflict Location & Event Data Project (hereafter, ACLED), which provides day-by-day, geocoded political event data drawing on media sources, and has compiled records of more than a million events already. The second is the Global Protest Tracker from the Carnegie Endowment for International Peace (hereafter, CEIP), which takes a less comprehensive but more curated approach, grouping individual events into protest movements, which are the main unit of analysis. We extend our thanks to both organizations for compiling such rich resources for political risk research.
Source: WTW analysis of data from ACLED. “Cost of living protests” are identified via text search of event descriptions. Note that in this context “person-days” does not refer to a 24-hour period, but rather the participation of a person at a given protest or riot on a given calendar day. For further details see the technical appendix.
A rising tide: estimated person-days dedicated to cost-of-living protests, March 2022 through March 2023
Consider a traditional view of cost-of-living protests: a society’s poorest people, perhaps on the verge of starvation, roam the streets looking for food, which has become increasingly unaffordable. Their neighbours see them, join in, and the crowd builds. The police or even army may intervene, but these people are desperate, and soon the crowd turns violent, producing a “food riot.”
That image of starving people taking to the streets was invoked as global food prices rose in the first half of 2022. Many commentators predicted that social unrest would result. For instance:
“When people can’t feed their families… it’s an almost inevitable recipe for large-scale civil unrest” - Financial Newswire
“Food is a key element of any community’s culture. When that element is removed civil unrest can occur” - Business Insider
* The classic text on this subject is Tilly, C. and Tarrow , S. 2016. Contentious Politics: Second Edition.
Is that what happened? No. With the benefit of hindsight, perhaps surprisingly, there was no significant correlation between food price inflation in 2022 and protest intensity (although, to be sure, some specific protests took place regarding food). Nor was there a positive correlation between poverty and protest intensity. Quite the opposite: in 2022, the richer people were, on a country-average basis, the more likely they were to take to the streets (in the section ‘What happened in 2022?’ we will look at such correlations in more detail).
Moreover, if we think of recent major protests, they do not fit the image of the desperate poor swarming onto the streets. Take, for instance, the protests that occurred in France in January 2023 regarding the government’s efforts to raise the retirement age by two years. According to the CEIP database, these protests involved crowd sizes exceeding a million people and, as of this writing, had lasted at least four months.
What is it that makes French workers so willing to protest? Their extreme anger? Extreme poverty? Extreme desperation? None of these potential explanatory factors appears to find the mark. Indeed, while one can empathize with the desire to retire with full pension at 62, by global standards, French workers appear to be relatively privileged.
Therein lies a crucial clue to the correlates of mass political action. As a general rule, protests are an expression of political power. From French workers to #blacklivesmatter to #metoo, protests tend to occur when social groups have the capacity to organize to press their political agenda by contentious means – and believe that taking to the streets is an effective method to achieve their political objectives.*
The traditional view of protest imagines actions that are driven by personal rage or desperate circumstances, occur largely spontaneously, and are the last resort of those without other alternatives. Social science presents a different picture. Mass protests tend not to be spontaneous but rather organized; they tend to be joined not necessarily by those who are poor or desperate but rather those who believe their protest can be effective; and they tend to be driven not by personal anger or the “madness of crowds,” but rather by the political agendas of social groups.*
In 2022, there was indeed a global food crisis, as commentators had predicted. This crisis, which continues as of this writing, put hundreds of millions at risk of starvation or malnourishment. For better or worse, however, the risks associated with this food crisis were mainly humanitarian, rather than riots.
That said, there was extensive mass unrest in 2022 – indeed, one study identified 12,500 protest incidents involving food or energy prices occurring in 148 countries in the year to 31 October 2022. The current research differs from that study, both in time period and approach. Our analysis suggests as many as three million person-days were devoted to cost-of-living protests in 122 countries and territories in 2022, and adding in the first quarter of 2023, up to four million person-days.
If not “food riots,” then, what happened? To understand the sources of this extensive unrest, we need a more realistic view of the causes of protest. That is the subject to which we turn in the next section.
* For a review, see van Stekelenburg, J. and Klandermans, B., 2013. “The Social Psychology of Protest.” Current Sociology Review.
If “food riots” do not form the main link between rising living costs and protest, what are the links? We here outline three scenarios based on historical cases, which are not mutually exclusive. In the section that follows, we will test these scenarios against the protest event databases for 2022.
The government enacts an unpopular policy. People take to the streets in protest, with the intention of having the policy reversed. This relatively undramatic scenario has played a role in some of the most extraordinary incidents of mass unrest in recent history, including the Chilean riots of 2019-2020, which began in response to increases in subway fares, and the Gilets Jaunes protests in France, which started following an increase in fuel taxes.
Indeed, the “policy protest” scenario has likely contributed to the idea of the “food riot,” because governments often regulate the prices of staple goods. For instance, many protests that occurred at the time of the French Revolution occurred because the French government had attempted to deregulate the price of food (leading to sharp increases in prices). This policy change resulted in “bread riots,” which contributed to the revolutionary atmosphere at the time – and which, hundreds of years later, have been reimagined as unrest fomented by desperate peasants searching for food.
This historical case aside, most policy protests differ very obviously from food riots. When the news agency Reuters researched riots relating to the food and fuel price shock of 2022, they found a picture starkly at odds to the image of people rioting because they were unable to feed their families. For instance:
Tunisian farmers protested to force the government to raise state-controlled food prices
Thousands of students marched through Santiago demanding higher food stipends
Thousands of farmers protested in Buenos Aires against government policies to contain food prices
One person was killed in Guinea’s capital during protests over hikes in government fuel price
Price protests in Iran were sparked by the government’s fuel subsidy cut decision
Indonesian farmers protested in Jakarta against the government’s palm oil export ban
These protests were not instigated by heads of households unable to obtain bread; they were usually sparked by special interest groups that had been adversely impacted by specific government policies. In many cases (such as Chilean university students), these groups were relatively prosperous. In several cases, the protesters (including Tunisian and Argentine farmers) wanted food prices to be increased.
Movements in the global price of food and fuel are particularly dangerous as a source of policy protests because in many emerging market countries staple goods are subsidized, or prices are regulated, by the government. A comprehensive 2018 study found that, at that time, more than 70% of emerging market countries regulated the prices of food and beverages, and more than 80% imposed price controls on energy products.
One of the most extreme cases is Egypt. Egypt maintains a generous subsidy regime for bread, providing food subsidies to nearly 70% of the population. Egypt is also one of the world’s largest importers of wheat. The combination of subsidies and import dependence can be dangerous. Food and fuel subsidies have historically consumed about 10% of the Egyptian government budget. When global commodity prices are high, as they were on the eve of the Arab Spring, that figure can rise to 40% – threatening to bankrupt the country.
Egyptian governments, however, know that they cut the program at their peril. Protests against efforts to cut the bread subsidy contributed to the Arab Spring uprising (see graph). In the late 1970s, when global food prices were high, an attempt by the Egyptian government to cut the subsidy provoked the so-called “bread intifada” – three days of rioting in which some 800 people died in clashes with the police – until the Egyptian government backed down and restored the subsidy program (ironically, another case that has often been confused with a “food riot”).
Policy riots, not food riots: global food prices and the Arab Spring
SOURCE: https://www.fao.org/worldfoodsituation/foodpricesindex/en/
In recent decades, the circumstances of the average American worker have, in some respects, not been enviable. Wages for US workers tend to be set by the market, and various market forces – from technological advances to international competition – have arguably tended to push wages down even as worker productivity has increased.1
In Europe, by contrast, most wages are set by collective bargaining arrangements, in many cases involving negotiations with labor unions. For European countries including Austria, Belgium, France and Italy, more than 95% of wages are set via collective bargaining (against only 12% in the US). In recent decades, the political power of organized European workers may have tended to keep real wages and labor’s share of national income high, at least in comparison to the United States (see graph).2
1 https://www.epi.org/productivity-pay-gap/. For an alternate view see https://www.aei.org/articles/the-productivity-pay-gap-a-pernicious-economic-myth/2 Regarding real wage growth; regarding the labor share of income; literature review of possible alternative explanations
At least, that was the story historically. In 2022 and 2023, demand for workers in the US, particularly in services, surged. Many US workers, their wages dictated by market forces, saw their pay rise accordingly – indeed, since July of 2022, US wage growth has been at a 25-year high. In Europe, by contrast, political negotiations over salaries can take time. As of March 2023, real compensation per hour in the eurozone had fallen by 7% compared with 2021.
That dynamic is at the heart of our second scenario: the wage protest. If workers whose wages are set by collective bargaining want their wages to rise quickly, they may decide that public protests are their best strategy. That scenario might be particularly likely to occur when prices are rising, and when the government sets wages directly, for instance for public sector employees, such as UK doctors, nurses, and railway workers – who have recently been on strike, sometimes drawing on broader protests in support.
The wages of bargaining? Annualized changes in labor’s share of national income, US vs. selected European countries, 1995-2013
SOURCE: https://www.oecd.org/economy/outlook/Decoupling-of-wages-from-productivity-november-2018-OECD-economic-outlook-chapter.pdf; the European countries selected have more than 90% of wages set by collective bargaining
For this final scenario, let us for the moment leave the world of the protester – although we will quickly return – and consider the decisions of voters. It is conventional wisdom that voters will sometimes punish a government that presides over an economic downturn. In the United States, after President George W. Bush went down to defeat after a single term, commentators were quick to credit a phrase coined by an opposition campaign strategist: “it’s the economy, stupid.”
This phenomenon of “economic voting” (when election results correlate with a country’s economic performance) is sometimes known as the “punishment vote,” because it targets the incumbent (currently serving) government. But why does it occur? Consider the following table of possible voter motivations.
* Lewis-Beck, M. S. & Stegmaier, M., 2013. The VP-function revisited: a survey of the literature on vote and popularity functions after over 40 years.
Here the views of the general public and political scientists tend to diverge. Most people tend to imagine that voters vote their pocketbooks – number (3) in the table. Obviously, if voters cast their ballot based on their personal economic circumstances, then during a recession, a large number of voters are likely to have poor personal finances, thus producing a “punishment vote.”
Political scientists, by contrast, would hope that voters would consider the relative merits of both the serving government and the opposition, consider the economic policies of each candidate, and cast the vote they believe would deliver the best outcome – number (2) in the table. That is a lot to ask of a voter; but such voting strategies would be necessary to deliver the best possible government.
The truth appears to be in the middle: voters do not vote their pocketbooks; but nor do they appear to consider the future. Social science research has found that economic voting tends to be a “punishment vote” phenomenon, considering only the performance of the economy under incumbent government (and not, unfortunately, the future capabilities of either the government or the opposition). Research also finds that voters appear to make little effort to distinguish fault: if the economy turns down for reasons entirely unrelated to the government’s policies, the government is still punished.
On the other hand, voters do not simply vote their pocketbooks – they appear to make an effort to understand the situation in the national economy, and vote on that basis. Remarkably, even voters who do not read newspapers appear, in aggregate, to have a good idea of how the economy is doing. Hence when the economy turns down, inflaton rises, or unemployment surges, a punishment vote may be in store.*
Sometimes, however, voters may not have an opportunity to enact punishment. For instance, elections may not be free or fair; or may not be held at all. Or, in some cases, voters may be impatient. They may not want to wait for the next election to see the president step down. In prime ministerial systems, they may hope, via mass protest, to force an early election.
How might punishment protests start? Presumably, as with a punishment vote, poor economic performance (including high inflation) could be a driver. But policy protests can surely play a role as well. For instance, the famous cases of the Chilean subway fare riots or France’s Gilet Jaunes started as policy protests and appeared to evolve into more general punishment protests against the sitting government.
In the Arab Spring as well, scholars found evidence that policy protests had evolved into punishment protests – and indeed that the dividing lines between the two scenarios were often unclear. One study of Arab Spring unrest, for instance, found that “a change in food access motivated protest and violence involving existing grievances rather than explicitly addressing food access. In this way, food changed the meaning and severity of existing grievances.”
Protest intensity: estimated proportion of the population engaged in cost-of-living protests in 2022 and early 2023
SOURCE: WTW analysis of data from ACLED. Protest descriptions are quoted from CEIP. “Cost of living protests” are identified via text search of descriptions. “Protest intensity” is calculated as the number of person-days of protest divided by population. Protests for which no participant data are available are assumed to involve 100 people. For further details see the technical appendix.
Perhaps the most surprising case of mass unrest in 2022 and early 2023 followed a similar pattern to the Arab Spring: policy protests (scenario 1) turned into a punishment protest against a serving government (scenario 3).
In Sri Lanka, protests over a government decision to raise fuel prices eventually spiraled into forceful anti-government unrest, causing the president to flee the country. These developments, and Sri Lanka’s subsequent debt default, were particularly striking because the country had been a development success story, enjoying extremely rapid growth over three decades, rising from a per capita income of roughly $2,000 per person in 1990 (on par with Afghanistan or Zimbabwe today) to more than $13,000 per person in 2019 (on part with Mexico or Serbia).
There are also numerous less-extreme examples of policy protests in the country profiles in this Index. Iran, already experiencing an intense year of unrest, saw protest after fuel subsidies were cut. India saw protests against inadequate food subsidies. Cameroon saw protests after fuel subsidies were reduced, as did, more dramatically, Haiti. Protests relating to fuel prices, nearly a tenth of them violent, engulfed Ecuador in June 2022. Jordan experienced demonstrations after the IMF demanded subsidy cuts (as did countries not covered in this Index, including Lebanon).
But our statistical analysis of correlates of cost-of-living protests suggests that focusing only on these policy protest cases from the Index could be misleading. Rising energy prices did correlate strongly with protest intensity, at least for the ACLED database, suggesting that policy protests (scenario 1) played an important role in recent turmoil.
Across both databases, however, many other correlates related to people power factors and to labor markets, especially real wages and the proportion of the workforce subject to collective bargaining arrangements (see graph), suggesting that wage protests (scenario 2) have also been an important driver.
Including multiple variables in the analysis reinforces this picture: the best-fitting models contain indicators of rising energy prices, declining real wages, and pay deals set by collective bargaining.* As noted in the previous section, the prevalence of collective bargaining may have helped make worker pay slower to respond to rising inflation, and incentivized workers to resort to street protests to accelerate the process.
The best model based on ACLED data contains energy price inflation, greater democratic freedoms, and collective bargaining. The best model based on CEIP data contains only two factors: real wage growth and the proportion of the workforce with collective bargaining pay arrangements. Note that due to limitations on the availability of real wage data for 2022, this latter model applies mainly to advanced economies and wealthier emerging markets, and has a poor statistical fit – for further details see the technical appendix.
* The best-fitting models were identified by starting with the most highly correlated independent variable, and then adding variables that correlated with residuals. For further details, see the technical appendix.
Power and dissatisfaction: correlates of cost-of-living protest intensity in 2022
SOURCE: WTW analysis of protest data from ACLED and CEIP, correlated with indicators from World Bank, ILO, IMF, IEA, and Freedom House. For details see the technical appendix.
ACLED database
CEIP database
This analysis suggests that wage protests (scenario 2) were a major driver of the cost-of-living protest surge in 2022 and early 2023. Where wages are set by collective bargaining rather than market forces, wages will tend to be slower to respond to broader inflation, sapping workers’ purchasing power. That phenomenon is dangerous for the economy as a whole; it also appears to have triggered wage protests. Last year, the political power of workers in Europe (and other regions where collective bargaining is common, including some parts of Latin America) became something of an Achilles’ heel.
This impression of wage protests as a driving force in 2022 and 2023 is reinforced by an analysis of the frequency of word appearance in the databases to event descriptions. The most frequently-appearing words, particularly in the ACLED data, are those that would seem to relate to wage protests (including “workers,” “members [presumably, union members],” “prices,” “demand,” and “inflation”) – although labor organizations can also play key roles in fomenting policy protests.
Actors and issues: word cloud of cost-of-living protest descriptions worldwide. NOTE: ACLED LEFT; CEIP RIGHT
SOURCE: WTW analysis of protest and riot description data from ACLED and of protest movement trigger and motivation data from CEIP.
Knowing that mass protest represents the exercise of political power (in contrast to the image of the “food riot”), we should not be surprised to find that “people power” indicators also correlate strongly with the relative intensity of cost-of-living protests. Countries that were more democratic tended to experience more protests; countries that were richer tended to experience more protests; and countries with higher levels of literacy tended to experience more protests.
In general, such indicators likely reflect the public’s ability to take political action. People who are wealthier are more likely to be able to take days off to join protest events. People who are literate and have internet access are likely to be more able to organize strikes and demonstrations, and to understand their country’s political system and how public pressure might best be applied to achieve their political goals.
It appears that the combination of a greater reliance on collective bargaining, and higher levels of people power, helped to make the cost-of-living protest a rich-world phenomenon. When inflation rates are persistently high – a situation most countries in Western Europe have not faced in some decades – these pay arrangements can have worrisome consequences.
It should also be noted that energy price inflation (and energy import dependence) correlate significantly with protest intensity, while food price inflation (and food import dependence) does not (although the country profiles in this Index do point to several protests over food subsidies, for instance in India). In the word clouds, words relating to energy prices also appeared more frequently than food prices. This finding mirrors a 2022 study which found that energy price protests were far more common than food protests.
There are likely two reasons why energy prices appeared to be more important in 2022 and early 2023. The first reason has to do with people power. While food makes up a significant portion of the income of the poor, spending on fuel often rises for the relatively rich – particularly those with personal automobiles. Hence higher fuel prices are more likely to be a concern with better-off social groups, who may be better able to organize protests.
A second likely reason energy prices were a more important protest driver is that energy prices were more volatile. Below, the graph of food prices and the Arab Spring from the previous section is reproduced, but with world energy prices added to the chart – and volatility in energy prices was far more extreme, particularly over the past year.
Were there punishment votes, or punishment protests, in 2022 and early 2023? Quite possibly. While elections cannot be boiled down to single factors, economic conditions were said to have contributed to the ouster of incumbent governments in Nigeria and Kenya, and the Philippines and Colombia also saw dramatic changes of power.
Many countries where inflation was high also saw dramatic anti-government protests, although these were often focused on political factors, such as the ouster of President Pedro Castillo in Peru, proposed limits to regional autonomy in Uzbekistan, measures taken against former Prime Minister Imran Khan in Pakistan, measures seen to limit democracy in Central African Republic, Chad and Azerbaijan, and restrictions on women in Iran.
For further analyst assessments of these countries and other cases, see the country profiles in this Index.
Worse for fuel than food: global food prices vs global energy prices, 2000 to 2023
SOURCE: FAO and IMF. For the food price series, 100 is the 2014-2016 average. For the energy price series, 100 is the 2016 average.
Looking at the graph of world food and fuel prices, one might expect the cost-of-living crisis to have abated. Global fuel prices have fallen almost by half from their peak; food prices are down as well. Two factors, however, appear to have extended the crisis and related protests. First, especially in the emerging world, local currency weakness and the strength of the US dollar have created a “double burden” of elevated global prices and rising costs of imports. Second, generalized inflation has proved to be stickier than expected in the US and Europe.
Of course, not all countries covered in this Index were struck by cost-of-living protests. Some countries cut subsidies or raised state-controlled prices without triggering extensive unrest, including Ethiopia. The government of Bangladesh, with one eye on events in Sri Lanka, approached the IMF early for support; the government of Uganda, noting the cost-of-living riots in Kenya, proactively began messaging about rising prices.
Many other countries took advantage of strong fiscal positions to protect consumers. Saudi Arabia and Qatar poured money into domestic subsidy programs and kept inflation low. Other oil exporters, including Iraq and Nigeria, did the same, although at greater risk to government finances. Russia provided such extensive new support programs, particularly for soldiers’ families, that poverty rates appear to have declined.
But in many cases, the experts who compiled the country profiles in this Index worry that the show cannot go on. The government of Egypt, having learned the lessons of history, made great efforts to avoid cutting subsidies – and now appears to be missing targets agreed with the IMF. There are rumors of subsidy cuts in Laos. In Vietnam and South Africa, energy subsidies may be pushing the state-run power utilities into distress. The government of Iran has failed to publish inflation numbers in recent months, presumably fearing a political backlash.
Many other countries have been forced into international bailouts where subsidy cuts may eventually be required, including Republic of Congo, Ghana, and Angola. In Nigeria and Morocco, large subsidy programs are due to be phased out. Some of the bills for such subsidy programs are eye-watering: Turkey and Malaysia are estimated to have spent almost 2% of GDP controlling energy prices.
What countries might be at risk if the cost-of-living crises continue? We have here compiled two country indices – one for scenario 1 (policy protests) and one for scenario 2 (wage protests). We show the top 20 countries (countries at greatest risk) for each index.
The index for policy protests looks at the classic scenario of severe policy protests that countries including Sri Lanka in 2022 have experienced. Beginning on the left-hand side of the index are indicators of the level of government subsidy or control of food and energy prices. These policies can protect consumers, but can be dangerous if (moving to the right in the index) a country is dependent on imports and either its exchange rate falls or global prices surge.
In that case, the cost of subsidies will increase and – moving farther to the right in the index – if the government’s fiscal position is weak, these subsidies may become unaffordable and prices allowed to rise. If the public has sufficient people power (on the far right-hand side of the index), a scenario in which protests against these cuts eventually topple a government can result. Because we are concerned about the rare but dangerous case in which policy protests and punishment protests combine, we do not include indicators of democracy in the people power measures (even though those performed strongly in our statistical tests).
Most of the countries that top this index are those that have experienced severe unrest already, including Lebanon, Sri Lanka, Haiti, Iran, Sierra Leone, Kenya and Pakistan. The delicate predicament of countries in bailouts, including Egypt, is highlighted. There are also some unexpected inclusions, notably Tajikistan, which according to our analysis of the ACLED database has not yet experienced any cost-of-living protests, and Georgia, which is enjoying something of an economic boom due to migration from Russia.
SOURCES: WTW calculations based on various indictors; for details, see technical appendix
Policy Protest Risk Index for 2023
The second index looks at scenario two, the wage protest. This index is dominated by countries from the developed world. Starting at the left hand side of the index, the indicators reflect a scenario in which generalized inflation is high, or particularly inflation in fuel prices. These rising prices undermine living standards.
In countries where pay arrangements are set via collective bargaining – moving to the right in the index – workers’ compensation may be slow to adjust, even in the face of generalized inflation. Governments with strong fiscal positions may be able to alleviate some of the pressure, via consumer subsidies or reduction in taxes (the EU spent some $350 billion on such subsidies in 2022). If the government is unable to act, in countries where people power is great enough and the right to protest is guaranteed (on the right-hand side of the index), labor organizations and sympathetic groups may stage protests to raise pay more quickly.
The top countries in this index include many of the countries where protests have recently been most intense, including Argentina, France, Spain, and Italy. Sierra Leone has the unfortunate distinction of appearing in both indices, and indeed has become a case study in price protests (despite relatively low levels of people power, and an exceptionally violent government response). The top of the index also includes some surprises, such as Iceland, which according to the ACLED database has yet to experience any cost-of-living protests. Perhaps worryingly, several developing countries, such as Ghana, Sudan and Zimbabwe, appear among the top 20, suggesting that in 2023 the burden of cost-of-living protests may become increasingly severe in the emerging world.
Wage Protest Risk Index for 2023
As of this writing, many countries around the world continue to grapple with cost-of-living crises of unexpected scale and duration. The research published here highlights another reason (in addition to direct economic costs) why it is so vital for governments to get a grip on inflation. High inflation can sap the purchasing power of consumers; it can also have indirect costs, by triggering protests.
And such protests are expensive. Major protest events have been found to reduce economic output by about a percentage point and stock market returns by 1.4% (equivalent to wiping about $500 billion off US equity market capitalization). France’s gilets jaunes protests specifically caused an estimated $220 million in outright property damage and an estimated $4.8 billion in economic losses.
In advanced economies and wealthier emerging markets, where recent cost-of-living protests have been most intense, the damage tends to be greater because greater value is concentrated in those locations – especially in the urban areas where protests tend to occur. Especially if inflation continues to be elevated, cost-of-living protests, and social unrest more broadly, are likely to pose severe challenges to companies and the insurance sector in the years ahead.
For both the ACLED and CEIP databases, the target of our analysis is a measure of “protest intensity,” which is the number of people involved in cost-of-living protests divided by the population (for CEIP), or the number of person-days dedicated to protest divided by the population (for ACLED). (Note that in this context, a “person-day” refers not to a 24-hour period, but the participation of a single person at a given protest or riot on a given day.)
Essentially, the idea is to capture the portion of the population engaged in protesting over 2022 and early 2023. For ACLED, the period of analysis is April 2022 to March 2023. For CEIP, the period is the entirety of 2022 and January to April of 2023.
Cost-of-living protests were identified using text searches of protest notes and descriptions. For both ACLED and CEIP the same keywords were used (“food,” “bread,” “energy,” “fuel,” “prices,” and so on). We attempted to exclude climate protests by searching the keywords “climate” and “global warming.” This approach differs somewhat from earlier research which attempted to exclude wage actions by labor unions.
We then measured the correlations of protest intensity to a number of potential explanatory variables. To create our multivariate models, we first used the most highly correlated variable, and then selected the variable that was most correlated with the residuals until the adjusted goodness-of-fit measure began to decline.
Outliers were a significant issue, particularly for the dependent variable. We truncated both the independent and dependent variables to one standard deviation from the mean, but outliers may still influence the results.
The candidate explanatory variables were:
Import dependence (imports of food or fuel as a percentage of consumption). The sources were http://data.un.org/Data.aspx?q=dependency+ratio&d=FAO&f=itemCode%3A21035 and https://databank.worldbank.org/source/world-development-indicators
Consumer price inflation, Sourced from https://www.imf.org/en/Publications/WEO/Issues/2023/04/11/world-economic-outlook-april-2023, with some gaps filled using https://www.oecd.org/economic-outlook/march-2023/
Real wage growth, sourced from https://www.ilo.org/digitalguides/en-gb/story/globalwagereport2022-23
Share of wages subject to collective bargaining, sourced from https://ilostat.ilo.org/topics/collective-bargaining/ For index construction, missing values were imputed using a regression based on unionization rates and public sector employment, from ILO and the World Bank respectively
Energy price inflation, sourced from https://www.worldbank.org/en/research/brief/inflation-database. For 2023 figures, we compared midyear 2022 prices with April 2023 prices from https://www.globalpetrolprices.com/gasoline_prices/
Unionization rates, sourced from https://ilostat.ilo.org/topics/union-membership/
Social trust, sourced from https://www.worldvaluessurvey.org/wvs.jsp
Mobile phone penetration, sourced from https://databank.worldbank.org/source/world-development-indicators
Urbanization rates, sourced from https://databank.worldbank.org/source/world-development-indicators
Literacy rates, sourced from https://databank.worldbank.org/source/world-development-indicators
Income per capita, sourced from https://databank.worldbank.org/source/world-development-indicators
Proportion of households with access to the internet, sourced from https://databank.worldbank.org/source/world-development-indicators
Democratic freedoms, sourced from https://freedomhouse.org/. Measures of political rights were used. Measures of democratic backsliding were tested but were not significantly correlated.
Food and fuel subsidies. Here we relied on indirect indicators: the price of bread at the store, and the price of gasoline at the pump, on a U.S. dollar basis. Countries that have very low prices, by international standards, are likely to be those that subsidize (especially if those countries import these products). The sources were https://databank.worldbank.org/source/icp-2017 and https://www.globalpetrolprices.com/
Government debt as a percentage of GDP. Sourced from https://www.imf.org/en/Publications/WEO/Issues/2023/04/11/world-economic-outlook-april-2023
Budget deficit as a percentage of GDP. Sourced from https://www.imf.org/en/Publications/WEO/Issues/2023/04/11/world-economic-outlook-april-2023
Sovereign risk rating. The source was https://credendo.com/en/country-risk. The medium/long-term risk rating was used.
Debt interest burden (interest payments as a share of government revenues). The source was https://databank.worldbank.org/source/world-development-indicators
All “people power” factors tested correlated very strongly with level of development (0.7 or greater). Hence in the final ACLED model “people power” was represented solely by democratic rights. Regression results are presented below.
For both databases, we preferred to use measures of the population engaged in protests rather than incident counts. The difficulty with this approach is that the number of people engaged in protests is often unreported (or may be misreported for political reasons). Incidents with unreported participant counts were excluded from our regression analysis.
Additionally, protest size reporting may be better in advanced economies, so one may find an exaggerated significance for level-of-development variables. However, aggregated incident counts had the same significant correlations as our protest intensity variable (correlating with both energy price inflation and “people power” factors), which gave us confidence to proceed. (Although incident counts correlated more strongly with the proportion of the workforce in the public sector than the proportion of wages set by collective bargaining.) In addition, the performance of the multivariate models gave us confidence that that the significance of variables was not based solely on this level-of-development phenomenon.
For the ACLED database, we attempted to integrate incidents without participant counts by assigning various standard numbers to these incidents (e.g., 10, 100, 1000). For the CEIP database, we attempted to create a measure of person-days spent on protests using protest duration, and to integrate considerations of protest outcomes (a successful protest might end more quickly) into the results.
Both approaches produced interesting initial findings – specifically, both measures were correlated with food price inflation and general inflation, an expected result that was not apparent in other data. In addition, a measure of intensity for all protests in the ACLED database (not just cost-of-living protests) correlated with general inflation, and a measure of intensity of violent protests correlated with food price inflation. We left these interesting results for further research.
