10 Geographic Information System(GIS) mapping of the data. Finally, each protest event in the dataset contains a ‘notes’ cell, providing a succinct description of the event and the claims that the protesters advanced. We used the ‘ notes’ to determine how to categorise each event by protest type. Our three protest category types include: food, energy(fuel, gas and electricity), or cost of living. Analysis We analysed the ACLED dataset using targeted keyword searches. The targeted keyword searches used the following terms:‘food’;‘food price’;‘essentials’;‘commodit*’;‘fuel’;‘fuel price’;‘gas price’;‘gas’;‘diesel’; ‘petrol*’;‘energy’;‘electric*’;‘power outage’;‘blackout’;‘cost of living’;‘living cost’;‘inflation’. 6 As noted, each protest event in the ACLED dataset contains a cell labelled‘notes,’ which provides information about the event pulled from local news sources. When one of our keywords appeared in an event ’ s notes cell, we read the description of the protest event to determine if the coding matched a food, energy or cost of living protest type. If we determined that a protest event was a positive match, we colour-coded the protest event entry to match the corresponding protest type. Protest events that contained one of our keywords but were not related to claim making about food, energy or rising cost of living were coded as non-events. For example, an event about petroleum workers protesting delayed payment of wages by their employer might show as a match in our keyword search, but it is not a relevant event in our study. An event about climate activists protesting fuel extraction would also be a non­event. Another non-event would be when a region such as Venezuela’s‘Vargas’ produced a keyword match, but the protest event was entirely unrelated to energy, food or cost of living. Accordingly, our research required close examination of each protest event that proved an initial keyword match to filter relevant events from non­events. Given the volume of relevant protests in our initial review, we employed a secondary audit of our initial protest type categorisations. The purpose of the audit was to catch events that had possibly been mis-coded as representing a specific protest type when they could more aptly fit into a different category. For the‘food’ and ‘energy’ sub-datasets, we ran a full second keyword search. If any of the protest events initially coded specifically as a food or energy event produced a match on any of our other keywords, indicating that the protest included multiple claims, they were recoded as general protests over the rising cost of living and moved to the cost of living sub-dataset. For example, a protest note could read,‘On 20 July 2022, members of TRS along with a Minister held a protest in Mahabubnagar city(Mahabubnagar district, Telangana) against the imposition of GST on milk products as well as rising food and fuel prices’(ACLED ID 9421769). If this event was initially categorised as a food protest, it would be recategorised as a cost of living protest, given that the protester grievances had been over both food and fuel. Limitations The ACLED data and our search strategy provide us with a dataset of protests triggered by grievances about access to and prices of food, fuel and electricity, as well as broader discontent about inflation and the rise in the cost of living that we are confident includes the types of events in which we were interested. However, our relatively strict inclusion criteria mean that the dataset errs on the side of caution, and we have deliberately 6 The use of* denotes the search term captured permutations of the base word. For example,‘petrol’ captured entries of ‘petrol’ and‘petroleum’.