A view to a kill…ing of productivity: The allocative cost of war in Ukraine

Estimates of the cost of Russia’s war on Ukraine are dominated by what the war destroys. This column measures a second cost: the labour and capital that survive but end up stuck in low-value uses. Firm-level financial statements filed throughout the full-scale invasion show allocative productivity approximately 20% below its 2021 level by 2024, with roughly half of the wartime output loss attributable to misallocation rather than destruction. Attack intensity, not distance from the front, determines where the damage falls. Unlike destroyed capital, this loss can in principle be recovered without new investment.

Four years into the full-scale invasion, the cost of Ukraine’s reconstruction and recovery is almost $600 billion over the next decade (World Bank 2026). That figure anchors a difficult financing debate. EU leaders agreed a €90 billion loan for 2026 and 2027 in December 2025 after failing to bridge differences over a loan backed by immobilised Russian assets, and several member states have since pushed to reopen the case. Almost all of that discussion concerns physical things – housing, energy, transport – and who will pay for them.

That framing follows from how the cost of war is usually measured. The cross-country literature works with aggregates, documenting large and persistent output losses across conflicts (Federle et al. 2025). Granular work on Ukraine has begun to complement it: Anastasia et al. (2026) find that the labour market worked markedly worse where the war was worst. Matching between workers and vacancies became about 15% less efficient nationally after February 2022, but the average hides a wide gap: 4% to 8% in the western oblasts against 17% to 24% in those closest to the front line, where some local markets ceased to function altogether. 

But aggregate output depends not only on how much capital and labour an economy has. It also depends on how these resources are distributed across firms. Hsieh and Klenow (2009) made this channel measurable. In a new paper (Amann et al. 2026), we quantify losses in productivity due to misallocation. 

In a sign of remarkable resilience, Ukrainian firms kept filing financial statements throughout the invasion. We build a balanced panel of Ukrainian firms for 2018 to 2024 and five peer economies, and use the Hsieh-Klenow framework to turn the dispersion of firms’ marginal revenue products into an index of allocative productivity. Intuitively, it asks how much extra output the same capital and workers would generate if spread across firms more efficiently. We weigh firms by their current employment, so that a plant with a thousand workers counts for more than a workshop with five, which makes the index informative about output.

Figure 1 shows that allocative productivity in Ukraine was roughly flat through 2021, then fell sharply after February 2022. By 2024 it stood ~20% below its 2021 level, while none of the peer economies deteriorated comparably.

Figure 1 Allocative productivity in Ukraine and peer economies, 2018–2024 (2021 = 1)

Note: Firms weighted by current employment. Peers are Bulgaria, Croatia, Estonia, Poland, and Romania.

The magnitude matters for the reconstruction debate as well as for the defence effort. A growth accounting decomposition of the 2022 collapse leaves about half of the output loss unexplained by the labour that left or the capital that was destroyed. Our estimates of misallocation can rationalise that residual. In other words, Ukraine did not merely lose factories and workers. It also lost a significant part of its ability to make good use of the ones it kept.

The firm accounts are not the only evidence. We also use three million online job postings from the aggregator Jooble.org. If posted wages track the marginal product of labour, the spread of wages offered for the same occupation should widen as allocation deteriorates. Within-occupation dispersion in posted wages was flat at about 0.41 before the war, compressed briefly in 2022, and then rose to roughly 50% above its pre-war level in 2023-2024. Two independent datasets tell the same story.

Where does that damage fall? The intuitive answer is geography: the closer to the fighting, the worse the disruption. Using geocoded incident data from the VIINA project (Zhukov 2023), we measure local exposure from war-related property damage events per capita across 96 raions (districts). Attack intensity, not geography, drives the damage to allocative efficiency. According to our estimates, a one standard deviation increase in local war damage is associated with an 11 percentage point decrease in allocative productivity (Figure 2). War intensity alone accounts for more than 60% of the variation in productivity losses across districts. Once it is included, distance to Russia and to the line of control become statistically insignificant. Kyiv illustrates the point: far from the front line, heavily struck from the air, and with productivity losses to match. There is also a sectoral dimension: construction and retail trade lost about half their allocative efficiency, while agriculture, which is geographically dispersed and less dependent on urban infrastructure, proved the most resilient.

We also find that liberated raions show no permanent scar from occupation once attacks are accounted for. What damages the allocation of resources is being under attack now, not having been overrun in the past. The gradient mirrors what Anastasia et al. (2026) find across oblasts in the labour market, at finer resolution and for capital as well as labour.

Figure 2 Change in raion-level allocative productivity against local war damage

Note: Each bubble is one of 96 raions, sized by firm count, and the fitted lines are firm-count weighted. War intensity is that year’s property damage events per 100,000 pre-war residents, logged and standardised within year.

The picture is not one of passive decline. Employment moved during the war towards firms that proved more resilient under attack, especially towards firms located in less exposed regions and industries. Comparing the actual path with a counterfactual in which the composition of employment stayed frozen at its pre-war pattern shows that this reallocation was worth more than 8 percentage points of allocative efficiency by 2024. Without it, the loss of around 20% would have been substantially larger.

There may be scope to recover more still. Internal displacement, the largest single reallocation of people in the war, has been largely local: households moved to the nearest district that would take them, often one still under regular attack. Set against our district-level results, that pattern concentrates displaced labour where its marginal product has fallen disproportionately. Better housing, transport links and job placement in safer districts could raise the allocative gain from reallocation further.

The geography result also makes air defence a productivity policy. If attack intensity rather than distance explains where productivity is lost, intercepting strikes protects the output of an entire region and not merely the assets that would otherwise have been hit. For Ukraine, this is a reason for European donors to treat air defence deliveries as economic as well as military support, and to prioritise them accordingly: the return includes output that a protected region keeps producing and that appears in no damage assessment. Proposals for protected economic clusters combining air defence, war insurance and reliable energy (Dombrovskis et al. 2024) should be judged on that broader return. The same logic applies to European rearmament: how much output an economy retains under attack depends on whether resources can move quickly, and on whether attacks behind the front stay sparse enough for moving to be worthwhile.

The mechanism should run in the other direction too. Our data cover years in which Russia held a near monopoly on long-range strikes. That balance has since shifted. Ukraine now reaches deep into Russia, hitting oil refineries and logistics centres, in a campaign Kyiv calls “long-range sanctions”. The refinery strikes in particular have bitten: Mikula and Sabatini (2026) find that nighttime radiance around struck refineries falls by 15% to 18% within five kilometres and stays below its previous path more than a year later, with no comparable effect at 106 other deep-strike targets. If sustained attack degrades allocative efficiency in Ukraine, it should do the same in Russia, in which case these strikes erode Russia’s capacity to fund its aggression by considerably more than the physical damage alone would suggest.

Gorodnichenko and Obstfeld (2026) estimate that Ukraine will need at least $40 billion a year in new investment to rebuild its capital stock and begin converging on its EU peers. Nothing here reduces that bill. But our analysis identifies another margin that requires far less spending: the workers and machines already in Ukraine are producing less than they could. Recovering their productivity calls for different instruments – air defence, mobility for displaced workers, war insurance, and the administrative capacity to let firms and workers move – which are cheap by comparison.

War kills the economy twice: once by destroying capital and labour, and again by crippling its ability to use what remains. The first loss is expensive; the second can, in principle, be reversed by reforms that make better use of existing resources. Ukraine is leaving money on the table. Picking it up is the fastest, cheapest form of recovery.

Source : VOXeu