HomePublicationsMain points
PDF
U4 Helpdesk Answer

Anti-corruption reform in Ukraine: what does the empirical evidence say?

Conditions enabling and hindering reform success since 2005

Empirical evidence on anti-corruption reform in Ukraine is thinner than the wider literature suggests but offers several consistent lessons. The strongest evidence indicates that reforms work when they reduce discretion, align incentives and embed transparency and accountability into institutional design. Decentralisation and police reform show that participatory and visible institutional change can improve public trust, while awareness raising in isolation can have limited or counterproductive effects. External conditionality has been important for reform adoption, but sustained implementation depends on domestic pressure, institutional safeguards and continued international involvement. Major evidence gaps remain, particularly on enforcement, political finance and reforms undertaken since the full-scale Russian invasion.

Anti-corruption reform in Ukraine: what does the empirical evidence say?

Main points

  • Ukraine has produced a modest body of empirical evidence on the outcomes of anti-corruption reforms, concentrated mainly on the 2014-2022 period and in two intervention areas above all others: public procurement and decentralisation. Fewer than 20 studies meet a robust empirical bar; the wider literature is chiefly descriptive.
  • The strongest evidence concerns e-procurement and e-auctions. Effects on prices, speed and revenue in public contracting and the sale of public assets are real but depend on reforms to the design that account for vested interests. For example, the mandatory shift of land lease auctions to Prozorro.Sale raised revenue by around 175%, whereas an earlier electronic pilot project run by the state geocadastre, which simultaneously monopolised cadastral surveying, organised the auctions and could disqualify bidders at short notice, produced prices 12–19% lower than offline auctions. Digitisation appears to reduce rent-seeking behaviour only where this reduces discretion on the part of officials with vested interests in the outcomes.
  • Decentralisation reforms between 2015 and 2020 increased trust in local government through the participatory process of fostering a stronger sense of local community and stake in government, not fiscal devolution alone, with effects persisting four to five years and underpinning wartime community resilience. However, effects did not spill over into heightened trust in national institutions.
  • Police reform indicates why political will can be scarce: abolishing the old traffic police and replacing it with a newly recruited, vetted and better paid patrol force durably improved perceptions of police integrity but produced no electoral return for reformers and only brief spillover to national institutions. Citizens only credit reforms they can observe, and that credit decays fast.
  • Awareness raising and integrity training deliver limited or mixed effects. Publicising the scale of corruption can normalise it among audiences with no prior personal experience of corrupt practices, while anti-corruption training changes behaviour only when participants know their peers were trained too. Social norms and expectations regarding the behaviour of others, rather than individual integrity or conviction, appear to constitute the operative mechanism.
  • The most consistent enabling conditions operate at different stages of the reform process. EU and IMF conditionality, especially combined with civil society pressure (the “sandwich” strategy of complementary top-down and bottom-up approaches), is the strongest driver of formal adoption, but is consistently weaker at securing meaningful delivery once reforms are adopted. Delivery and durability depend more on international involvement in appointments to independent institutions and on transparency-by-design, which protect institutions from capture and embed accountability into implementation itself.
  • The main hindering conditions are elite resistance, ranging from capture of some institutions to sustained attempts to limit the independence of others (NACP, NABU, SAPO and HACC), judicial capture, the persistent gap between legislative adoption and implementation, and, since 2022, the use of direct contracting, data classification and concentration of power.
  • Critical gaps remain across both intervention areas and time periods. Enforcement and sanctions (NABU, SAPO and HACC performance), political party finance, whistleblower protection, e-declarations, digital public services and the DoZorro civic monitoring mechanism all lack rigorous counterfactual evaluation. The 2005–2014 period remains understudied, reflecting both data scarcity and a comparatively inactive reform period; the post-2022 period is the least empirically studied of the three, since research has not yet caught up due to wartime conditions.

Query

What is the state of empirical evidence on anti-corruption interventions and reforms in Ukraine since 2005, and what conclusions can be drawn on conditions that enable or hinder success?

Introduction

Development partners are channelling significant resources into Ukraine’s recovery and reconstruction while the country simultaneously pursues EU accession under wartime conditions. Decisions about how to sequence, condition and safeguard that support should ideally be grounded in evidence of what has worked in Ukrainian anti-corruption reform to date. This Helpdesk Answer takes stock of what the empirical record shows. It reviews the state of empirical evidence on anti-corruption interventions and reforms in Ukraine since 2005 and synthesises what that evidence says about the conditions that enable or hinder success.

The answer focuses on studies that evaluate the impact of specific anti-corruption interventions using empirical methods: experimental trials, quasi-experimental designs, rigorous qualitative research such as process tracing and structured comparative case studies, as well as robust independent programme evaluations. General assessments of the state of corruption in Ukraine, expert commentary and descriptive overviews of reform processes are used as supporting context where they help interpret the empirical findings, but they are not the primary evidence base. This distinction matters because the volume of anti-corruption writing on Ukraine vastly exceeds the volume of anti-corruption evidence. A two-decade scoping review of the corruption literature on Ukraine confirms that the field is dominated by descriptive research, relies heavily on old data and offers limited comparability across studies (Belyaev, Villeneuve and Mugellini 2025).

As such, readers should expect the evidence base assembled here to be thinner than the literature might suggest; the paucity of robust empirical studies could in fact be considered a finding in itself.

For analytical purposes, this paper organises Ukraine’s reform trajectory into three periods: post-Orange Revolution (2005–2014), post-Maidan (2014–2022) and the period of Russia’s full-scale invasion (2022–present). The paper flags where this periodisation is supported or complicated by the evidence, for instance, where a study’s own sample period straddles two of these windows or where an institution’s trajectory carries through a boundary rather than resetting at it.

It then provides a brief narrative of the reform trajectory, followed by a review of the empirical findings by intervention type, following a thematic taxonomy designed for this paper: formal legal and institutional reforms; public governance reforms; transparency and accountability initiatives; political finance integrity measures; enforcement and sanctions; preventive approaches; and awareness raising and education. Each study is labelled with its place in a hierarchy of confidence, where the nature of the study design (for example, experimental or process tracing) is used as a proxy for the reliability of its findings. Full criteria are set out below. It then synthesises the conditions that enable or hinder reform success, organised under political, social and capacity conditions and drawing on the same empirical and contextual evidence, rather than a separate literature review. Finally, it sets out the evidence gaps. The annex presents in tabular form the empirical studies discussed in the intervention type review, categorised according to intervention type, period, method and confidence tier.

Background: Ukraine’s anti-corruption reform trajectory

Post-Orange Revolution (2005–2014)

The decade after the Orange Revolution produced few lasting anti-corruption reforms. Initiatives were fragmented and largely tokenistic: Ukraine acceded to the Council of Europe’s Group of States against Corruption (GRECO) in 2006 and ratified the UN Convention against Corruption in 2009, but neither commitment translated into a coherent institutional architecture, and earlier anti-corruption programmes going back to the late 1990s are assessed as having imitated the process of reform rather than delivered it (Lough and Dubrovskiy 2018). The one reform of the period that stands out as both genuine and durable is the introduction of external independent testing (ZNO) for university admission from 2008 (Klein 2014), which removed admission decisions from the discretion of individual institutions and is widely credited with curbing entrance related bribery, although, no rigorous impact evaluation has been identified.

Contrasting Ukraine’s post-2005 trajectory with Georgia’s after the Rose Revolution, Nasuti (2016) argues that Georgia’s success rested on administrative cohesion, more precisely, a ruling administration uniformly oriented toward reform. Ukraine’s coalition cabinet, by contrast, distributed ministries across parties with divergent backgrounds and priorities, and most ministers were career politicians rather than the Western-educated technocrats who dominated Georgia’s cabinet. Nasuti (2016) attributes the resulting inconsistent implementation, and the space this left for vested interests to reassert themselves, directly to this pattern. Quasi-experimental research on Ukraine’s non-accession to the EU points to a compatible dynamic through a different empirical route. Focacci, Kovac and Spruk (2025) estimate that Ukraine paid a measurable cost for remaining outside the EU integration process during 1996–2020, framing the early 2000s as a missed window of opportunity with compounding costs over time.

Additionally, the Orange Revolution reshuffled which business clans held influence but did not dismantle the underlying system. Rival oligarchic groups backed competing camps within the nominally reformist coalition, including the split between President Yushchenko and Prime Minister Tymoshenko. Markus and Charnysh (2017) find that where oligarchs entered political office directly, this did little to protect their wealth during systemic shocks. More effective oligarchic strategies to translate their economic clout into political power related to indirect channels of influence peddling, including financing political parties and expanding their ownership of media outlets. This allowed business elites to shape policy and public discourse without the scrutiny that direct office would invite.

This helps to explain why formal institutional reforms in this period coexisted with little substantive change. The rent extraction model did not disappear; it moved from the direct control of state office and assets that had characterised the Kuchma era toward the indirect channels of party financing and media ownership that Markus and Charnysh document, and the individuals occupying those channels also changed as business clans reshuffled. No systemic measure closed off the indirect channels, so the underlying extraction logic persisted even as its personnel and mechanisms shifted. For instance, the natural gas intermediary arrangements of the period, most notably the RosUkrEnergo scheme, are the most frequently cited illustration of energy sector rent extraction by politically connected actors (Wolczuk 2016), though this paper has not identified a rigorous empirical evaluation of that specific episode.

A descriptive comparative account by Pleines (2016) of oligarch-politics links across the Kuchma, Yushchenko and Yanukovych periods corroborates the pattern of oligarchic adaptation rather than retreat. Pleines observes, for instance, that the constitutional reforms of 2004 – reinstated only after Yanukovych’s 2014 removal – had themselves been a product of oligarchic bargaining rather than representing a clean break from it.

Post-Maidan (2014–2022)

The Revolution of Dignity triggered the most reform dense period in Ukraine’s modern history and is the era best covered by empirical literature. Between 2014 and 2019, Ukraine built a dedicated anti-corruption infrastructure consisting of five independent bodies:

  • National Agency on Corruption Prevention (NACP)
  • National Anti-Corruption Bureau (NABU)
  • Specialised Anti-Corruption Prosecutor’s Office (SAPO)
  • Asset Recovery and Management Agency (ARMA)
  • from 2019, High Anti-Corruption Court (HACC)

The same period produced the ProZorro e-procurement system (which became mandatory from 2016), the Prozorro.Sale auction platform, one of the world’s most ambitious electronic asset declaration systems, beneficial ownership transparency requirements, decentralisation reform creating amalgamated territorial communities (hromadas), and the replacement of the Soviet-era traffic police with a new patrol police. Judicial reform proceeded separately, in two successive waves, the first beginning after 2014, the second after 2022, and remains incomplete. VoxUkraine’s White Book of Reforms records over 1,650 reform oriented legal acts adopted since early 2015, alongside nearly 40 anti-reform measures,b32496e706a5 and the OECD’s 2025 integrity and anti-corruption review of Ukraine reports a fall in citizens’ reported corruption experiences from 70% in 2015 to 15% in 2021 (VoxUkraine 2025; OECD 2025; both as compiled by Econ4UA 2025).

The period was equally defined by resistance. Political elites repeatedly sought to limit NABU and SAPO independence, delayed the creation of the HACC, and contested the NACP’s mandate. In October 2020, the constitutional court struck down criminal liability for false asset declarations while several of its own judges were under investigation, a decision reversed only by legislative workaround (Reznik et al. 2023; Lough and Dubrovskiy 2018).

An assessment of the national 2014–2017 anti-corruption strategy, adopted by the Verkhovna Rada in October 2014, found that only 83 of 210 tasks were completed on time, with measures tied to EU visa liberalisation and IMF programmes implemented most efficiently (Marchuk and Nesterenko 2017). This was an early demonstration of the pattern, recurrent across all periods, in which external leverage has driven formal adoption while practical implementation lags.

Full-scale Russian invasion (2022–present)

Nonetheless, after Russia’s full-scale invasion of Ukraine in 2022, the array of anti-corruption institutions proved resilient. NABU opened 456 investigations and sent 54 indictments in 2022 alone, high-level verdicts continued, and leadership competitions for NABU and SAPO were completed during wartime (Kalitenko 2023).

EU candidate status, granted in June 2022, anchored a new round of conditionality. By mid-2026, the HACC had cut average case processing times from 618 days (as of 2024) to 405 days (TI Ukraine 2026b)

At the same time, martial law introduced structural constraints: direct contracting cut published tenders by 75% and published contracts by 61% against 2021 levels, and state registers and asset declarations were suspended for extended periods (Kalitenko 2023). ProZorro activity more broadly fell by over half after the invasion and did not recover, with a shift toward offline procurement that raised prices even as completion rates improved (Klymak and Vlandas 2024). State registers and asset declarations were also suspended for extended periods, though mandatory ProZorro use was restored for most non-defence procurement by June 2022, while the September 2023 dismissal of the defence minister over a procurement scandal suggested oversight, though relaxed, had not disappeared (Brick Murtazashvili et al. 2024).

Ukraine’s Open Budget Survey score fell from 65 in the 2021 survey to 38 in the 2023 survey, which assessed the period through December 2022, after seven years of steady improvement since 2015, as several key budget documents were withheld or produced for internal use only under martial law (International Budget Partnership 2024).

The period has also seen the sharpest institutional confrontation to date. In July 2025, parliament adopted legislation subordinating NABU and SAPO to the prosecutor general, although this was reversed within days following the largest street protests since the full-scale invasion and coordinated pressure from the EU, IMF and G7 partners (TI Ukraine 2026b; Popova 2026). This was not an isolated episode. Markovska, Serdiuk and Soldatenko (2025) document an earlier instance of the same dynamic in President Zelensky’s August 2023 proposal to treat corruption as equivalent to wartime treason, arguing that Ukrainians’ sharply elevated wartime trust in the presidency and security services is emotionally rather than institutionally grounded and can be invoked to bypass ordinary scrutiny. These episodes illustrate both the fragility of Ukraine’s independent institutions and the continued potency of the combination of civic mobilisation and international leverage, a combination this paper returns to when discussing enabling and hindering conditions. Through mid-2026, Ukraine’s anti-corruption institutions continued to expand their enforcement track record, and parliament adopted a new EU aligned public procurement law. At the same time, targeted pressure on NABU detectives intensified, while the Ministry of Justice continues to block the NACP’s mandatory external audit (TI Ukraine 2026b).

What does the empirical evidence say? Findings by intervention type

Note on the evidence base

Literature was identified through keyword searches, in both English and Ukrainian, on Google and Google Scholar (up to ten pages of results per query), followed by a second round of terms once an initial set of interventions had been identified, and snowballing through the reference lists of located studies. Given the volume of Ukrainian language legal and doctrinal literature returned by concept level searches, searches were also directed at the publication pages of specialised research organisations (including CPLR, DEJURE, CHESNO, AntAC, VoxUkraine, KSE and ACREC) rather than relying on bibliographic databases alone. Two selection criteria were applied: credibility of the publisher and alignment with topic. As with any Helpdesk format review, this was conducted over a bounded period with limited resources and should not be read as fully exhaustive.

The search identified around 17 studies that evaluate specific anti-corruption interventions in Ukraine using robust empirical methods, supported by around 35 descriptive and contextual sources used for interpretation. The empirical studies cluster heavily: most examine the post-2014 period, while public governance reforms (procurement, auctions, decentralisation) and awareness raising are far better covered than any other category of anti-corruption intervention.

Throughout this section, each study is labelled according to a three-tier hierarchy of confidence developed for the purposes of this Helpdesk Answer. This hierarchy is indicative, not absolute; as such, the reliability of a study’s findings depends not only on which method was used but on how rigorously it was implemented and the robustness checks the authors applied (Maslen and Bergin 2026). For example, a well-triangulated process tracing study can be more reliable than a poorly specified quasi-experimental design, just as a badly designed experiment is no more trustworthy than a weak observational study. Tier placement here reflects study design, not an independent audit of execution; where execution problems in a specific study were identified, they are flagged in the discussion of that study rather than reflected through a tier downgrade.

  1. Quasi-experimental and experimental studies (difference-in-differences, regression discontinuity, randomised and field experiments) are deemed to provide the strongest warrant for causal claims.
  2. Process tracing and rigorous qualitative comparative studies are considered to provide medium confidence as they cannot isolate causal effects with the same precision but are often better at identifying mechanisms and conditions.
  3. Programme evaluations and structured assessments by organisations (or those affiliated with them) implementing anti-corruption measures are useful for context but carry a risk of favourable framing and are treated accordingly. This differs from the lowest tier of purely descriptive and monitoring sources as programme evaluations typically involve systematic data collection against a defined methodology, even where that methodology is set by an interested party, whereas descriptive and monitoring sources make no such methodological claim at all. Both are treated as context rather than evidence of effect, but the former offers structured detail on implementation and mechanisms that the latter cannot.
  4. Descriptive and monitoring sources sit below this hierarchy and are cited only as context.

One caveat applies across the board. Many of Ukraine’s reforms were introduced concurrently, especially in 2015–2016, when procurement reform, decentralisation, police reform, e-declarations and the creation of the anti-corruption bodies overlapped. Attribution of observed changes to any single intervention is therefore a systemic challenge. The procurement literature confronts this problem explicitly, but it applies to every category reviewed below.

This category covers the establishment and performance of the specialised anti-corruption bodies, the HACC, national anti-corruption strategies and the broader institutional framework, including its anchoring in EU integration.

Focacci, Kovac and Spruk (2025) (quasi-experimental, highest confidence) provide the only study in this category with a counterfactual design. Using machine-learning supported estimates of institutional quality for Ukraine’s 28 regions over the period 1996–2020 and a synthetic control style comparison, they compare Ukrainian regions against similar regions in central and eastern European countries that joined the EU from 2004 onwards. Their central finding is that Ukrainian regions suffered a measurable, compounding institutional quality cost from staying outside the EU accession process, and that the failure to open an accession pathway in the early 2000s constituted a missed window of opportunity whose costs grew over time. The study directly supports treating EU integration, and not only itemised conditionality, as an enabling condition, though it evaluates a macro-level trajectory rather than a specific intervention.

Pop-Eleches and Robertson (2024) (quasi-experimental, highest confidence) exploit the staggered 2015–2016 rollout of the new patrol police across Ukrainian localities with a difference-in-differences design built on a two-wave nationally representative panel survey. The reform, a wholly new institution with officers recruited through a competitive process that combined skills testing with integrity vetting by mixed panels including civil society representatives, and better paid than their predecessors, improved perceptions of police corruption by roughly half a standard deviation, an effect that proved durable. Spillover to perceptions of the presidency and parliament was much smaller and dissipated within about six months, and the reform produced no measurable electoral gain for the incumbent government. The study is discussed further under the enforcement section, but its institutional lesson belongs here: creating an entirely new institution, recruiting fresh personnel with no prior service history or local patronage ties, rather than reconstituting the old militia’s existing local staff, can shift entrenched perceptions, yet the political return to reformers is weak, narrow and short-lived.

Marchuk and Nesterenko (2017)(programme evaluation, contextual confidence) systematically assessed implementation of the 2014–2017 anti-corruption strategy over its first two years (2015–2016, as reported May 2017): 83 of 210 tasks implemented, 65 partially in progress and 66 unaccomplished despite lapsed deadlines. Their most policy relevant finding is the pattern of completion: legislative and regulatory tasks had the highest completion rates, practical implementation was far weaker and tasks tied to EU visa liberalisation and IMF conditionality were implemented most efficiently. Overall political will was characterised as weak.

TI Ukraine (2020)(structured programme assessment, contextual confidence) evaluated the capacity, governance and interaction of the five anti-corruption bodies from 2015 to 2020 using a methodology derived from Transparency International’s National Integrity System:17bae8ef9dbb scenario-anchored 1–5 scoring split between law and practice, 50 interviews and external peer review. Overall scores placed the HACC highest (4.0), followed by NABU (3.6), SAPO (3.3), NACP (3.3) and ARMA (3.0), with the recurring pattern that legal frameworks scored well while practice lagged.

Kuz and Stephenson (2020) and Vaughn and Nikolaieva (2021)(descriptive and practice analysis, low confidence) document the design and early operationalisation of the HACC. The design feature they highlight, the Public Council of International Experts, with an effective veto over judicial candidates, was intended specifically to protect the court against political capture, and selection bodies mixing international and Ukrainian members have since been replicated across Ukraine’s institutional appointments.

Process tracing and comparative work (medium confidence) fills in the mechanisms. Zaloznaya and Reisinger (2020) derive a typology of how states “decouple” international anti-corruption prescriptions from domestic practice – sabotage, appropriation, intransigence and accommodation – and find that post-2014 Ukraine predominantly accommodated international norms while implementation remained contested. Králiková (2022) reaches a complementary conclusion on EU norm import through the association agreement, using a comparable process tracing approach: progress moved in fits and starts, and domestic actors – the balance of power between reformers and resisters – determined what was actually implemented, not external pressure alone.

Evidence gap: the institutional independence literature largely relies on process tracing and expert assessment. No study provides a counterfactual estimate of the contribution of the specialised anti-corruption institutions to documented outcomes.

Public governance reforms

This category, which includes reforms on procurement, public asset auctions, decentralisation and civil service reform, contains the paper’s richest and most policy relevant empirical evidence.

Public e-procurement (ProZorro)

Kovalchuk, Kenny and Snyder (2019) (observational regression with fixed effects, medium confidence, with caveats98c24b19f29a) provide the most careful independent estimate of ProZorro’s effects, using regression analysis on more than 77,000 above-threshold contracts before and after the platform became mandatory. For contracts that were competitively bid in both periods, ProZorro is associated with 3.5–5.8% lower prices and procurement roughly five to six days faster. Notably, once entity and goods type fixed effects are included, they find no statistically significant increase in the number of bidders, a challenge to the stronger competition claims often made for the platform. The authors are explicit that concurrent legislative changes, including a ban on above-estimate contracting and tightened timelines, make it difficult to isolate ProZorro’s own contribution.

EBRD (2025) and World Bank IEG (2021) (implementer-affiliated assessments, contextual confidence) report larger headline numbers: US$8.7 billion in savings since launch, growth from 14,000 bidding firms in 2014 to 140,000 in 2024, and registered bidders rising from 64,000 to 240,000 between 2016 and 2020.

The descriptive literature converges on the mechanism: ProZorro’s “golden partnership triangle” – a state-owned central database, competing private e-platforms and integrated civic monitoring through DoZorro – combined with the “everyone sees everything” transparency-by-design principle, embedded civil society oversight from its launch and secured the reform in legislative terms through accession to the WTO government procurement agreement. Nizhnikau (2022) (process tracing, medium confidence) adds an institutional legitimacy explanation for why this design survived where comparable reforms regressed. He contends that ProZorro’s early, visible results reduced the uncertainty facing state and non-state stakeholders and secured a co-productive coalition that continuously fine-tuned the platform, in contrast to reforms that lacked an equivalent early demonstration effect.

Klymak and Vlandas (2024) (quasi-experimental, highest confidence) analyse wartime procurement dynamics by analysing more than one million ProZorro transactions from January 2021 to October 2022 using a buyer, seller, month and industry fixed effects design. They find that purchasing activity fell by more than half after the invasion and did not recover, that the completion rate of tenders rose from 93% to around 95%, and that this came at a cost: a shift away from more competitive online auctions toward direct, offline procurement reduced the average number of bidders per contract and raised the ratio of winning to expected prices, meaning government buyers paid closer to or above their own estimates. The authors interpret this as evidence that wartime authorities traded cost-efficiency for speed rather than abandoning the platform. Brick Murtazashvili, Mylovanov, Shapoval and Murtazashvili (2024) (political economy analysis, contextual confidence) offer a complementary account of why ProZorro persisted. They contend that martial law provisions created carve outs for direct contracting, particularly for defence and security purchases, but the government reinstated mandatory ProZorro use for most other procurement in June 2022, and they cite the 2023 dismissal of the defence minister over a procurement scandal as evidence that oversight, though relaxed, did not disappear. An earlier, narrower assessment of the platform’s 2015–2016 rollout period found a temporary decline in competition and in the relative share of budget saved as ProZorro moved toward mandatory status (Dmytryshyn et al. 2018); the authors partly attribute this to customers learning to set more realistic budget estimates over time, and it does not extend to the sustained savings and bidder growth reported for the platform’s mature years by Kovalchuk et al. (2019) and EBRD/World Bank IEG above.

Public asset auctions (Prozorro.Sale)

Deininger, Ali and Neyter (2023) (quasi-experimental, highest confidence) using regression discontinuity in time around the 1 November 2021 mandate that shifted public agricultural land lease auctions to the Prozorro.Sale electronic platform –organised by local councils rather than the central cadastral agency (SGC) – and fixed effects regressions on 34,235 auctions from 2015–2022, find the reform raised lease revenue by around 175%, with placebo tests returning null results. The study’s value comes from its within-case comparison: an earlier 2018–2019 electronic pilot, also organised by the SGC, produced prices 12–19% below offline auctions, and SGC organised auctions returned 20–33% less than council organised ones. The SGC simultaneously held a monopoly on cadastral surveying, organised the auctions and could disqualify bidders or move auction times at short notice. It also administered a separate “free privatisation” channel through which land could be transferred to individuals at no cost, giving it the discretion to route valuable parcels away from competitive, revenue generating auctions altogether. The authors note that this channel was widely abused. The study offers strong comparative value by examining two electronic platforms implemented in the same country, sector and asset class but that yielded opposite results. It shows that digitisation reduces rents only where it removes discretion from officials with vested interests and where the auctions are organised by the body that directly benefits from higher revenue, which was local councils once the 2021 reform made them both organiser and beneficiary, unlike the SGC, which organised auctions but had no financial stake in their outcome. The authors estimate that enacting the 2021 reform four years earlier, in 2017 (a delay caused by a flood of amendments requiring individual parliamentary votes, COVID-19 disruption and political resistance including calls to veto the law) would have generated roughly US$500 million per year in additional local revenue.

Decentralisation and participatory governance

Ukraine’s decentralisation reform, launched in 2014, combined two distinct elements: an administrative dimension, amalgamating small municipalities into larger hromadas, with elected councils and mayors holding executive authority, and a fiscal dimension, devolving a substantial share of local income tax revenue, and direct budgetary relations with the central government. Arends, Brik, Herrmann and Roesel (2023) (quasi-experimental, highest confidence) exploit the staggered formation of these hromadas between 2015 and 2020 in a difference-in-differences design. Decentralisation increased trust in local government and participation in local elections, with effects persisting four to five years. The mechanism, they argue, was the participatory process of communities building new institutions, not fiscal devolution per se. The improvements in trust that citizens had in local institutions did not translate into higher levels of trust in national institutions. The authors credit these same community structures with underpinning local resilience after the full-scale invasion, drawing on wider observations of hromadas’ wartime performance, although rapid wartime centralisation may erode these gains. Keudel (2024) provides context on local anti-corruption capacity that helps interpret these findings.

Khutkyy and Avramchenko (2019) (mixed-methods independent evaluation, highest confidence) evaluated participatory budgeting across the 154 Ukrainian communities that had adopted it between 2015 and 2019, combining regression analysis of official data from 141 local governments with expert interviews in nine purposively selected communities. Their principal finding is conditional: participatory budgeting delivers its expected outcomes only where public activism and genuine local authority readiness coincide. Notably, structural design variables, like funding share, number of cycles, prior completion rates, showed no statistically significant relationship with voter turnout; pre-existing civic activism and proactive civic values were the strongest predictors of engagement. Expected effects on social cohesion and inclusion of marginalised groups were not confirmed, which the authors attribute to Ukraine’s individualised, social-entrepreneurship model of participatory budgeting rather than a deliberative one. The evaluation also documents a “revolving door” mechanism where those involved with participatory budgeting became public officials in multiple cities.

External independent testing (ZNO/EIT)

External independent testing (ZNO/EIT) for university admission is one of the earliest reforms identified, and the most cited for the post-Orange Revolution period. Klein (2014) (process tracing, medium confidence) interviewed reform stakeholders, including university rectors, and attributes the reform’s unusually successful implementation (in contrast to comparable reforms elsewhere in the post-Soviet space) to a short post-Orange Revolution window in which political will, domestic civil society pressure and external donor capacity building were aligned. Mishchuk, Krol and Bilan (2026) (three-wave nationally representative survey, n=5,123, with ordinal regression modelling, medium-high confidence) provide the first evidence on how the reform’s successor, the National Multisubject Test, is faring under wartime conditions. Trust in the testing mechanism is consistently highest among younger respondents relative to older age groups, including parents, and the survey identifies socio-economic and spatial factors – displacement, regional disparities – now shaping the perceived fairness of access. Neither study evaluates a discrete before/after policy change with a comparison group, so the underlying causal claim that ZNO/EIT reduced corruption in admissions still rests on descriptive statistics rather than a counterfactual design.

Evidence gap: no empirical studies were identified on civil service reform, including merit based recruitment. The closest available evidence is adjacent: Dimitrova and Mazepus (2026) (survey and survey-experimental data859b44056e5b) find a causal link between citizens’ evaluations of Ukrainian public administration reform and their trust in government. The study covers administrative service quality more broadly rather than corruption specifically, so it is noted here for completeness rather than as evidence for the synthesis of the enabling conditions described below.

Transparency and accountability initiatives

Setting aside the procurement platforms (ProZorro and Prozorro.Sale), the transparency and accountability literature on Ukraine is dominated by descriptive assessment, and its three most consequential interventions – DoZorro, beneficial ownership transparency and Diia – share a common gap: none has ever been subject to an independent evaluation with a counterfactual design. The one dedicated study identified in this category, Kobets et al. (2024) on the relationship between Ukraine’s “digital state” agenda and corruption prevention and detection, sits below the quasi-experimental tier and its findings should be treated as indicative. The strongest empirical evidence touching this category remains the participatory budgeting evaluation discussed earlier. The monitoring literature nonetheless documents a consistent wider trajectory across the category: Ukraine’s Open Budget Survey score rose steadily after 2015 before declining once wartime martial law and document classification took hold (International Budget Partnership 2024), and civil society monitoring of parliamentary transparency documents the specific constraints wartime openness now faces (CHESNO 2025).

DoZorro illustrates the pattern directly. Open Contracting Partnership’s 2020 reporting found that, over three years, civil society organisations using DoZorro had registered violations in over 30,000 tenders worth an estimated US$4 billion and helped correct violations in more than 4,200 tenders worth around US$500 million, corresponding to a reported “fix rate” of 14%. Earlier reporting from 2018 found that approximately 25% of all suspicious cases identified through the platform had been resolved, rising to 48% when calculated only among cases that had reached a resolution (Open Contracting Partnership 2018; 2020).

In 2015, Ukraine launched the world’s first public beneficial ownership register. Open Ownership (2022) (implementer network scoping review, contextual confidence) interviewed register users and implementers and found that public accessibility allowed civil society and journalists to combine the register with other datasets –including politically exposed persons lists – to identify undisclosed relationships that authorities alone had not uncovered, and credits this with driving incremental improvements in data quality over time. The same review documents persistent quality gaps: a 2019 EITI evaluation found only a small share of extractive sector companies had reported a natural person as the ultimate beneficial owner, with most instead reporting another company or no beneficial owner at all.

Diia, the digitalisation of public services, is by far the largest unevaluated intervention in this category (by public expenditure and by political attention) and most available evidence comes from studies commissioned by the same ministry that runs the programme. Civitta (2026) (government commissioned quantitative study, contextual confidence, significant caveats) estimates cumulative savings of UAH184 billion since 2020 from digitalising public services, including a 76% cost reduction on services identified as most vulnerable to corruption, based on before-and-after price comparisons for specific services (for example, online vehicle registration falling from UAH5,398 to UAH1,369) combined with interviews used to estimate typical bribe amounts under the previous paper based system. The E-Governance Academy (2021) reached similar conclusions using expert interviews and a user poll, but the poll’s sample size, sampling method and instrument were not disclosed, and it returns a uniform result: 100% of respondents reporting never having faced corruption using any Diia service, across every one of roughly fifteen documents and services assessed. This appears more like self-selection or leading design rather than a representative finding. Neither study uses a comparison group or controls for concurrent price and inflation changes, and both were commissioned by the Ministry of Digital Transformation or its close partners. The before-and-after price comparisons are informative but should not be read as causal estimates, and the E-Governance Academy’s self-reported corruption avoidance percentages should not be cited as standalone statistics.

Evidence gap: neither DoZorro, beneficial ownership transparency nor Diia have been subjected to an independent evaluation with a counterfactual design. Given how much of the enabling conditions argument rests on integrated civic monitoring and transparency-by-design, and how much donor and government attention digitalisation now receives, this is a conspicuous gap that an independent quasi-experimental study (exploiting, for example, the staggered municipal or service-by-service rollout of Diia services) could plausibly close.

Political finance integrity

This remains the most glaring gap in the review: no empirical study of any method was identified anywhere in the political finance literature. What exists is election cycle compliance monitoring. IFES Ukraine (2019, 2020) (election monitoring assessments, contextual confidence) document the 2019 presidential and parliamentary campaigns in detail: oversight bodies’ formal analysis of financial reports was described by the monitors as largely procedural rather than substantive, sanctions for violations went unenforced, wealthy donors and personal funds remained the dominant financing sources, and complex in-person donation procedures were found to disproportionately discourage individual donors to women candidates. These are useful for describing the compliance landscape but contain no evaluation of any specific intervention’s effect.

Evidence gap: no empirical studies meeting the evidence bar were identified for this category. Monitoring reports document the reform landscape and its persistent weaknesses: party finance reporting was suspended during the COVID-19 period and again after the full-scale invasion before being restored in 2023, leaving multi-year gaps in oversight; sanctions are assessed as disproportionate; and the POLITDATA reporting system has been dysfunctional (Biletskyi et al. 2024; Marchuk and Nesterenko 2017). In addition, restoring the reporting mechanism has not produced full compliance as more than 130 of Ukraine’s roughly 400 registered parties failed to submit required quarterly reports in 2025 (Judicial and Legal Newspaper 2025).

Enforcement and sanctions

Pop-Eleches and Robertson (2024) (quasi-experimental, highest confidence), introduced earlier in this paper, is the only study in this category with a counterfactual design, and it concerns the police rather than the specialised anti-corruption bodies. The authors note that the improvement in perceptions was driven by what citizens could observe at street level: new officers, new uniforms, new vehicles – the elements most easily dismissed as window-dressing actually signalled change most obviously. The authors draw the uncomfortable implication that the citizens are unlikely to register reforms of institutions, including parliament, that they cannot observe. Combined with the finding that reformers reaped no electoral return for their efforts, the study adds empirical weight to a simpler explanation for under-investment in anti-corruption enforcement: such reforms are costly and difficult, confront entrenched vested interests and, as this finding shows, do not reliably pay off electorally even when pursued, making them a poor claim on scarce political capital regardless of whether individual politicians have internalised this specific pattern.

The remainder of the enforcement literature is descriptive or doctrinal. Lough and Dubrovskiy (2018) (rigorous qualitative analysis, medium confidence) offer the period’s most influential assessment: by mid-2018, NABU had filed 207 cases yielding only 19 convictions and none at a high level, which they attribute to vested interest control of the judiciary and unreformed prosecution structures. Their broader conclusion – that closing the space for corrupt practices (restructuring Naftogaz, electronic VAT administration, ProZorro, banking clean-up) showed greater success than punitive enforcement – is one of the most cited claims in the Ukrainian reform literature and functions as a design principle in the enabling-hindering conditions section below.

Reznik et al. (2023) (doctrinal with an empirical element, medium-to-contextual confidence) document judicial resistance directly, including the constitutional court’s October 2020 decision striking down e-declaration liability, and analyse 50 court verdicts for false declarations, finding predominantly light sanctions.

Timofieieva (2025) (descriptive quantitative content analysis, contextual confidence) reviews 457 corruption related verdicts from the Odesa region between the start of the full-scale invasion and September 2024, categorised by offence article and sanction type, and concludes that sentences frequently fail to meet the principle of proportionality as fines and suspended sentences dominate even for offences directly enabled by wartime conditions. Wartime enforcement performance is tracked descriptively: 33+ high-level verdicts in the first year of the full-scale invasion, accelerating HACC throughput (Kalitenko 2023; TI Ukraine 2026b).

Evidence gap: after political finance, this is the most significant gap in the evidence base, and the most consequential given the centrality of NABU, SAPO and the HACC to Ukraine’s anti-corruption architecture and to donor investment. Prosecution and conviction outcomes lack any counterfactual analysis; no study estimates the deterrent effect of the enforcement architecture.

Preventive approaches

Ukraine’s electronic asset declaration system is among the most ambitious in the world, covering hundreds of thousands of officials with public disclosure. Yet its effectiveness has never been empirically evaluated. Monitoring by the OECD’s Istanbul Anti-Corruption Action Plan fifth round (OECD 2024) and Biletskyi et al. (2024) document the system as technically sophisticated but persistently constrained by verification gaps and by the NACP’s capacity limits; the suspension of declarations under martial law and their subsequent restoration created further discontinuities.

Conflict of interest regimes show the same pattern: a legal framework exists, enforcement is weak or inconsistent, and the evidence consists of normative compliance assessments rather than impact studies (Biletskyi et al. 2024). Lustration and vetting measures adopted after 2014 have likewise been examined only descriptively and comparatively, not evaluated for effect (Gricius 2019).

Whistleblower protection follows the same trajectory but has drawn considerably more legislative attention. The framework rests on the 2014 Law on Prevention of Corruption, which first established reporting person status, but the substantive reform was signed into law in October 2019 providing a financial reward of up to 10% of the sum recovered for disclosures involving damage above UAH10 million, protection from dismissal with back-pay for up to a year, free legal aid and psychological support, and anonymous, direct and external reporting channels. The reward mechanism took years to become operational in practice (the first confirmed payouts were reported only in October 2024) and a Unified Whistleblower Reporting Portal, intended to standardise status confirmation and enable court enforcement, followed as a further digitalisation step.

Research by the UNDP and the Government of Japan (2025) (implementer commissioned survey, contextual confidence) comes closest to empirical evidence on whether the whistleblower framework functions in practice, without meeting this Helpdesk Answer’s threshold for an impact evaluation.7cd180667675 Surveying 888 employees across 24 state-owned enterprises under the Ministry for Communities and Territories Development, it finds 85% of employees willing to report corruption but only 46% of enterprises with an actual whistleblower support policy in place, 31% of respondents fearing retaliation despite this, and the anti-corruption officer’s email (63%) used far more than the national unified portal.

Evidence gap: no experimental or quasi-experimental studies exist on asset declarations, conflict of interest management, lustration and vetting, or whistleblower protection in Ukraine. The UNDP survey is the closest available proxy for outcome data anywhere in this category, and it still covers only two institutional settings.

Awareness raising and education

This is the most experimentally studied intervention category in Ukraine, and its findings are cautionary. They also sit squarely within a wider international pattern rather than reflecting anything distinctive about Ukrainian responses to anti-corruption messaging. Mixed and backfire effects from information campaigns have been documented elsewhere, including field experiments in Nigeria and Peru finding that anti-corruption messaging either had no effect or increased bribery (Cheeseman and Peiffer 2022; Incio and Seifert 2024), and a broader review concluding that injunctive norm-based campaigns typically produce only small, temporary effects (Erlich and Gans-Morse 2025).

Denisova-Schmidt, Huber and Prytula (2015) (field experiment, highest confidence) evaluated an information campaign among university students in Lviv using materials built on Transparency International definitions. The campaign produced modest positive effects on willingness to join anti-corruption activity overall, but a backfire effect for a key subgroup: students without prior exposure to corruption who learned from the materials how widespread it was became more tolerant of it, consistent with descriptive norms normalisation (“everyone does it”).3aac57ab933c

A follow-up study by the same authors (Denisova-Schmidt, Huber and Prytula 2019) (online randomised controlled trial, highest confidence) tested three different anti-corruption videos against a control among a much larger, national sample. Only one, an emotionally engaging narrative following a family harmed by a chain of bribery culminating in a birth injury, produced a robust negative shift in attitudes toward corruption, with viewers more likely to call it a crime and to see it as damaging to health, safety and the economy, a result that held up under the study’s own correction for multiple hypothesis testing. The other two videos, a statistics led news style segment and a piece on the ethics of academic cheating, showed no such effect and, on measures that did not survive the same correction, nudged toward greater tolerance of corruption instead. The comparison suggests that message format, not just message content, determines whether awareness raising helps or backfires.

Borcan, Grabher-Meyer, Heger and Patel (2023) (lab-in-the-field experiment, highest confidence) tested integrity training with Ukrainian law students, measuring behaviour through an incentivised bribery game. Training significantly improved attitudes. Behaviour, however, changed only when trained participants knew they were interacting with other trained participants. Individual conviction was not the operative mechanism; shared expectations were.

Gorodnichenko, Sologoub and Fedyk (2026) (randomised controlled trial, highest confidence) ran information treatments on a representative sample of more than 7,000 Ukrainians during wartime. They document a large and persistent gap between experienced corruption, which fell sharply between 2015 and 2021, and perceived corruption, which remains stubbornly high. Informing citizens raised their perceptions about the government’s willingness to counter corruption but did not reduce overall corruption perceptions, information about the scale of corruption had no significant effect, and neither treatment passed through to prosocial behaviour. Perceptions, in short, are sticky even when reforms are working.

Khambekova, Los and Huss (2021) (qualitative, 242 CSOs, medium confidence) turn to how the organisations delivering anti-corruption messaging frame the problem, with implications for their own effectiveness. Across 242 regional anti-corruption civil society organisations (CSOs) interviewed from 2017 to 2019, those framing corruption systemically – as an institutional and collective action problem – cooperated with authorities, used coalition building and advocacy, and reported faster results. Those framing it as individual moral failure defaulted to “shame and blame” tactics and rated their own effectiveness as low. The individual moral failure framing was 1.5 times more prevalent than the broad one.

Conditions enabling and hindering anti-corruption reform success

This section draws on the empirical and contextual evidence reviewed earlier in this paper to identify cross-cutting conditions, organised under the rubrics of political, social and capacity factors. A note on approach: most studies do not directly test enabling or hindering conditions; they identify them as mechanisms or contextual explanations for their findings. This section therefore synthesises across studies rather than reporting the output of any single one, and it flags where a condition rests on strong empirical evidence and where it rests on descriptive consensus. It also addresses whether the conditions identified are genuinely specific to Ukraine or generic enabling conditions found anywhere. In most cases the mechanism is generic, but the Ukrainian evidence gives it unusual precision or scale, and a small number of conditions, noted below, are distinctively Ukrainian in form.

Political conditions

Enabling

International pressure and EU/IMF conditionality: this is the most consistently documented enabling condition, with an equally consistent qualification. Measures attached to EU requirements (visa liberalisation, the association agreement, candidate status) and IMF programme conditions show the highest implementation rates across every period examined. Marchuk and Nesterenko (2017) find EU and IMF linked tasks were implemented most efficiently under the 2014–2017 strategy, Focacci et al. (2025) quantify the institutional cost of remaining outside the EU integration process altogether. The attempt to introduce Law No. 12414, which would have undermined the independence of NABU and SAPO, triggered mass street protests and a strong response from the European Commission, which explicitly linked the issue to Ukraine’s accession prospects. A subsequent law restoring the independence of both institutions was adopted unanimously, by 331 votes, later that same week (Popova 2026). Process tracing research shows domestic power balances determined what was actually implemented (Králiková 2022), Ukraine’s typical response to international prescriptions was formal accommodation (Zaloznaya and Reisinger 2020), and independent monitoring of delivery on EU accession commitments continues to find completion rates far below adoption rates (TI Ukraine 2026a; see also Huss 2020). This paper therefore codes conditionality as a mixed condition: powerful, but biased toward formal compliance.

The European Court of Auditors’ (2021) review of EU anti-corruption support for the 2016–2019 period found that, despite substantial funding, grand corruption remained a central problem. The commission’s own assessments treated benchmarks as “achieved” on the strength of legislative adoption alone in cases where implementation had not followed, and only 1,005 of the more than 900,000 asset declarations submitted in 2019 were fully checked (0.1% rate) of which 674 showed irregularities. The system had produced no convictions for illicit enrichment. Comparatively, Popova (2026) finds Ukraine’s corruption levels are not exceptional against Bulgaria, Romania and Croatia at equivalent stages of their own EU accession, and argues that domestic co-ownership (pre-existing institutions combined with mobilised civil society) is what sustains reform under conditionality, a comparative reading consistent with the sandwich strategy evidence discussed under social conditions.

Windows of opportunity: the post-Orange Revolution and post-Maidan moments both disrupted elite networks and temporarily expanded reform capacity, and the empirical record treats their use or missed opportunity as consequential. Focacci et al. (2025) frame the unexploited early-2000s window as carrying lasting institutional costs; the density of durable reform achieved in 2014–2016 shows what an exploited window yields. The July 2025 crisis demonstrates that windows also reopen under pressure and can close again quickly. Popova (2026) reads the episode as evidence that domestic co-ownership and external leverage operate jointly even under wartime conditions.

Administrative cohesion: the comparative evidence adds a condition often missing from Ukraine specific accounts: Nasuti (2016) argues that Georgia’s post-revolution success depended on a uniformly reform oriented administration, while Ukraine’s coalition cabinet distributed posts across parties with divergent backgrounds and priorities, producing inconsistent implementation. Political concentration mattered, but only because it let a cohesive, reform oriented team implement without internal veto players. Cohesion of the administration, not concentration of power per se, was what translated into reform.

Hindering

Political elite resistance and institutional capture: this was the most consistent hindering condition across all three periods. The descriptive and monitoring literature documents sustained attempts to limit NABU and SAPO independence, delay the HACC and capture the NACP (Marchuk and Nesterenko 2017; Biletskyi et al. 2024), with the July 2025 episode its sharpest recent expression and continued pressure on NABU detectives recorded through 2026 (TI Ukraine 2026b). Oligarchic spoilers are a recurring feature of these accounts; rather than disappearing, they have historically adapted their strategies to changing constraints, shifting from direct political office toward indirect channels of influence such as party financing and media ownership (Markus and Charnysh 2017; Pleines 2016).

Judicial capture: distinct from generic elite resistance, and Ukraine specific in a particular respect, the country’s constitutional court (not merely lower courts susceptible to bribery) became a vehicle for capture, striking down the e-declaration system in October 2020 while its own sitting judges were under scrutiny (Reznik et al. 2023). Lough and Dubrovskiy (2018) attribute NABU’s early conviction failures to court level capture more broadly. Two waves of judicial reform, after 2014 and after 2022, remain incomplete.

The legislative-versus-implementation gap: formal adoption of laws far outpaces practical implementation in every period for which measurement exists: 83 of 210 strategy tasks were completed by 2017 (Marchuk and Nesterenko 2017), roughly 7.4% of measures under the 2023–2025 state anti-corruption programme were completed on time as of late 2023 (Biletskyi et al. 2024), with low delivery rates on EU accession commitments in current monitoring (TI Ukraine 2026a). The gap is the domestic mirror of conditionality’s compliance bias.

Wartime concentration of power: martial law provisions enabling direct contracting, security based data classification and expanded executive authority create structural conditions for corruption that are difficult to monitor or contest. The measurable expression is the post-2022 decline in open budget and procurement transparency scores after seven years of improvement (International Budget Partnership 2024; Kalitenko 2023). Zaloznaya (2026) argues that when institutional guardrails are weakened by external crises, such as war, the combination of strong civic pressure and an aggressive anti-corruption agenda can be co-opted by populist actors, sidelining quieter constructive reformers, which is a hindering dynamic that cuts across the political and social categories. Markovska, Serdiuk and Soldatenko (2025) document a related risk using survey data from Kharkiv (2013–2023) and national polling. They show that trust in the president, the army and the security services rose sharply after the invasion even as concern about elite level corruption persisted, and they argue this trust is emotionally rather than institutionally grounded and can be invoked to deflect scrutiny, pointing to President Zelensky’s August 2023 proposal to treat corruption as equivalent to wartime treason as an example of the risk in practice. Maletova and Utkina (2026) identify defence procurement and humanitarian aid as the sectors where this risk has materialised most concretely under martial law, and recommend finalising leadership at NABU and ARMA and establishing a civil society led corruption hotline for humanitarian sector reporting as partial remedies.

Social conditions

Enabling

Civil society integrated into accountability mechanisms: Ukraine’s civil society is exceptionally active by regional standards, and the evidence indicates that its effect is greatest when it is built into reform architecture rather than commenting from outside. For instance, the DoZorro monitoring was built around ProZorro from its launch, civic representatives sit on oversight commissions for Prozorro.Sale, and civic mobilisation combined with international pressure reversed the July 2025 law. This two-sided pressure – bottom-up civic mobilisation plus top-down international leverage – can be described as a deliberate “sandwich” strategy. Keudel, Grimes and Huss (2023) explain this mechanism at the local level: sustained political will for reform emerged where officials, civil society and business built communication coalitions and shared narratives, not where formal mandates existed alone. A larger companion study drawing on the same 242-interview dataset across 57 cities and towns finds this operates at the organisational level too. The regional anti-corruption CSOs that achieve concrete impact are those that convincingly resolve either a capacity deficit (financial and human resources) or a legitimacy deficit (a credible support base), rather than both at once, since in practice few organisations manage both (Bader et al. 2019). Local political will is accordingly important but not decisive on its own: several regional capitals where every interviewed CSO rated local political will as absent nonetheless reported medium to high success, and Huss et al. (2020) trace this to the structure of local elite competition rather than to will. Decentralised systems, where several elite groups compete for access to public resources, created more openings for transparency and answerability than either monopolised or highly cooperative local systems.

Participation as the active ingredient: two of the strongest studies in the review converge on this point from different sectors: decentralisation built trust through the participatory process of institution building rather than fiscal transfers (Arends et al. 2023), and participatory budgeting produced engagement only where pre-existing civic activism and receptive authorities met; as such, structural design variables alone predicted nothing (Khutkyy and Avramchenko 2019). The post-Maidan movement of activists into public office, documented concretely in the participatory budgeting evaluation and hypothesised as a reform sustaining mechanism in the research programme underlying Keudel et al. (2023), deserves a qualified mention as a partially evidenced enabling condition.

Street level visibility: Pop-Eleches and Robertson (2024) show that citizens register and give credit to reforms they can observe visually; the visible elements of police reform carried the signal, and the effect was uniform across partisan groups. For communication and sequencing strategy, this is one of the most directly actionable findings in the review.

Hindering

Sticky perceptions and the perception–experience gap: experienced corruption fell sharply between 2015 and 2021, yet perceived corruption barely moved, and information treatments cannot close the gap (Gorodnichenko et al. 2026). This matters practically: public and international judgements of reform success track perceptions, and the political capital and electoral benefits to be derived from reform is accordingly weak. Pop-Eleches and Robertson (2024) measure that return as narrow, small and fast-decaying, which in turn feeds the political will problem.

Messaging that backfires: Denisova-Schmidt et al. (2015) and Borcan et al. (2023) jointly show that raising awareness of how widespread corruption is can increase tolerance among previously unexposed groups, and that training changes behaviour only with peer-norm reinforcement. How anti-corruption messaging is designed matters as much as whether it is delivered.

Capacity conditions

Enabling

Transparency-by-design and multi-stakeholder architecture: the procurement evidence consistently identifies design, not digitisation, as the active ingredient: a state-owned database with competing private platforms and integrated civic monitoring, with all data public in real time (Kovalchuk et al. 2019; EBRD 2025). The reconstruction policy literature argues for carrying the same principle into recovery frameworks, including the DREAM platform (Becker et al. 2022; Richter 2023, 2024).

Separation of functions and aligned incentives: Deininger et al. (2023) likewise emphasise the centrality of careful design: electronic auctions underperformed when the SGC organised them and succeeded when organisation moved to local councils whose revenue incentives matched the reform objective, with the system administrator separated from competing trading platforms, immediate disclosure of bidder identities, open-source code and an oversight commission spanning government, business and civil society. Because both versions of the reform ran in the same country, sector and asset class, the comparison isolates incentive alignment, not digitisation, as the decisive variable. This within-case contrast sharpens the generic principle that what matters is removing discretion from the stakeholders with vested interests, not the technology itself.

International participation in appointments: the hybrid selection model mixing international and Ukrainian members, pioneered for the HACC’s judicial selection and since extended to other bodies,a49ed25faf41 is documented as a deliberate anti-capture design (Kuz and Stephenson 2020) and regarded by practitioners as a Ukrainian institutional innovation with transferable potential.

Hindering

Chronic understaffing and monitoring gaps: the NACP’s policy coordination and monitoring functions have been constrained by understaffing throughout. At one point, five staff were monitoring more than 1,000 measures across 95 agencies (Biletskyi et al. 2024, drawing on OECD monitoring), while the agency’s mandatory external audit remains blocked (TI Ukraine 2026b). Wartime mobilisation has compounded this gap throughout, as government bodies more broadly have lost staff to the war effort since 2022.

Uneven civil society capacity: the enabling role of civil society is not uniform. The majority of regional anti-corruption CSOs operate with a narrow, punishment focused framing associated with low effectiveness, while access to international funding and recognition concentrates in Kyiv (Khambekova, Los and Huss 2021). Human capacity has tightened further since 2022, with widespread understaffing across both government bodies and CSOs as personnel were diverted to the war effort.

Weak follow-through from legal frameworks to enforcement: asset declaration verification, conflict of interest enforcement and whistleblower protection all combine sophisticated legal frameworks with weak or inconsistent application, which was a recurring finding across OECD monitoring, European Commission reports, GRECO recommendations and civil society shadow reporting (Biletskyi et al. 2024).

Jurisdictional fragmentation: overlapping mandates among NABU, the State Bureau of Investigation, the security service (SSU), the Bureau of Economic Security and the national police produce parallel investigations, lost evidence and inadmissible materials. SSU involvement in anti-corruption investigations, for which it holds no mandate, is identified as a particular problem (Biletskyi et al. 2024). A broader institutional-mapping exercise by Bondarenko et al. (2021) reaches a similar conclusion: anti-corruption competencies have been dispersed across nearly every level of state and local government, and this dispersion and duplication of powers, rather than any single body’s weakness, is itself identified as complicating measures to curb corruption. This jurisdictional overlap gives political resistance a second, quieter channel; the same elite pressure that shows up as direct legislative attack under political elite resistance and institutional capture (above) can also work through routine bureaucratic interference here, such as parallel investigations, diverted evidence, agencies with no clear mandate stepping into NABU’s cases, rather than through open confrontation.

Finally, the design principle that emerges most strongly across the capacity evidence is Lough and Dubrovskiy’s (2018): restricting the space for corrupt practices, through system redesign, transparency and the removal of discretion, has shown greater success than punitive enforcement in Ukraine. The procurement and auction evidence corroborates it; the enforcement track’s two-decade struggle against judicial capture illustrates its converse (see judicial capture, above).

Evidence gaps and directions for further research

The evidence base reviewed here supports several conclusions, but its boundaries need to be stated.

  • How much of Ukraine’s specificity drives these findings is unresolved. This paper draws heavily on Ukraine focused studies where they exist, but robust experimental and quasi-experimental evidence from other countries (on transparency-by-design, integrity training or e-procurement, for instance) has not been systematically weighed against them. Whether such findings would replicate in Ukraine, or whether something about its institutional and political context would alter them, is an open question this paper cannot resolve.
  • Most empirical studies examine the post-2014 period. The 2005–2014 period is largely unexamined empirically; what is known about it comes from comparative case analysis (Nasuti 2016) and descriptive sources.
  • The post-2022 period is the least studied. Evidence on wartime reforms, reconstruction integrity and the effectiveness of candidate status conditionality is almost entirely descriptive, with Gorodnichenko et al. (2026) the main experimental exception.
  • Enforcement and sanctions lack any rigorous counterfactual evaluation of NABU, SAPO or HACC performance. Given the centrality of these institutions to Ukraine’s anti-corruption architecture and to donor investment, this is the most consequential single gap.
  • Political finance lacks empirical evaluation studies; whistleblower protection has only the UNDP implementer commissioned survey, which falls short of a full impact evaluation.
  • The DoZorro civic monitoring mechanism, NACP’s e-declarations system, beneficial ownership transparency and the digitalisation of public services lack independent impact evaluations, despite their prominence in reform narratives.
  • Given how much of the enabling conditions argument rests on integrated civic monitoring and transparency-by-design, and how much donor and government attention digitalisation now receives, this digitalisation gap is conspicuous; an independent quasi-experimental study exploiting, for example, the staggered municipal or service-by-service rollout of Diia services, could plausibly close it.
  • Early screening of Ukrainian language academic literature suggests much of it is concentrated in doctrinal, legalistic scholarship rather than empirical research.
  • Many reforms were introduced concurrently, especially in 2015–2016, making it difficult to isolate the effect of individual interventions. The procurement literature addresses this openly, and it applies across the board.

For researchers and evaluation commissioners, the highest value additions to this evidence base would be counterfactual evaluations of the enforcement institutions, an independent evaluation of DoZorro and of digitalised public services more broadly, potentially exploiting the staggered rollout of Diia services as a natural experiment, empirical work on the wartime and reconstruction period while administrative data remain retrievable, and evaluation designs planned into reconstruction programming from the outset rather than retrofitted.

Empirical evidence of anti-corruption effectiveness in Ukraine is still relatively thin. Rather than drawing definitive conclusions from this overview about which interventions work, policymakers should treat the findings as indicative rather than conclusive, and consider funding research to expand the evidence base on effectiveness.

Annex: Empirical studies by intervention type, period and confidence

The table below maps the studies meeting the empirical bar against intervention type, period covered, method and confidence tier. Intervention categories follow the taxonomy designed for this paper: formal legal and institutional reforms; public governance reforms; transparency and accountability initiatives; political finance integrity measures; enforcement and sanctions; preventive approaches; and awareness raising and education.

Table 1. Interventions that address corruption

Study

Intervention type

Period studied

Method

Confidence tier

Focacci, Kovac & Spruk (2025)

1 – Formal legal & institutional (EU integration)

1996–2020 (treatment year 2007)

Synthetic control and synthetic difference-in-differences; 26 treated regions (24 oblasts plus Kyiv City and Sevastopol-Crimea); in-space placebo

Quasi-experimental (highest)

Pop-Eleches & Robertson (2024)

1 – Institutional/5 – Enforcement (patrol police)

2015–2016 (post-Maidan)

Difference-in-differences on staggered rollout; two-wave panel survey, n=1,291 across 107 localities (29 reformed)

Quasi-experimental (highest)

Deininger, Ali & Neyter (2023)

2 – Public governance/3 – Transparency (Prozorro.Sale land auctions)

2021–2022 (post-Maidan)

Regression discontinuity in time

Quasi-experimental (highest)

Arends, Brik, Herrmann & Roesel (2023)

2 – Public governance (decentralisation)

2015–2020 (post-Maidan)

Difference-in-differences on staggered hromada formation

Quasi-experimental (highest)

Gorodnichenko, Sologoub & Fedyk (2026)

7 – Awareness raising

2022–present (full-scale invasion)

RCT, n=7,000+

Experimental (highest)

Denisova-Schmidt, Huber & Prytula (2015)

7 – Awareness raising

2015 (post-Maidan)

Field experiment, four public universities in Lviv, n=600

Experimental (highest)

Denisova-Schmidt, Huber & Prytula (2019)

7 – Awareness raising

2017 (post-Maidan)

Online RCT, four video arms, n=3,196

Experimental (highest)

Klymak & Vlandas (2024)

2 – Public governance (wartime procurement)

2021–2022 (post-Maidan, full-scale invasion)

Fixed effects regression, >1 million transactions

Quasi-experimental (highest)

Borcan, Grabher-Meyer, Heger & Patel (2023)

7 – Awareness raising

2021 (post-Maidan, pre-invasion)

Lab-in-the-field bribery game, 2×2 design; n=252 intermediaries; ITT and 2SLS/LATE

Experimental (highest)

Khutkyy & Avramchenko (2019)

3 – Transparency & accountability (participatory budgeting)

2015–2019 (post-Maidan)

Mixed methods: regression (141 LSGs) + expert interviews

Independent mixed-methods evaluation (highest)

Kovalchuk, Kenny & Snyder (2019)

2 – Public governance/3 – Transparency (ProZorro)

2014–2017 (post-Maidan)

OLS with ProZorro dummy plus entity and goods-code fixed effects, 77,000+ contracts; RD attempted and rejected after McCrary density test failed

Observational regression with fixed effects (medium)

Keudel, Grimes & Huss (2023)

Cross-cutting (political will, local level)

2021 fieldwork (post-Maidan)

71 interviews, 6 cities; comparative case study; coded semi-structured interviews

Rigorous qualitative (medium)

Khambekova, Los & Huss (2021)

3 – Transparency/ 7 – Awareness (CSO effectiveness)

2017–2019 (post-Maidan)

Qualitative content analysis, 242 CSOs, 23 regions; framing codes plus case studies

Rigorous qualitative (medium)

Nizhnikau (2022)

2 – Public governance (ProZorro institutional design)

2015–2021 (post-Maidan)

Single-case explanatory study with brief comparative juxtaposition

Rigorous qualitative (medium)

Klein (2014)

2 – Public governance (ZNO/EIT implementation)

1999–2014 (post-Orange Revolution and after)

Process tracing; 30 semi-structured interviews

Rigorous qualitative (medium)

Mishchuk, Krol & Bilan (2026)

2 – Public governance (ZNO/NMT wartime trust)

2021, 2023, 2025 (post-Maidan/full-scale invasion)

Three-wave representative survey, n=5,123

Rigorous quantitative survey (medium-high)

Lough & Dubrovskiy (2018)

5 – Enforcement and sanctions

2014–2018 (post-Maidan)

Qualitative institutional analysis of NABU caseload and prosecution outcomes

Rigorous qualitative (medium)

Table 1: Empirical studies meeting the review’s evidence bar, by intervention type, period, method and confidence tier. Source: author’s evidence matrix (August 2026).

  1. The White Book of Reforms is an annual VoxUkraine publication that tallies discrete reform oriented legislative and regulatory acts adopted since 2015; it is distinct from VoxUkraine’s separate, ongoing Reform Index tracker.
  2. The National Integrity System (NIS) is Transparency International’s methodology for assessing a country’s institutions relevant to safeguarding accountability and preventing corruption. It scores designated “pillars” (e.g., legislature, executive, judiciary, anti-corruption agencies, media, civil society) against standardised indicators split between legal framework and practice. See Transparency International, National Integrity System Assessments - Transparency.org.
  3. The authors originally intended a regression discontinuity design around the mandatory adoption threshold, but abandoned it after a McCrary (2008) density test indicated manipulation/sorting of entities around the cutoff, invalidating the design’s core assumption. They instead used fixed effects regression on a before/after comparison, a weaker identification strategy than true regression discontinuity, though still an improvement on simple observational comparison.
  4. By design, this study would meet the highest confidence tier used elsewhere in this paper. It is excluded from the evidence synthesis not for a methodological reason but for its scope as its subject is public administration reform and trust generally, not an anti-corruption intervention specifically.
  5. This is a descriptive cross-sectional survey of current attitudes and practices. It measures awareness, policy existence, fear of retaliation, and channel use at a single point in time, rather than testing the effect of a specific whistleblower protection intervention against a comparison group or a pre/post baseline. It is also implementer affiliated: UNDP and the Government of Japan jointly fund and provide technical support for this same ministry’s anti-corruption programming, the same affiliation caveat applied elsewhere in this answer to donor or implementer commissioned assessments (e.g., EBRD and World Bank IEG on ProZorro; Civitta and E-Governance Academy on Diia). Both factors place it below the threshold for inclusion as intervention effectiveness evidence, though its descriptive findings provide informative context.
  6. Descriptive norms convey perceptions of how common a behaviour is (“everyone does it”); injunctive norms convey whether it is socially approved or disapproved (“this is wrong”). The backfire here is a descriptive norms effect: learning how widespread corruption is normalised it for previously unexposed students, independent of any change in their view of its acceptability. The wider messaging literature finds comparably mixed effects from both norm types (Erlich and Gans-Morse 2025; Köbis et al. 2022; Cheeseman and Peiffer 2022; Incio and Seifert 2024).
  7. Since the HACC’s 2018–2019 selection process, the model of involving international experts in integrity sensitive appointments has been extended to other key anti-corruption institutions, including the selection of the NABU director (2022–2023 competition), the SAPO head (2020–2022 competition), and the NACP head (2023–2024 competition). Subsequent reforms adopted in December 2023 further strengthened the role of international experts in future SAPO leadership selection.

References

Transparency International

Cite this publication


Rechica, V. 2026. Anti-corruption reform in Ukraine: what does the empirical evidence say? Conditions enabling and hindering reform success since 2005. Bergen: U4 Anti-Corruption Resource Centre, Chr. Michelsen Institute (U4 Helpdesk Answer 2026:15)

Vlora Rechica

Disclaimer


All views in this text are the author(s)’, and may differ from the U4 partner agencies’ policies.

This work is licenced under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International licence (CC BY-NC-ND 4.0)

Photo