Navigation – Plan du site

Can big data revolutionise development economics? A literature review on Latin America

¿Pueden el big data revolucionar las políticas del desarrollo? Una estudio critico sobre América Latina
Le big data peut-il révolutionner l’économie du développement ? Une analyse critique des publications scientifiques sur l’Amérique latine
Tove Forsbacka Karlsson, Paula Höhne Tarragona et Milindee Jeewandara
p. 155-175


Le domaine du développement a été freiné par un manque général de données. L’émergence récente du big data a attiré l’attention des chercheurs et des représentants du gouvernement, qui se sont efforcés de concevoir de meilleurs programmes de développement en exploitant la quantité croissante de données. Dans cet article, nous analysons le potentiel du big data pour répondre aux défis du développement en Amérique latine. Nous passons en revue la littérature actuelle et les programmes de développement existants pour montrer comment le big data a été utilisé pour trouver des solutions innovantes dans des secteurs tels que la santé, l’agriculture et l’éducation. Le big data a un énorme potentiel pour améliorer les statistiques et notre compréhension des problèmes rencontrés par les pays d’Amérique latine. Si les données sont partagées et compilées de manière appropriée, le big data peut éclairer les politiques et améliorer la manière dont les programmes de développement sont construits. Cependant, plusieurs préoccupations ont été soulevées concernant la représentation erronée des populations et le manque de fiabilité des données, et le potentiel des données massives dépend de leur utilisation responsable.

Haut de page

Notes de la rédaction

Texte reçu le 12 septembre 2018, accepté le 30 novembre 2018.

Texte intégral

We want to express our gratitude to Nadja Friedl for her valuable contributions to an earlier version of this paper.


1Every day, a huge volume of data is created through the interactions of billions of people using computers, GPS devices, and cell phones, as well as through the passive observation of satellites and sensors [World Economic Forum, 2012]. As the digital revolution has taken hold of the world, it is also reaching low-income populations. Governments, organisations and researchers alike are investigating the potential of the surge in data coming from poor nations for the field of development economics, where scarcity of data is a major problem. Big data, as it is called, is used to identify needs, provide services, and predict and prevent crises in low-income countries [World Economic Forum, 2012]. Just as the surge in available information has radically transformed the fields of marketing and healthcare in the developed world, it is possible that the availability of big data from developing countries will completely change development economics.

2In this paper we focus on Latin America, where data scarcity in urban slums and large informal sectors inhibits efficient and targeted public policy. This paper investigates current applications of big data in development policies in Latin America, and examines whether big data has the potential to solve challenges in development economics in the region. We review the current literature and empirical evidence of existing programmes to discuss the potential and the limitations of big data. In addition, we analyse whether big data could improve impact evaluations, which are an economic method used to analyse the impact of programmes and policies.

3This paper is structured as follows: first we identify Latin America’s main data-related challenges in development economics and introduce the concept of big data and its common applications. After that, we provides a literature review of studies where big data has been used to address development challenges in Latin America. This is followe by the potential of big data in impact evaluation is examined. We then analyse common concerns about big data. Finally, we conclude with a discussion on policy implications and governance issues.

Development challenges and data concerns

4Over the past decades, growing inequality and rapid urbanisation have been the two most prominent socioeconomic trends in Latin America. Latin America is today the most unequal region in the world. The richest 10 percent of the population control over 70 percent of the wealth [Bárcena Ibarra and Byanyima, 2016]. In 2015, poverty rates were estimated to 30 percent, and more than 20 percent of youths were unemployed [Hewco, 2017]. Since 1950, Latin America has experienced an youth unparalleled rate of urbanisation. Today, 80 percent of the population lives in cities, making Latin America the most urbanised region in the world [Paranagua, 2012]. Urbanisation can be both an opportunity and a threat: it is crucial for development, as cities enhance efficiency and productivity, but unorganised urbanisation risks resulting in widespread slums and poverty. In 2012, 24 percent of the urban population lived in slums, lacking access to housing and public services. Almost half of the Latin American population was employed in the informal sector [European Parliament, 2016]. Informal sectors make tax collection difficult and tax revenues generate only 17 percent of GDP [Corbacho, Fretes Cibils & Lora, 2013]. Large informal sectors also make governance and city planning less efficient, and people employed in the informal sector lack protection through social and labour regulations.

5These trends are reflected in several sectors. High urbanisation now means that human and financial resources for healthcare are highly concentrated in urban areas, while rural and remote areas often struggle with inadequate access to resources. Many countries in the region struggle with the spread of infectious diseases and remote areas in tropical countries are often disproportionately plagued by so-called neglected diseases, constraining these area ability to develop [Franco-Paredes et al., 2007]. In education, the income gap further translates into a secondary education gap with dropout rates higher for children living in poverty or in remote areas, and for indigenous children. The key development challenge in the financial sector is unequal financial inclusion, with the highest inequalities in Central America. In Nicaragua only a third of the population has a bank account [World Bank, 2017]. Common reasons for underdeveloped financial inclusion are a lack of trust in financial institutions arising from a weak rule of law or highly unstable financial systems [Rojas-Suarez, 2016]. In the agriculture sector, there are also large discrepancies in the access to financing for agricultural activities, and a relatively large number of cases where price setting and inflationary issues are key problems for those people earning a living from agriculture. In particular, the threat of climate change disproportionately affects farmers and the rural poor [Janvry & Sadulet, 2008].

6Governments, organisations and researchers have tried to address these challenges and lift people out of poverty through targeted policies and development programmes. These interventions are highly dependent on data and statistics to ensure that the right people are targeted, resources are effectively used and progress is made. Development economics has been held back by a general lack and low quality of data [Easterly, 2009]. Data challenges are more prevalent in developing countries, where there is a lower degree of price equalization and economic integration, and the infrastructure of the official macroeconomic statistic is weaker [Henderson, Storeygard, and Weil, 2012]. Traditionally, the two main sources of data have been national accounts and household surveys [Deaton, 2005]. Although most Latin American countries have long traditions of conducting censuses and surveys, and have relatively good national statistical institutions, many issues remain. Budget cuts and poor long-term planning due to political instabilities often result in fallouts of planned data collections. Vast income gaps and almost one fourth of the population living in urban slums [Magalhães, 2016] means that the quality of national statistics coverage is poor.

7The large size of the informal sector in Latin America is also of particular importance to the quality of data available for studies in development economics. Socioeconomic data is not accurate when much economic activity happens outside what is captured by national statistics [Soto et al., 2011]. For instance, Henderson, Storeygard and Weil [2012] point out that official GDP measures often understate the true output growth due to the uncaptured growth of informal sectors. Attempts to estimate the size of the informal sectors also often yield very different results, and there is little consensus about what methodologies should be used for such estimates. By using new data approaches [Ghosh et al., 2009] find that the actual size of the informal sector in Mexico was 150 percent larger than official estimates indicated.

8While the exceptional increase in the use of independent household surveys over the past decades has advanced our insights into the living condition of the world’s poor [Deaton, 2005], surveying is both costly and time-consuming. There is also an issue of comparability of household surveys across countries [Smith and Wills, 2016]. In Latin America, researchers have found difficulties with survey data collection across all classes: upper income households often live in gated communities that are hard to reach, and collection in the lowest income households and slums is often hindered by security issues [Parker, 2015].

9Some of the most central measures in development economics are particularly difficult to estimate and to find reliable data on. There is, for example, a long-standing discussion in economics on how to measure poverty. The conventional methods have serious drawbacks. Income is difficult to measure since households in poor areas tend to diversify their activities and usually have several sources of income. Consumption data is often noisy and asset-based data is unsuitable as a short-term measure of poverty. Households rank quite differently depending on which approach is used [Steele, 2017]. On a macroeconomic level, GDP data suffers from reliability issues in many developing countries and is rarely available at subnational levels [Henderson, Storeygard, and Weil, 2012]. In regions, such as Latin America, where inequality is high across various indicators, a lack of sufficiently disaggregated statistics means that tracking the progress of individual groups or communities is difficult [Data-Pop Alliance, 2016].

10A much-discussed way to tackle data challenges in development economics is to leverage big data. Big data seems like the latest buzzword of the moment, discussed by policymakers, businesses and researchers alike, all trying to exploit the benefits of big data. With the rapid digitization of society and the ever-growing role of electronic devices and storage, people’s actions, behaviours and preferences have a digital footprint. Big data is used to broadly denote this large volume of data—both implicitly and explicitly collected—that is created daily. However, when used correctly, this information is incredibly powerful for all users as it offers an almost unparalleled ability to observe the activities of individuals, groups and nations in real-time. It can enhance productivity in the public sector, increase the competitiveness of businesses and create substantial consumer surpluses [McKinsey & Company, 2011]. Technology research company Gartner [2017] offers a formal definition of big data as “high-volume, high-velocity and/or high-variety information assets that demand cost-effective forms of information processing that enable enhanced insight, decision-making, and process automation”. Compared to traditional datasets, big data essentially has the qualities of being of a higher volume, of a wider variety and of enabling users to make quicker decisions [Kshetri, 2014]. Boyd and Crawford [2012] further clarify the constitution of big data, arguing that big data “is less about data that is big than it is about a capacity to search, aggregate, and cross-reference large data sets”. To this end, big data also has the characteristics of variability and complexity—data flows can be highly inconsistent, and data may come from complex sources, requiring multiple data linkages and cleansing [SAS Insights, 2017].

11The power of big data is not new and has been harnessed effectively by the private sector for a long time. US retailer Walmart for example, has been tracking the sale of every single item since 1991 using a system called Retail Link [Anderson, 2017]. Internet sites now tailor advertising based on the browsing profile of users, our daily commutes are tracked through our phones and health insurance companies use people’s fitness trackers (such as FitBits) to assess our insurance premiums. Governments are slowly starting to harness the power of big data and the use of big data is now commencing in developing countries, particularly to assess issues of development which were previously plagued by a
lack of data problems.

Applications of big data in development economics

12Latin America has a young, urban population that is hungry for technology [Letouzé, 2017]. Big data, generated through unconventional devices, could, hence, provide an alternative to low quality national statistics and expensive survey data. Individual data from mobile devices and social media, could be aggregated and used in a variety of sectors, including healthcare, agriculture and education. In fact, many early studies of the application of big data in development economics were conducted in Latin America [Letouzé, 2017]. This section provides a literature review of studies that have used big data to tackle development challenges in Latin America.

13There are several ways in which big data can be used in development economics. The UN Global Pulse, the United Nations’ big data initiative [2016] points out that cheaper access to large amounts of data through satellites, mobile tracking, social media and transactional data, can provide information on otherwise difficult-to-reach groups. Recently, there have been various attempts to address the data problems faced in development economics using satellite data. Several studies have used night time luminosity captured by satellites to estimate a country’s level of development. Elvidge et al. [2012] created a Night Light Development Index, using satellite data, which proved to be strongly correlated with the Human Development Index. Chen and Nordhaus [2011] have shown that luminosity is a good proxy of GDP at national levels. Other studies have used satellite data to analyse poverty on subnational levels. Smith and Wills [2016] measure rural poverty as the number of people living in darkness at night. Light is a basic need and research has shown that when people escape extreme poverty they tend to quickly change to electricity from kerosene. Data on light, collected by satellites, is both cheaper and faster to collect, compared to household surveys and can be compiled to a globally balanced panel data set. Their measure is remarkably accurate, as they correctly classify households as living above or below the poverty line in 83 percent of the cases. Henderson, Storeygard, and Weil [2012] use a similar approach, measuring economic growth by studying changes in night light that can be observed from space. This method allows them to identify regional differences in economic growth, which otherwise is difficult using official statistics. Klemens, Coppola and Shron [2015] try to estimate poverty on a very local scale in Guatemala, distinguishing between rural and urban areas. They use satellite data on luminosity, leaf coverage and verdancy, and find that combining this data with survey data significantly improves poverty data.

14In 2010, there were five billion mobile phones in the world. In Latin America, the usage of cell phones is growing faster than anywhere else in the world, and 98 percent of the population today has access to cell signal [Kyte, 2012]. The spread of mobile phones in the developing world has had an enormous impact [UN Global Pulse, 2012]. Mobile phones are no longer only a communication tool; today they are also used for banking and financial services, to search for jobs, and to trade goods and services. Data on how people use their phones is used by researchers to measure poverty in innovative ways. In a study on Latin America, aggregated data derived from mobile phone usage was used to model and predict the population’s different socio-economic levels, by using supervised machine learning techniques. People’s socio-economic status was predicted by matching data on location of the recorded phones with data on the distribution of socio-economic levels. Their prediction was correct for more than 80 percent of the urban population of 500,000 citizens [Soto et al., 2011].

15As a result of the increase in mobile phone usage, more and more people are using social media platforms. Social media reveals information about emerging concerns and can be used to detect signs of crises and outbreaks of diseases. Twitter data, particularly, is most often used as a source of opinion and sentiment data. In a study by UN Global Pulse, Every Woman Every Child and United Nations Foundation [2013], Twitter data was used to observe changes in attitudes to health issues for women and children. The study filters out keywords (such as “breastfeeding” and “maternal health”) from tweets and then analyses the tweets to identify spikes and trends, especially in connection to organised events and campaign efforts. They found that during the time analysed there was an increase in tweets related to women’s and children’s health by 1,000 percent, and spikes in conversations moved together with advocacy events, showing that public awareness reacted to the campaign’s efforts. In Mexico, Twitter has been used to monitor tourism. The study focused on analysing domestic traveling behaviour during a national holiday. 60 million tweets were analysed in order to estimate the number of people who were travelling on the roads. [Data Pop Alliance, 2016].

16A further source of data is transactional data. There is a self-evident link between economic activity and development. Cash withdrawals and bank card payments have much to tell about human behaviour. Records of transactions have been used to identify behavioural differences between different socio-economic groups. Alfaro Martínez et al. [2016] use transactional data to study economic resilience to natural disasters in Mexico. They measure the economic recovery time, defined as the time it takes for the community to return to the baseline economic activity. In this way, the communities are evaluated based on how prepared they are for disasters.

17Big data is increasingly employed in a range of different sectors and can be used to glean detailed insights into the behaviour and preferences of individuals in developing countries. In the health sector, data may be collected through mobile devices by health workers or submitted by individuals, providing insights into population health trends and treatments [World Economic Forum, 2012]. Not only can the effectiveness and cost efficiency of treatments against outcomes be observed, but the movement of affected individuals can also be tracked in a way that was not previously possible. For example, following the 2010 earthquake in Haiti and a cholera outbreak in the capital later in the year, researchers mobilised the power of big data to examine the population movements in and around Port-au-Prince. Bengtsson et al. [2011] tracked the geographic position of 1.9 million active SIM cards to examine the net outflows of individuals from affected areas. SIM card movement reports with a high level of validity could be produced within 12 hours of receiving data and in this way, real time monitoring of the outbreak could be observed and appropriate treatment and targeted information was disseminated. This information also helped humanitarian and aid organisations prepare for future outbreaks. In addition, Twitter data provided information about the cholera outbreak in Haiti two weeks before official statistics [Chunara et al. 2012]. During the 2014 World Cup in Brazil, UNAids also used Twitter data to evaluate how discrimination affected take up of extra healthcare services set up to combat HIV. The researchers filtered 8,000 tweets in Portuguese for HIV topics and compared tweets and their sentiment to the number of people using extra health services such as mobile testing stations or condom giveaways [Erkkola & Luengo-Oroz, 2014]. Satellite data has also long been used to predict the risk of malaria, using remote sensing satellite data on population density, vegetation density, soil moisture and spatial pattern of infrastructure. Thanks to technological advancements, large datasets can today be used in these types of studies. In Brazil, researchers have created malaria risk maps for the Amazon Basin, using remote sensing satellite data, health records and statistics on economic, social and cultural factors. The study focuses on predicting morbidity burden in pregnant women [Data Pop Alliance, 2016]. Big data has also been used in healthcare provision to improve user information and choice. In Uruguay, the Ministry of Health has collaborated with the open data organisation Data Uruguay to create a big data-based web portal, where citizens can get clearer information on the number of doctors at a particular hospital, waiting times, responsiveness to feedback, and costs. The portal is meant to discourage monopolistic practices and corruption in the country’s healthcare system [Hwang, 2018].

18In the education arena, big data is often used by educational publishers who track how many times various resources are accessed or even track the clicks of students using online learning resources to understand and report on students’ interpretation and analytical patterns so that future resources may be better tailored to individuals [Trucano, 2013]. Given the potential benefits of big data, developing nations are also introducing methods to effectively collect big data for decision-making in the education sector. In Mexico, a civil society organisation collects and processes open data on schools to create a platform where people can consult the educational quality and attainments to make better choices on their children’s education1. On a school learning level, an initiative by ProFuturo, a joint organisation by the Fundación Telefónica and the Fundación Bancaria “la Caixa” is developing a programme aimed to use data algorithms and machine learning to optimise education in Latin America, Africa and Asia. Through the integration of digital learning tools and data analytics the programme strives to allow teachers to be continuously informed about their students learning, to predict accomplishments and new scenarios and determine the best learning tools suited for different students and situations [ProFuturo, 2018].

19In the financial services sector too, big data from mobile money services can reveal spending and savings habits within populations [World Economic Forum, 2012]. Kirkpatrick [2015] suggests that “analysis of mobile financial services data could be used to assess availability and uptake of mobile money transfers, access to mobile loans, and repayment default rates”. This could contribute to the Sustainable Development Goal (SDG) 1 (specifically Target 1.4—equal rights to economic resources). Transactional data is used to tackle the problem of tax evasion in Latin America. Tax revenues are essential for governments to fund development projects but tax revenues in Latin America are unusually low. Big data solutions offer opportunities for tax authorities to fight tax evasion. OECD [2016] has investigated the role of big data in tax administration and concludes that “Big Data technology, combined with advanced analytics and increased digital presence […] offer the opportunity to consider moving more to a real-time tax system and to explore new options for managing taxes, countering fraud, dealing with the shadow or hidden economy and assessing how assets are actually being used”. The tax authorities in Brazil have implemented big data technology in its public digital bookkeeping system, which has given them real time access to transactional data. This increases transparency, gives the authorities more detailed information about the taxpayers’ accounts and enables them to check the quality of data by benchmarking taxpayers with similar characteristics. Big data technologies enable authorities to identify broader patterns and trends, detect anomalies and analyse the risk of non-compliance [KPMG, 2016].

20Agriculture is another main domain where big data is already offering significant value. Open data resources such as Global Open Data for Agriculture and Nutrition (Godan) promote the proactive sharing of data to ultimately promote global food security [Godan, 2017]. The use of big data can considerably decrease the possibility of crop failure, and provide up to date information about crops, yields and harvest through satellite and GPS tracking systems to farmers. In a study in Colombia, conducted by the UN, satellite images were used to collect national agricultural data, e.g. to measure and monitor coca crops [The Economic Commission for Latin America and the Caribbean, 2014]. Furthermore, mobile payments for agricultural products and seeds may help governments understand expected food production levels [World Economic Forum, 2012] and ensure the availability of proper crop storage. Real-time data from transactions using SMS transactions can reveal information about markets, the spread of animal and plant diseases and the credit risk of farmers, which has in turn led to more efficient and timely provision of credit lines [Protopop and Shanoyan, 2016]. In Chile, small-scale farmers can subscribe to text messaging services and receive daily information about market prices and weather forecasts [Kyte, 2012]. Another project by Fedearroz and the International Centre for Tropical Agriculture in Colombia aims to improve farmers’ resilience to climate variability and climate change by providing farmers with almost real-time and relevant climate and cropping system information. The organisation used machine learning and network techniques to analyse a combination of commercial harvest data and climate and weather data for optimal conditions. The analysis of favourable sequences coupled with forecasts provides farmers with better data driven information on when and how to maximise rice crop yields. Following the analysis’ advice to not sow rice in the 2014 sowing season, farmers saved an estimated USD 3.6 million [Cariboni, 2014].

21Another sector where big data is being used to an increasing extent is city planning. The urbanisation trend in large parts of Latin America has resulted in a need for better organised cities, and many are becoming leaders in using mobile or sensor data to make cities more productive and liveable, as well as to ensure access to key resources for citizens. Rio de Janeiro, Brazil, is widely considered one of the largest smart cities in the world, where city operations, transport systems and repairs are constantly monitored through video surveillance. The system also works to improve communication with citizens and allows citizens to get involved in proposing policies that benefit them [Hwang, 2018]. In Montevideo, Uruguay, mobile data is used to encourage recycling of household waste in cities. Through an app, users can find the location of recycling stations. The application uses open data from the city council and matches the data with the user’s location from their mobile phone GPS or computer [The Economic Commission for Latin America and the Caribbean, 2014]. In Guadalajara, Mexico, data collected from city wide sensors on street lights and water pipes, for example, allow officials to deploy services and solve issues before they break or stop working [Hwang, 2018].

The potential of big data in impact evaluation

22Beyond examining conditions of development, development agents are concerned with how to help countries develop. Impact evaluations are used to assess the impact and effectiveness of development policies and programmes, and since they are empirical in nature, big data may also change this evaluation process. Most authors seem to agree that big data’s main contribution to impact evaluation is to strengthen current evaluation methods (e.g. randomised control trials, matching and IV), rather than completely changing the way impact is measured. This is because evaluators are interested in causal effects, and these are not necessarily addressed by big data analytics. Instead, big data’s contribution lies in more data of higher quality and the potential to strengthen counterfactuals. The fact that big data makes it cheaper to collect information for larger samples, for instance, can improve the statistical power and precision of an empirical analysis by decreasing variance. Furthermore, the existence of big data in the form of huge databases alone can facilitate the impact evaluation of a programme even when take-up of the intervention is very low. A study by the World Bank and the BBVA Bancomer bank in Mexico used big data to evaluate a financial education programme where take-up was only 0.8 percent. With such low attendance, traditional evaluation methods found no significant impact of the treatment. To find the impact of the workshop the researchers used a large time-series database on credit card clients to match participants to clients in the control group with similar levels and trends in key outcomes. This allowed them to find an impact of the programme despite the low take-up [Lara Ibarra, Mckenzie & Ruiz Ortega, 2017].

23Big data can also solve issues linked to misreporting. In a study, Barstow et al. [2013] used sensors in cookstoves and water filters distributed in a randomised control experiment in Rwanda and found that survey respondents overreported the use of the programme’s water filters and stovesand continued to use old methods of cooking even though they were given new cookstoves as part of the intervention. Similar programmes to provide families with cleaner cooking stoves and water filters exist in Latin America (e.g. a UNDP programme in poor areas of Peru, a Natural Capital Partners initiative in Guatemala). Big data could potentially be used in similar ways as in the experiment in Rwanda, to improve the understanding of the programme’s impact.

24Big data may also facilitate the evaluation of complex programmes [UN Global Pulse, 2016; Raftee & Bamberger, 2014; Letouzé, Areais & Jackson 2016]. Complex development programmes usually include many components and have multiple objectives, are usually non-linear, and often covers large areas. These would benefit from more observed variables, real-time data, and insights into behaviour and attitudes through social media data [UN Global Pulse, 2016].

25Novel applications of data that are currently being investigated include predictive modelling based on Bayesian probabilities [UN Global Pulse, 2016; Latouzé, 2014] and big data driven simulations [Hilbert, 2016]. Hilbert [2016] points out that the use of big data in combination with theory-driven simulation models enables researchers to explore scenarios that never existed, and thus creates a sort of bridge between the simple patterns found in big data analytics and the theory-based understanding of causality. It is already used in the private sector, such as in product development [Cavanough, 2016], as well as for predicting effects of large-scale policies in developed countries [e.g. Abraham, 2012].

26To conclude, most authors agree that big data may improve impact evaluation, but it is unlikely to revolutionise the field in the short term. This is mainly because big data analytics do not completely solve the traditional problems of causality—the focus of impact evaluations.

Concerns regarding big data

27Despite recognising its potential to revolutionise research in the fields of development economics, the literature has also identified potential dangers arising from big data analytics. Ethical concerns have been a prominent criticism to the emergence of big data analytics. While big data itself is described as “fundamentally ethnically neutral” [Isacoff, 2014, p. 161], this does not apply to its use and analysis. Data theft, unauthorised access and manipulation are commonly addressed in the literature [Hamami, 2014; Hilbert, 2016; McFarland and McFarland, 2015]. Some scholars consider ethical issues to be especially severe in developing countries, as the institutional framework plays a crucial role in safeguarding citizens’ privacy and security. Weak institutional frameworks makes monitoring and enforcement of big data companies’ commitments hard to achieve [Hilbert, 2016]. Similarly, developing countries often lack national regulation on subsequent data sharing [Taylor and Schroeder, 2015] and data misuse [Kshetri, 2014]. Data protection regulations vary greatly among Latin American countries, with some richer countries like Chile, Colombia and Mexico reforming its data supervision and protection agencies [Garrigues, 2018] while others, like Brazil, have comparably low standards of data protection laws [Vera, 2017]. As other regions, Latin American countries also need to find balance between protecting their citizens privacy and tackling cybersecurity and allowing the use of big data.

28The opportunities arising from big data analytics cannot be assumed to be shared equally across countries. High-income countries have an advantage in creating, collecting and analysing big data, leading to biases in data production and research [Boyd and Crawford, 2012]. Consequently, Boyd and Crawford argue that this unequal “access to big data creates new digital divides” [2012, p. 673]. Developing countries with lower ICT spending and inferior hardware infrastructure are disadvantaged in developing big data capabilities [Hilbert, 2016] and further suffer from insufficient training capabilities to produce high-skilled analysts [Noormohammad et al., 2010].

29Big data results in more data but not necessarily better data. As large quantities of data do not solve the problem of biases, big data often suffers from the same drawbacks as small data sets do, and the issue of representativeness is often raised. Marginalised groups and the poorer part of the population often produce less data due to lacking technological equipment. For instance, women are 10 percent less likely to own a mobile phone in low- and middle-income countries [GSMA, 2018] which questions the representativeness of mobile phone data. Twitter users are also commonly criticized for not being representative [Kshetri, 2014, Hilbert, 2016]. Considering Latin America’s vast income differences, which translates into a digital divide, poor people and informal sectors will likely be underrepresented in data collected from social media and technological devices, if this issue is not actively addressed.

30Algorithms make social media data difficult to use for economic analysis and additional scrutiny might therefore be needed. Algorithms that determine which data is created can greatly affect the conclusions drawn from its analysis. There are growing concerns that algorithms operate as intransparent “black boxes”, with researchers being unable to determine whether the algorithms propagate certain voices and create biases. The comparability of data over time may be questionable as companies change algorithms without transparency. If a social media platform, due to its algorithms, propagates certain users more than others, resulting in a duplication of these voices (e.g. through retweets of the most visible users), the analysis of such data will disproportionately represent one view, without being representative of people’s actual opinions. People might also take more provocative or extreme stances on social media in order to get more attention, as algorithms often favour highly contested content. Furthermore, there may also be issues with non-human internet traffic by bots that skew results. Considering this, researchers and policymakers using social media should adjust data management practices to take these factors into account. Having access to algorithms, and knowledge of how certain platforms collect and create data will be crucial to avoid misinterpretations of social media data.

31There are concerns about the quality of big data from other sources than social media as well. Letouzé, Areais and Jackson [2016] find limitations specific to the quality of data. For data coming from mobile devices there may be issues of spatial autocorrelation, as ownership is often concentrated in certain areas. Bamberger [2016] also points out the issue of data changing rapidly in form and quality without a way to control for this, which can make the data difficult to interpret continuously and reliably. He recommends the development of dynamic integrated data platforms that permit comparability across sectors, integrating carefully determined and representative national statistics with big data for best quality.

32Several authors have criticised the missing link between big data and development economics. Despite the potential of enhanced analysis and insight from big data as discussed above, these possibilities are no remedy for economic development. Data mining and machine learning methods are common in big data analysis. For example, Chile has made clear that they prioritise finding any patterns over finding explanations given limited resources [Hilbert, 2016]. Without a theory underpinning the trends and patterns seen in the data, they are difficult to explain, generalise and build policy recommendations on. Consequently, Hilbert [2016] emphasises the importance of supplementing big data analytics with theory-driven models and Raftee [2015] encourages using big data only when it is answering the question of interest, rather than using it simply because it is available. In addition, there is a risk that the link between big data analysis and the region-specific factors is weak [e.g. Taylor and Schroeder, 2015]. For example, data analysis is preferably done in higher-technology facilities, far from the problems being studied. Innovative analyses may be disjoint from local conditions, which limits their potential to produce targeted development policies. Finally, we highlight some of the risks linked to the use of big data in impact evaluations. Problems of attributions and spurious correlations may become more serious as data mining becomes more attractive, and is even encouraged by data scientists. Concerns about selection biases also persist and present themselves in new ways: big data is often available only for certain groups, and large numbers of data do not automatically imply a representative sample [UN Global Pulse, 2016].

33“Big data is getting bigger” [LaValle et al., 2011, p. 29]—this statement summarises the general agreement in the literature concerning the future of big data. Most scholars do not doubt that the use and analysis of big data will accelerate and gradually affect more countries, institutions and aspects of development economics. As analysed above, large quantities of data are likely to produce critical insight into questions concerning development economics and to facilitate research, which clearly contributes to big data’s potential in this field. Through the large amount of data collected and analysed from various new sources such as satellites, phones, social media, and sensors, researchers are able to get a better picture of the level of development and poverty in Latin American countries. Among different fields, big data has been widely used in healthcare and disaster relief, from facilitating prevention of disease spread to quicker humanitarian action after disasters. In agriculture, there is also huge potential to leverage big data especially as more farmers are dealing with climate change issues and can benefit from real time or predictive analytics of weather, potential yields and prices. In comparison, Latin American projects leveraging big data for education are scarce in the literature despite considerable challenges in educational attainment in this region. On a general note, it should also be noted that most of the current uses of big data are relatively small-scale examples of its use for quite specific purposes. Big data is not yet widespread across large-scale development projects or actively used by national or large institutions in Latin America.

34Without reliable statistics, it is hard to identify the actual barriers to development and to implement efficient poverty reduction programmes and tailored development policies. It is difficult for governments and non-profit organisations to address problems that they cannot quantify. Big data could offer important information about living conditions and well-being, by providing large amounts of descriptive statistics. With access to real time data, time lags in response strategies could be reduced. Giving warning signals before crises are about to break out, big data sends “digital smoke signals of distress” [UN Global Pulse, 2012, p. 6]. Better information could lead to better interventions, if the data is analysed correctly and the right political measures are taken.

35It is hard to predict how helpful big data will be at addressing development challenges in the region since there have been few studies on this topic. As our review of the literature shows, there is little evidence on how institutions and development agencies in Latin America have applied big data analytics. In contrast to the vast literature concerning private organisations applying big data [LaValle et al., 2011], there is currently no comparable literature for governments and development agencies. This is a vital area for future research, as more and better data will not automatically lead to better development policies if policy-makers are reluctant to apply the insight from big data or if they cherry-pick their favourite results.

36There are still significant problems to be overcome for big data to fully achieve its potential. Technical issues, such as different types of data storage and format, continue to impede the transfer and sharing of data [UN Global Pulse, 2012]. Furthermore, an important requirement for the use of big data is an extensive hardware and software infrastructure, including powerful computers and storage units, analytics software and human resources to process and interpret big data. Data access and sharing are further key issues that will determine the impact big data can have in revolutionising development economics. To fully harness big data’s potential, access to the sources of big data is crucial [Raftee, 2015]. There is still a high reluctance to share data, both among private and public organisations, as well as academia, that is driven by legal and reputational consequences, competitiveness and lack of incentives to share [UN Global Pulse, 2012; Kshetri, 2014].

37Some of the concerns discussed in the previous section are hard to address through public policy, such as misrepresentation of the population and the missing link to development economics. On the other hand, there are multiple suggestions on how to address some other concerns to enhance big data’s potential. Data protection laws are needed to address ethical concerns. To address the technical issues, Hilbert [2016] points towards using financial incentives and subsidies to encourage investment in data repositories and scientific data management systems. One way to account for the unreliability of big data is to supplement it with national statistics. Large amounts of unrefined big data coming from various sources can be combined and compared with statistical information from national agencies of higher quality but lower volume [Durand, 2012]. A Data Pop Alliance report [2016] highlights Latin America’s already advanced statistical systems compared to other developing regions, suggesting that the region’s institutions would be especially suitable to incorporate big data and create best practices in its use by national agencies. A commonly emphasised way to address the issue of data access is the expansion of “open data”, where data is a public good provided especially by governments and the public sector [Lathrop and Ruma, 2010; World Economic Forum, 2012]. More countries could follow the example of Uruguay’s open data initiative for healthcare or Rio de Janeiro’s use of open data for city planning.

38Whether big data will be enough to revolutionise development economics in Latin America depends on how the concerns above are addressed. To achieve all this and fully utilise big data’s potential, there is a need to bridge the gap between data scientists and development agents to create a common language to facilitate collaboration and identify further areas for application.

Haut de page


Abraham Jean, 2012, “Predicting the Effects of the Affordable Care Act: A Comparative Analysis of Health Policy Microsimulation Models”, State Health Reform Assistance Network.

Anderson Mark, 2017, “The hunt begins for Africa’s big data – The Africa Report” [online], 2017 [accessed 4 November 2017]. Available at:

Bamberger Michael, 2016, “Special evaluation challenges facing UN agencies in the age of sustainable development goals, complexity and big data”, in Evaluation Connections, European Evaluation Society Newsletter [online] avril 2016. Available at:

Bárcena Ibarra Alicia and Byanyima Winnie, 2016, “Latin America is the world’s most unequal region. Here’s how to fix it”, World Economic Forum [online] 17 Jan. 2016 [accessed 18 June 2017]. Available at:

Barstow Christina et al., 2013, “Use of Remotely Reporting Electronic Sensors for Assessing Use of Water Filters and Cookstoves in Rwanda”, Environmental Science & Technology, 47(23), p. 13602-13610. DOI: 10.1021/es403412x.

Bengtsson Linus, Lu Xin, Thorson Anna, Garfield Richard and Von Schreeb Johan, 2011, “Improved Response to Disasters and Outbreaks by Tracking Population Movements with Mobile Phone Network Data: A Post-Earthquake Geospatial Study in Haiti”, PLos Med, 8(8): e1001083.

Cariboni Diana, 2014, “Colombia rice growers saved from ruin after being told not to plant their crop”, The Guardian [online] 30 September 2014 [accessed 30 July 2018] Available at:

Cavanaugh Nick, 2016, “Simulation & Big Data Analytics Driving Product Innovation”, HPCwire [online],5 May 2016. Available at:

Corbacho Ana, Fretes Cibils Vicente and Lora Eduardo, 2013, More than Revenue: Taxation as a Development Tool, Inter-American Development Bank, Londres, Palgrave Macmillan Publishers.

Chen Xi and Nordhaus William, 2011, “Using luminosity data as a proxy for economic statistics, The Proceedings of National Academy of Sciences”, 108 (21), 8589-8594.

Chunara Rumi, Freifeld Clark and Brownstein John, 2012, “New technologies for reporting real-time emergent infections”, Parasitology, 139(14), 1843-1851. Available at:

Data-Pop Alliance, 2016, “Opportunities and Requirements for Leveraging Big Data for Official Statistics and the Sustainable Development Goals in Latin America”, Data-Pop Alliance White Paper Series, Harvard Humanitarian Initiative, MIT Media Lab and Overseas Development Institute.

Deaton Angus, 2005, “Measuring poverty in a growing world (or measuring growth in a poor world)”, Review of Economics and Statistics, 87(1), p. 1-19.

Durand Martine, 2012, “Can big data deliver on its promise? OECD Observer”, no 293 Q4.

Easterly William, 2009, “Can the West Save Africa?”, Journal of Economic Literature, 47(2), p. 373-447.

Elvidge Christopher et al., 2012, “The Night Light Development Index (NLDI): a spatially explicit measure of human development from satellite”, Soc. Geogr., 7, 23-35, doi:10.5194/sg-7-23-2012.

Erkkola Taavi and Luengo-Oroz Miguel, 2014, “Can Twitter be a force against HIV discrimination” [online], The Guardian, 13 November 2014 [accessed 29 July 2018] Available at:

Franco-Paredes Carlos et al., 2007, “Commentary: improving the health of neglected populations in Latin America”, BMC Public Health, 7(11). Available at:

Garcia De Rosa Santiago, Guadalupe César and Ruiz Pozuelo Julia, 2015, “The Latin America learning barometer: Moving forward in access, lagging behind in learning”, Brookings Global Economy and Development Working Paper 93.

Garcia Ramirez Jorge Alejandro, 2016, “These are the 5 health challenges facing Latin America” [online], World Economic Forum, 16 June 2016. Available at:

Garrigues, 2018, Data protection regulation in Latin America and the impact of the GDPR [online], 24 May 2018 [accessed 08 August 2018]. Available at:

Gartner, 2017, IT Glossary – Big Data [online] [accessed 29 October 2017]. Available at

Ghosh Tilottama, Anderson Sharolyn, Powell Rebecca, Sutton Paul and Elvidge Christopher, 2009, “Estimation of Mexico’s Informal Economy and Remittances using Nighttime Imagery.” Remote Sensing, 1(3), p. 418-444. Available at:

Global Open Data for Agriculture and Nutrition (Godan), 2017, Godan website homepage [online], 2017 [accessed 7 November 2017]. Available at:

Global System for Mobile Communications (GSMA), 2018, “The Mobile Gender Gap Report 2018” [online], 2018. Available at:

Gómez Ramírez Enrique, 2016, “Latin America’s informal economy: Some formalisation strategies”, Briefing, September 2016. Available at:

Hamami Oren, 2014, “Big Data Security: Understanding the Risks”, Business Intelligence Journal, 19(2), p. 20-26.

Henderson Vernon, Storeygard Adam and Weil David, 2012, “Measuring Economic Growth from Outer Space”, American Economic Review, 102(2), p. 994-1028.

Hewco John, 2017, “Poverty rates are creeping back up in Latin America - Investing in entrepreneurs can help change this” [online], World Economic Forum, 5 Apr 2017, Available at:

Hilbert Martin, 2016, “Big Data for Development: A Review of Promises and Challenges”, Development Policy Review, 34(1), p. 135-174.

Isacoff Christian, 2014, Big Data – Big Opportunities?, 8th International Scientific Conference on Economic and Social Development and 4th Eastern European ESD, Zagreb, Croatie, p. 161-167.

Janvry Alain and Sadulet Elisabeth, 2018, Development: Agriculture for Development at a World Scale: A Striking Discrepancy [online] Center of Latin American Studies, University of California, Berkeley, 2008 [accessed 05 August 2018]. Available at:

Kirkpatrick Robert, 2015, Driving a Big Data Revolution for Sustainable Development and Humanitarian Action[online], Presentation at the UN EGM on Strengthening the Demographic Evidence Base for the Post-2015 Development Agenda, New York, 5-6 October 2015 [accessed 4 November 2017]. Available at:

Kirkpatrick Robert, 2011, Data Philanthropy: Public and Private Sector Data Sharing for Global Resilience [online], UN Global Pulse,16 September 2011 [accessed 30 July 2018]. Available at:

Kipf Andreas et al., 2016, “A proposed integrated data collection, analysis and sharing platform for impact evaluation”, Development Engineering, 1, 2016, p. 36-44. ISSN 2352-7285, Available at:

Klemens Ben, Coppola Andrea and Shron Max, 2015, “Estimating Local Poverty Measures Using Satellite Images: A Pilot Application to Central America”, Policy Research working paper; no. WPS 7329. Washington, D.C., World Bank Group.

KPMG, 2016, “Tax Digitalization: Latin America leads the change” [online], 2016 [accessed 30 July 2018]. Available at:

Kshetri Nir, 2014, “The emerging role of Big Data in key development issues: Opportunities, challenges, and concerns”, Big Data & Society, 1(2), p. 1-20.

Kyte Rachel, 2012, “Latin America leads Global Mobile Growth” [online] World Bank, 18 July 2012 [accessed 20 June 2018], Available at:

Lathrop Daniel and Ruma Laurel, 2010, Open Government: Collaboration, transparency, and participation in practice, Cambridge, MA: O’Reilly Media.

Lavalle Steve, lesser Eric and shockley Rebecca, 2011, “Big Data, Analytics, and the Path from Insights to Value”, MIT Sloan Management Review, 52(2),. p. 20-31.

Letouzé Emmanuel, 2017, Harnessing Big Data in Latin America [online presentation] Eclac-MIT Seminar Santiago, Chile, 6 March 2017 [accessed 20 June 2018]. Available at:

Letouzé Emmanuel, Areais Ana and Jackson Sally, 2016, “The evaluation of complex development interventions in the age of Big data”, in M. Bamberger, J. Vaessen and E. Raimondo eds., Dealing with Complexity in Impact Evaluation, Thousand Oaks, United States, Sage.

Magalhães Fernanda, 2016, “Slum Upgrading and Housing in Latin America”, Inter-American Development Bank. Available at:

Mcfarland Daniel and Mcfarland Richard, 2015, “Big Data and the danger of being precisely inaccurate”, Big Data & Society, 2(2), p. 1-4.

McKinsey & Company, 2011, Big data: The next frontier for innovation, competition and productivity [online] 2011 [accessed 3 November 2017] Available at:

Noormohammad Sheraz, Mamlin Burke and Biondich Paul, 2010, “Changing course to make clinical decision support work in an HIV clinic in Kenya”, International Journal of Medical Informatics, 79-3, p. 204-210.

Organisation for Economic Co-operation and development (OECD), 2016, “Technologies for Better Tax Administration A Practical Guide for Revenue Bodies: A Practical Guide for Revenue Bodies”, OECD Publishing, Paris. Available at:

Paranagua Paulo, 2012, Latin America Struggles to Cope with Record Urban Growth [online], The Guardian, 11 September 2012 [accessed 20 March 2017] Available at:

Parker Helen, 2015, The challenges of data collection in Latin America [online] Quirck’s Media. August 2015 [accessed 16 August 2018] Available at:

Protopop Iuliia and Shanoyan Aleksan, 2016, “Big Data and Smallholder Farmers: Big Data Applications in the Agri-Food Supply Chain in Developing Countries”, International Food and Agribusiness Management Review, Special Issue. Vol. 19 Issue A.

Profuturo, 2018, ProFuturo, UPSA y Telefónica se unen en en busca del “algoritmo de la calidad educativa” [online] ProFuturo,16 May 2018. Available at:

Raftee Linda and Bamberger Michael, 2014,“Emerging Opportunities: Monitoring and Evaluation in a Tech-Enabled World”. The Rockefeller Foundation.

Raftee Linda, 2015, Big data in development evaluation [online] Wait…What? 23 November 2015 [accessed 27 October 2017] Available at:

Rojas-Suarez Liliana, 2016, “Financial Inclusion in Latin America: Facts, Obstacles and Central Banks Policy Issues”, Inter-American Development Bank, Discussion Paper IDB-DP-464.

Sas Insights, 2017, Big Data and Global Development [online] SAS Insights, 2017 [accessed 4 December 2017] Available at:

Smith Brock and Wills Samuel, 2016, “Left in the Dark? Oil and Rural Poverty”, OxCarre Working Paper 164, University of Oxford.

Soto Victor, Frias-Martinez Vanessa, Virseda Jesus and Frias-Martinez Enrique, 2011, Prediction of Socioeconomic Levels using Cell Phone Records, International Conference on User Modeling, Adaptation and Personalization (Umap), Industrial Track, Girone, Espagne.

Steele Jessica et al., 2017, “Mapping poverty using mobile phone and satellite data”, Journal of The Royal Society Interface,14-127.

Taylor Linnet and Schroeder Ralph, 2015, “Is bigger better? The emergence of big data as a tool for international development policy”, GeoJournal, 80, p. 503-518.

The Economic Commission for Latin America and the Caribbean, 2014, “Big data and open data as sustainability tools”, A working paper prepared by the Economic Commission for Latin America and the Caribbean.

Trucano Michael, 2012, Citizens’ Monitoring of the Education Sector in the Philippines [online] EduTech – A World Bank Blog on ICT use in Education, 2012 [accessed 1 November 2017] Available at:

UN Global Pulse, 2013, Every Woman Every Child and United Nations Foundation, Advocacy Monitoring through Social Data: Women’s and Children’s Health.

UN Global Pulse, 2012, Big Data for Development: Challenges & Opportunities.

UN Global Pulse, 2016, Integrating Big Data into the monitoring and evaluation of development programmes.

Vera Francisco, 2017, Data protection laws and cybersecurity: Challenges for Latin America [online] Global Forum on Cyber Expertise, 07 June 2017. [accessed 8 August 2018] Available at:

World Bank, 2017, Global Findex Database [online] Updated 2017 [accessed 5 August 2018]

World Economic Forum, 2012, Big Data, Big Impact: New Possibilities for International Development, 2012. [accessed 5 August 2018] Available at:

Haut de page

Pour citer cet article

Référence papier

Tove Forsbacka Karlsson, Paula Höhne Tarragona et Milindee Jeewandara, « Can big data revolutionise development economics? A literature review on Latin America », Cahiers des Amériques latines, 92 | 2019, 155-175.

Référence électronique

Tove Forsbacka Karlsson, Paula Höhne Tarragona et Milindee Jeewandara, « Can big data revolutionise development economics? A literature review on Latin America », Cahiers des Amériques latines [En ligne], 92 | 2019, mis en ligne le 01 avril 2020, consulté le 27 septembre 2020. URL : ; DOI :

Haut de page


Tove Forsbacka Karlsson

Tove Forsbacka Karlsson est diplômée d’un master en « International Economic Policy » au sein du Paris School of International Affairs à Sciences Po Paris. Elle prépare actuellement un PhD en économie au Stockholm School of Economics.

Paula Höhne Tarragona

Paula Höhne Tarragona est diplômée d’un master en « International Economic Policy » au sein du Paris School of International Affairs à Sciences Po Paris. Elle travaille pour le secteur privé en Allemagne.

Milindee Jeewandara

Milindee Jeewandara est diplômé d’un master en « International Economic Policy » au sein de Paris School of International Affairs à Sciences Po Paris. Il collabore actuellement pour les Nations unies.

Haut de page

Droits d’auteur

Creative Commons License
Les Cahiers des Amériques latines sont mis à disposition selon les termes de la licence Creative Commons Attribution – Pas d’utilisation commerciale – Pas de modification 4.0 International.

Haut de page
  • Logo DOAJ – Directory of Open Access Journals
  • Logo CNRS – Institut des sciences humaines et sociales
  • Logo Presses Sorbonne Nouvelle
  • Logo Institut des Hautes Études de l’Amérique latine
  • Logo Centre de recherche et de documentation sur les Amériques
  • OpenEdition Journals