Digital Infrastructure as a Resilience Multiplier in Crisis-Prone Developing Economies

Digital resilience rests on three rails. Connectivity — phones and internet — carries warnings, information and transactions. Payments — accounts at banks or mobile-money providers — let money move without cash or queues. Data systems — population registers, social registries and the statistical capacity to measure what is happening — tell governments whom to help and whether help arrived. This report measures the first with internet use (ITU), the second with account ownership (Global Findex) and the third with the World Bank's Statistical Performance Indicators, a 0–100 score of how well a national statistical system produces and shares data.

All three have improved. Internet use in the median low- or middle-income country rose to 65% in 2024, and the median share of adults with an account rose from 22% in 2011 to 57% in 2024; the Global Findex 2025 reports that 75% of adults in developing economies, weighted by population, now have one. Statistical performance has improved more slowly. The gaps between income groups remain stark: in the median low-income country, 19% of people use the internet and 39% of adults have an account, against 94% and 97% in high-income countries, and its statistical performance score is 58 against 89.

The rails also depend on foundations outside this index. The World Bank's ID4D initiative estimates that more than 800 million people have no proof of legal identity and 2.8 billion no government-recognised digital identity, which keeps them out of many registries and accounts; the ITU estimates that 2.2 billion people remain offline, 96% of them in low- and middle-income countries. Those excluded from the rails are often the people most exposed to shocks.

The clearest evidence for digital infrastructure as a resilience tool comes from the 2020 cash response. Social protection was the main instrument governments used: the World Bank's living review counted 3,856 social protection measures across 223 economies, with cash transfers reaching about 1.36 billion people. How fast that money arrived depended heavily on what already existed. India announced its relief package on 26 March 2020 and credited payments to the accounts of about 200 million women between 3 and 9 April. Brazil's emergency aid law passed on 2 April and the first payments went out on 9 April, with more than 10 million digital accounts opened in the first week.

Pakistan launched its Ehsaas emergency cash programme on 1 April and began paying on 9 April through more than 17,000 payment points, reaching 15 million families by October. Togo's Novissi programme, built from voter rolls and mobile money, paid 567,002 people in its first phase, 65% of them women. The contrast is the Philippines: households already in the 4Ps conditional cash transfer programme, with existing payment arrangements, were paid by 5 April, while those outside it waited until June or later, and in a second round 30–35% of payroll records failed validation. Peru, starting with few beneficiaries reachable digitally, paid about 30% of its first bonus digitally and set a target of 80%.

Speed matters because losses compound. Households facing a shock run down savings, skip meals, sell assets and pull children from school, and the longer help takes, the more of those losses become permanent. In Bangladesh in 2020, WFP sent 4,500 taka by mobile money to more than 23,000 households about four days before a flood peak; recipients were 52% less likely to go a day without eating. In Pakistan's 2022 floods, BISP had paid PKR 25,000 each to more than 1 million families by 19 September. These are the gains digital rails make possible — and the macro sections that follow ask whether they add up to faster national recovery.

To test recovery at the level of whole economies, the analysis identifies every shock since 1991 in which GDP per person in a low- or middle-income country of at least 1 million people fell by 2% or more in a year — 207 episodes with five years of follow-up, from the post-Soviet collapses and the Asian financial crisis to commodity busts, conflicts and the pandemic. The typical shock cut GDP per person by 5% in its first year. Over the following five years the median economy lost the equivalent of 21% of a year's pre-shock output, and only 49% of episodes regained their pre-shock level within five years; 37 never had by 2024.

Comparing economies by their digital infrastructure just before the shock, the raw pattern looks promising: episodes in countries in the bottom third of internet use lost a median of 31 percentage-point-years of output, against about 16 in the middle and top thirds. But internet use rises with income, and richer economies tend to recover faster for many reasons. Once income and the size of the initial drop are held constant, internet use — measured as a country's rank among all countries that year, to remove the global diffusion trend — has no association with lost output. Government effectiveness, in contrast, is strongly protective: one standard deviation is associated with about 13 percentage-point-years less output lost.

Account ownership and statistical capacity look protective on their own — one standard deviation of account ownership goes with about 9 percentage-point-years less loss — but most of that disappears once government effectiveness enters the model, because good governments tend to build accounts and statistics too. The episode data cover account ownership only since 2011 and statistical performance only since 2016, so these estimates rest on fewer episodes. The chart lets you compare each measure alone and with governance held constant, and the scatter shows the individual episodes.

The pandemic recession is a natural test: one shock, at one time, everywhere. In 105 low- and middle-income economies, GDP per person fell by a median of 4.2% in 2020; by 2023 the median economy was 3.8% above its 2019 level, but 34% were still below it, and relative to each country's 2014–19 trend the median economy had lost about 19 percentage-point-years of output. The question is whether economies with more internet use, accounts and statistical capacity before the pandemic fared better, after allowing for income, tourism dependence — the single biggest driver of 2020 losses — and pre-pandemic growth.

On average, they did not by much. Internet use in 2019 is unrelated to the 2020 fall or to the gap by 2023. Statistical performance is associated with stronger recovery by 2023 — about 2.4 points more GDP per person per 10 points of the score, a statistically significant association — and account ownership, less precisely, with smaller cumulative losses; both shrink toward zero once government effectiveness is controlled. Government effectiveness itself is again the clearest predictor of how much output a developing economy recovered: one standard deviation is associated with about 11 percentage-point-years less cumulative loss.

That does not mean digital rails made no difference in 2020. Macro averages blend countries that used their rails for fast, broad cash transfers with others that did not, and GDP misses much of what transfers protect — consumption, nutrition, children's schooling, small firms kept alive. It does mean that digital readiness on its own is not a reliable shield for whole economies. The next section shows the pattern behind the average: the association between accounts and recovery depends sharply on the strength of the state.

Allowing the association between account ownership and recovery to vary with government effectiveness reveals the pattern the averages hide. Among developing economies with government effectiveness around −1 — roughly the weakest quarter of the sample, including much of the Sahel and several conflict-affected states — each additional 10 percentage points of adults with an account in 2017 was associated with a fall in GDP per person in 2020 about 1.1 points smaller, recovery by 2023 about 2.4 points stronger, and about 7 percentage-point-years less cumulative lost output. Among economies with government effectiveness around zero, the association was nil.

One interpretation fits this pattern and the household evidence: where the state cannot reach people through capable ministries, local offices and reliable registries, payment rails do some of that work for it. Mobile money lets cash reach a village without a bank branch; accounts let transfers bypass slow and leaky distribution chains; transaction data reveal who is affected when official statistics lag. In capable states, which have other channels to reach people and stabilise demand, accounts add less at the margin. Digital rails are a multiplier on thin capacity.

The finding should be held loosely. It rests on 78 countries and an interaction term, the account data predate the pandemic by three years, and countries that build accounts despite weak government may differ in other ways, such as remittance flows or private mobile-money networks. Nor does it say that weak states can skip building capacity: the best outcomes in the data still belong to countries with both. But it points to where the marginal dollar of digital infrastructure may buy the most resilience — in the states least able to respond any other way.

Each low- and middle-income country receives a digital readiness index: the average of its percentile ranks among developing economies on internet use, account ownership and statistical performance — connect, pay and count. Readiness is highest in parts of Latin America, Eastern Europe, Central Asia and East Asia, and in Africa in Mauritius, South Africa, Senegal, Kenya and Ghana; it is lowest across the Sahel, Central Africa and in fragile and conflict-affected states, where all three rails are thin at once.

Crossing readiness with government effectiveness gives four groups. Ready rails with capable states have the full toolkit. Ready rails with weak states — Ukraine, Kyrgyzstan, Bolivia and Palestine, for example, at the default thresholds — are where section 5 suggests digital infrastructure substitutes most for missing capacity, and where protecting and extending it matters most. Capable states with thin rails can add digital channels to institutions that already work. The hardest group combines thin rails with weak states, and holds many of the people most exposed to climate, conflict and price shocks.

Move the thresholds to see how the groups change, and pick a map layer to see each rail separately. The country card shows each country's three rails, its governance score and how its economy fared through the pandemic. The index is relative — a country can rise or fall because others moved — and it omits identity systems, data protection and cyber security, for which comparable data are not available for all countries.

The first risk is exclusion. Digital channels reach those already connected, and the people left out — without identity documents, phones, accounts or digital skills — are disproportionately poor, rural, older and female. The Global Findex 2025 estimates that 1.3 billion adults still have no account, and the World Bank's State of Social Protection 2025 report finds 1.6 billion people with no social protection coverage at all. A response that runs only through digital rails can leave the most vulnerable furthest behind, which is why the most successful 2020 programmes kept cash-out points, agents and manual enrolment alongside digital channels.

The second risk is fragility of the rails themselves. In June 2024 a ransomware attack on Indonesia's temporary national data centre disrupted services at more than 200 government institutions, a reminder that centralising data and services concentrates risk. Power cuts, network outages and fraud can stop payments exactly when they are needed most, and shocks such as floods or conflict often knock out connectivity in the places worst affected. Resilient digital infrastructure needs redundancy, offline fallbacks, cyber defences and tested continuity plans, not only coverage.

The third risk is trust. Registries and transaction data that let governments target help can also be used for surveillance or exclusion, and weak data protection undermines the willingness of people to enrol. Early warning illustrates the gap between rails and outcomes: by 2025, 119 countries — 60% — reported multi-hazard early warning systems, but only 43% of small island developing states did, and a warning that reaches a phone only saves lives if people trust it and can act on it.

The evidence supports investing in digital infrastructure as part of resilience — but as a complement to state capacity, not a substitute for it. The rails deliver their biggest gains when they are pre-positioned: accounts already open, registries already populated, payment systems already linked to social programmes, and rules for scaling up already agreed. Countries that paid within a week in 2020 had spent the previous decade building that plumbing; those that started in the crisis took months.

The results also suggest where the next dollar does the most. In states with weak administrative reach, payment rails and registries can extend the state's capacity to help people quickly, so financial inclusion, identity and data systems deserve priority there — paired with simple, well-governed delivery programmes rather than complex ones. In capable states, the gains come more from integrating digital rails into institutions that already work: linking registries, automating triggers, sharing data across agencies and using real-time data to adjust policy.

Across all settings, resilience depends on designing for those left out and for failure. That means keeping non-digital channels open, investing in identity for all, protecting data by law and in practice, building redundancy and cyber defences, and testing systems before a crisis — and it means statistical systems capable of measuring a shock as it happens, because the response can only be as fast as the information behind it.

Data are from the World Bank's World Development Indicators (GDP per person in constant US$, internet use from the ITU, mobile and broadband subscriptions, tourism receipts), the Global Findex database (account ownership 2011, 2014, 2017, 2021 and 2024, and digitally enabled accounts in 2024), the Statistical Performance Indicators (overall score, 2016–2024) and the Worldwide Governance Indicators (government effectiveness). The sample covers countries with at least 1 million people in 2020; income groups are the World Bank's current classification.

A shock episode begins in a year when GDP per person falls by 2% or more, provided any previous episode has recovered; recovery is the first year GDP per person regains its pre-shock level, and five-year loss sums the shortfall below that level over the shock year and the following four. 207 episodes in low- and middle-income countries have five years of data. Digital measures are taken in the year before the shock; internet use enters as a percentile rank among countries in that year, and account ownership is interpolated between survey years. The COVID cross-section uses 105 developing economies with complete data; the cumulative gap compares GDP per person in 2020–23 with a projection of each country's 2014–19 growth rate.

Models are OLS with heteroskedasticity-robust standard errors, controlling for log GDP per person and the first-year drop (episodes) or log tourism receipts share and pre-COVID trend growth (COVID), with and without government effectiveness; standardised effects are per standard deviation of each measure. The weak-state result comes from an accounts × governance interaction on 78 countries. Results are associations and may reflect unmeasured differences between countries. 2020 programme facts are from official and World Bank sources. Related reports: resource scarcity and armed conflict and the network of interdependent risks.