“How Corruption Works in the Public Sector in One Easy Lesson”, a PFM blog post from March 2020, presented a paper that looked at corruption through a public finance lens. The paper used a “follow-the-money corruption cycle” to demonstrate how corruption works at each stage of the PFM cycle. It showed how, in the extreme, public finance systems can be used as the foundation for formidable cultures of corruption within the very institutions entrusted by the public to help deliver good governance. Once corruption is institutionalised it is very hard to shift with piecemeal (‘whack-a-mole’) solutions and creates incentives for corrupt individuals to put themselves forward for leadership positions.
The “follow-the-money corruption cycle” is a practical way to understand this underbelly of public finance systems. A follow-up paper from Artificial Fiscal Intelligence presents the results of a new costing model, which draws on publicly available data to estimate the costs of corruption. While corruption is a huge topic in the world of governance, the dollar costs of corruption at an individual country or sector level are not readily available. The aim of the paper is to fill this critical information gap. This should ultimately help governments and stakeholders understand, benchmark and track performance in tackling the problems of inefficiency and corruption in government.
The model estimates that globally the costs of corruption and efficiency losses are US$4.5 trillion at the general government level (about 5% of world GDP), or $1.7 trillion at the budgetary central government level.
The paper also estimates that up to 30% or more of all losses for a country were accounted for in the budget stage of the PFM cycle. This is much larger than losses from weak procurement or accounting systems for example. It supports the idea that corrupt budgeting (the allocation of resources to vested interests – something Artificial Fiscal Intelligence describe as an auction) is a major driver of inefficiencies and corruption in high-risk environments. Audit practices are extremely weak in many developing countries, and this is another major contributor to efficiency losses and corruption vulnerabilities.
The methodology presented in the paper also allows us, for the first time, to estimate the losses in different sectors of government. Under the general model, national risks are assumed as a proxy for sector level risks. The paper provides estimates of losses in the security sector in different countries as an example of what the model can produce. At the budgetary central government level for Afghanistan, for example, the estimated losses are US$765 million per annum (representing 31% of security expenditures or 4% of GDP). Other country examples provided in the paper include Argentina, Australia, Ethiopia, Philippines, Russia, and Ukraine.
Country income status matters. On average, lower income countries were found to have larger losses (about 30% of budgetary central government expenditure) compared to high income countries (23%). As a proportion of GDP, however, low-income countries on average have smaller losses (5.5%) compared to high-income countries (7.4%). This result may reflect the relative smallness of their governments and economies rather than lower corruption vulnerabilities of their PFM systems.
Regional results are also driven by country income status. Sub-Saharan Africa had the highest losses as percent of budgetary central government expenditure at 27.8%, while North America had the lowest at 18.9%. As a percent of GDP, North America had a particularly low level of losses (2.6%) compared to the region with the highest estimated losses (East Asia and the Pacific) at 9.0%. Not surprisingly, highly resource dependent countries tended to score poorly on measures of inefficiency and corruption, as did fragile and conflict-afflicted states.
Data for some countries (e.g., China, Zimbabwe, and Somalia) were not available to include in the model. Corruption and efficiency losses in public corporations were also excluded. The collection of country specific data from field work would make the model more robust. Comments on the methodology and the results are currently being sought from interested stakeholders.