Digital-Government Carbon Decline Steepens Where Electoral Democracy Is Stronger, 193-Country Panel Finds

Digital government has been sold to finance ministries and development banks as a machine for sharper policy: put services online, instrument the state with data, and the machinery of government will implement decisions more effectively. A new analysis of up to 193 countries observed across twelve United Nations E-Government Survey waves between 2003 and 2024 now adds a sobering political footnote to that pitch. The study finds that the relationship between digital-government development and national carbon intensity follows an inverted U — carbon intensity first rises as countries build energy-hungry digital infrastructure, then falls as the digital state matures — but the decline is not uniform.

The baseline curve is unambiguous on this panel. In a fixed-effects specification with country and survey-wave effects, controlling for industrial share, trade openness, urbanisation and population, the squared term on the UN E-Government Development Index is negative and highly significant. The curve turns at an index value of 0.486, near the middle of the observed range, and the Lind-Mehlum test confirms the inverted-U shape.

The slopes trace the arc directly: a 0.1 increase in the index centred at the tenth percentile is associated with a 12.17 percent rise in carbon intensity, essentially no change at the median, and a 13.52 percent fall at the ninetieth percentile. The result survives a wild cluster bootstrap and Driscoll-Kraay standard errors, so it does not rest on assumptions of independence across countries. The political conditioning emerges when the linear and squared index terms are interacted with the country-mean V-Dem electoral democracy index.

The joint test of the two interaction terms is significant at p = 0.025, and the squared interaction of −3.295 indicates that the curve bends more sharply where democracy is stronger. At the ninetieth percentile of digital-government, the slope of the curve runs from −0.965 at the tenth percentile of democracy to −1.915 at the ninetieth, a difference of −0.951 that is statistically significant. Translated into magnitudes, the same 0.1 increase in the index at high digital-government levels is associated with a 9.20 percent fall in carbon intensity in the least democratic tenth of countries, but a 17.43 percent fall in the most democratic tenth.

Crucially, the study shows that the decline is carried by emissions rather than by the denominator of the ratio. Because log carbon intensity is log emissions minus log output, the two margins decompose additively. Re-estimating the quadratic separately on each margin, the emissions margin shows pronounced and significant curvature in less constrained states — a coefficient of −1.937 — while the output margin shows little curvature and no statistical significance.

The same pattern holds on a consumption-based subsample of 115 countries and reproduces in the independent IEA-EDGAR emissions inventory. This matters because a fall in carbon intensity could otherwise reflect faster economic growth rather than cleaner production, and the commitment-versus-outcome gap documented in prior research warns that political characteristics influence pledges far more readily than actual emissions. The sectoral fingerprint is perhaps the most striking result.

Sorting emissions into a directly governable category — electricity and energy industries, manufacturing and construction combustion, and industrial processes — and a dispersed category of transport and buildings, the study finds the governable curvature is nearly three times the dispersed one: −1.815 against −0.638, with a formal difference of −1.177 that is significant at p < 0.001 in a stacked model pooling both series. At the ninetieth percentile of the index, a 0.1 increase corresponds to a 15.50 percent fall in governable-sector intensity against 8.74 percent in the dispersed sectors. The logic follows the governability of emissions: power plants and heavy industry can be reached by directed administrative measures, whereas transport and the built environment arise from millions of private decisions that respond only slowly to prices and technology.

A scope test on tax collection bounds the mechanism. If the carbon curve reflected a general administrative competence that digital government confers, a similar pattern should appear in tax revenue, a domain digitalisation is known to improve. No inverted U appears in either regime group, and equivalence tests exclude any parallel pattern larger than 0.62 percentage points of GDP in less constrained states and 0.95 points in more constrained ones.

The carbon result holds robustly on the same observations, so the null is not an artifact of the smaller tax sample. The mechanism, the authors conclude, pertains to governable emissions rather than to a diffuse boost in state capability. The findings come with substantial caveats that the authors state plainly.

The conditioning cannot be separated from resource dependence: the country mean of electoral democracy correlates at −0.511 with natural-resource rents, and when both moderators enter together, the rents interaction remains significant while democracy's weakens. No causal claim is made, and the fixed-effects design identifies conditioning through differences in curvature across countries, which may reflect any national characteristic correlated with democracy. Most consequentially, when a country-specific linear trend is added, the inverted U disappears altogether; only a modest decline remains at high digital-government levels, of 3.38 to 3.54 percent per 0.1 increase at the ninetieth percentile, compared with 13.52 percent in the baseline.

The curve may partly describe a shared temporal trajectory rather than digital-government itself. Other boundaries are equally candid. The inverted U does not hold for total greenhouse-gas intensity, which includes methane, nitrous oxide and other gases, and actually reverses among less constrained states — agriculture, which accounts for 48.1 percent of non-CO2 greenhouse gases in the panel, falls outside the governable mapping.

The sector contrast weakens and loses significance under a stricter specification allowing a separate trend for each country-sector. And the study's central conceptual construct, politically conditioned administrative reach, cannot be decomposed: electoral democracy, rule of law and public-administration quality correlate at 0.81 to 0.93 across countries, so no specification can precisely separate their contributions. Four designs built to distinguish the axes all fail to yield a significant separating term.

What the evidence does support is a refinement of the democracy-and-environment literature. Rather than democracies simply emitting less, the political character of the state conditions the form of the association: the curve bends more sharply where democracy is stronger, even though the expected finding that lower political constraint brings the turning point earlier is not borne out — the low-democracy turning point cannot even be located, and the two turning points cannot be told apart. The sectoral pattern, meanwhile, echoes the directed-action cases documented in China, where binding carbon-intensity targets were enforced through cadre evaluation and the decline was located in steel, building materials and chemicals.

For governments and development institutions financing digital-government programmes, the implication is that the carbon payoff is conditional, not automatic. The authors caution that the conditioning reflects enduring national characteristics that digitalisation alone cannot create, and that the findings are a positive empirical claim, not an endorsement of any political arrangement. The practical suggestion is to pair digital-government investment aimed at emissions with the capacity to monitor and enforce policy in electricity generation, energy industries and heavy manufacturing, where the decline is concentrated — and to expect far shallower returns in transport and buildings, where the reach of even a mature digital state runs out.

Subject of Research: Political conditioning of the association between digital-government development and national carbon intensity across 193 countries, 2003–2024

Article Title: When digital-government indicators predict falling carbon intensity: Evidence from 193 countries, 2003–2024

Article References: Cagle, M. N. (2026). When digital-government indicators predict falling carbon intensity: Evidence from 193 countries, 2003–2024. Environmental and Sustainability Indicators, 32, Article 101577. https://doi.org/10.1016/j.indic.2026.101577

Image Credits: AI Generated

DOI: 10.1016/j.indic.2026.101577

Keywords: digital government, carbon intensity, electoral democracy, state capacity, e-government, carbon emissions, inverted-U relationship, sectoral decomposition, climate policy, V-Dem, UN E-Government Survey, political institutions