← Back to home Christian Chiroque
EN / ES
Financial intelligence November 2022

Methods for measuring money laundering: a typology

Measuring money laundering is a task of enormous methodological difficulty. This article proposes a systematic typology of the methods used to estimate its scale, assessing their applications and limitations.

Measuring money laundering

Measuring money laundering (ML) is a task of enormous methodological difficulty. Unlike other criminal phenomena, ML leaves no identifiable direct victims, deliberately blends with legitimate transactions, and operates with a markedly transnational character. Since the 1970s, when money laundering was formally framed as a global public problem, economists and criminologists have proposed varied approaches to estimate its magnitude, without reaching consensus.

This article seeks to propose a systematic typology of methods for measuring money laundering, evaluating their main applications and limitations.

1. Money laundering as an object of measurement

To understand the methodological obstacles, it is first useful to define the phenomenon along three dimensions: what is laundered, how and where, and who carries it out.

What is laundered. The object of the crime covers any asset —cash, real estate, vehicles, metals, crypto-assets— originating from illicit activity and whose origin is to be concealed (UN, 2000). This breadth complicates the construction of homogeneous metrics.

How and where. The classic process involves three phases: placement (entry into the economic system), layering (operations to hide the origin), and integration (reincorporation as apparently licit funds). However, not every case study goes through each phase, and the transnational dimension multiplies the complexity of tracing.

Who launders. There is no single profile. Predicate offenses range from drug trafficking and corruption to illegal mining and financial fraud, and each generates distinct laundering typologies.

2. A typology of measurement methods

Five families of methods are proposed, organized according to their analytical approach and the nature of the data they use.

2.1 Methods based on capital-flow discrepancies

These are top-down methods with a macroeconomic approach. They assume that illicit money flows generate detectable distortions in macroeconomic records. The Hot Money method (Cuddington, 1986) identifies unexplained errors and omissions among capital flows in the balance of payments, interpreting them as movements of "dark-origin" funds. The residual method (Colaco et al., 1985) calculates capital flight as the difference between capital inflows and the increase in official reserves. Dooley's method (1986) focuses on capital flows that generate no income recorded in the balance of payments. Finally, the trade misinvoicing approach detects under- or over-invoicing of exports and imports to transfer assets disguised as commercial transactions, being especially useful for trade-based money laundering.

Their main limitation is that they do not distinguish between legal capital flight (tax evasion, savings abroad) and properly illicit flows. In other words, they are highly sensitive to the statistical quality of the balance of payments, which in informal economies tends to be low.

2.2 Methods based on estimates of the underground economy

Also top-down, these methods seek to estimate the size of the unrecorded economy as a proxy for laundering. The Currency Demand method (Tanzi, 1996) analyzes excess cash in circulation beyond what structural factors explain: since launderers tend to operate in cash to avoid financial controls, excess cash demand is interpreted as an indicator of illicit activity. This approach was applied in Canada by FINTRAC (2015) and in Italy by Ardizzi et al. (2012) through ordinary least squares regressions, with variables such as provincial GDP per capita, unemployment rates, and electronic payment volume.

The method based on the discrepancy between economic-activity indicators and official GDP follows a similar logic, comparing electricity consumption or other physical indicators with recorded output. The most sophisticated model in this family is DYMIMIC (Dynamic Multiple-Indicators Multiple Causes), which treats the underground economy as an unobservable latent variable estimated from causes (level of regulation, tax efficiency) and effects (cash demand, criminality). Schneider (2007) applied it to 145 countries.

The central limitation of this family is that the underground economy includes both illicit activities and informal ones of licit origin (undeclared work, subsistence economy), so it systematically overestimates ML.

2.3 Methods based on bottom-up predictive models

Unlike the previous ones, these methods start from data on the proceeds of specific crimes to extrapolate volumes of laundered money. The most influential is the Walker Gravity Model, developed by John Walker (1995) for Australia's Financial Intelligence Unit. Walker estimated the proceeds of 14 criminal categories from police statistics and expert surveys, and modeled the flows of laundered money between countries by incorporating distance variables —not only physical, but cultural, linguistic, and institutional.

The model rests on assumptions with solid criminological intuition: countries with high corruption attract launderers; tax havens and jurisdictions with weak Know Your Customer (KYC) controls are preferred destinations; ethnic and commercial ties orient the flows. The 1994 estimate for Australia placed laundered money between 400 million and 4.5 billion Australian dollars per year. Unger et al. (2006) refined the model by adjusting the relative weight of cultural versus physical distance. The Walker & Unger (2009) version applied by UNODC (2011) estimated that global ML equals 2.7% of world GDP (1.6 trillion dollars), a figure the FATF still cites as its main reference.

Its main weakness is its dependence on heterogeneous criminal statistics that are difficult to verify across jurisdictions, as well as the sensitivity of results to assumptions about variable weights.

2.4 Methods based on measuring money-laundering risk

This group does not seek to size how much money is laundered, but to estimate how likely it is that ML occurs in a given environment. The Basel AML Index, published annually since 2012 by the Basel Institute on Governance, builds a composite index aggregating information from Transparency International, the World Bank, the World Economic Forum, and the FATF itself, weighting variables of regulatory quality, corruption, financial transparency, accountability, and political risk.

The illicit financial flows risk index from Tax Justice Network (Cobham et al., 2021) introduces a quantitative methodology that measures the vulnerability of each economic channel (trade, FDI, banking, portfolio) according to the financial secrecy of counterpart countries. It calculates three components: vulnerability (share of the flow coming from opaque jurisdictions), intensity (weight of those flows on GDP), and exposure (a combination of both).

The FATF, for its part, requires its members to conduct National Risk Assessments (NRA) under a conceptual framework that defines risk as the combination of threats, vulnerabilities, and consequences, though without prescribing a unified quantitative methodology. Some countries —such as the United Kingdom with MoRiLe— have developed subnational quantitative scores (Hopkins & Shelton, 2019).

The fundamental limitation of this family is conceptual: it measures the environment conducive to ML, not the phenomenon itself. A jurisdiction may have high risk and little actual laundering, or vice versa.

2.5 Methods based on the results of the anti-laundering system

This bottom-up approach sizes ML from what the prevention and sanctioning system itself manages to detect. The first route is the analysis of Suspicious Transaction Reports (STR/SAR): FIUs receive reports from obligated entities (banks, insurers, casinos, notaries, lawyers), and their cumulative analysis can support extrapolations about the scale of the phenomenon. Levi and Gold (1994) applied this methodology to the British case.

The second route is the analysis of convictions: studying successfully prosecuted cases makes it possible to identify typologies, sectors, amounts, and predicate offenses. Recent examples include Chile's Typologies Report (UAF, 2021), which analyzed convictions between 2007 and 2020, and the SBS's VI Report on Money-Laundering Convictions in Peru 2012-2023 (2024).

The limitations are severe: the universe of data is bounded by what the system detects and successfully processes, which implies a structural underestimation of the real phenomenon. Moreover, tracking the layering of operations can generate double counting in STR analysis. Ferwerda and Reuter (2019), comparing the NRAs of Italy and Switzerland, found that the former based its analysis mainly on expert opinion, while the latter incorporated more quantitative STR data.

Table 1 · Typology of money-laundering measurement methods
Method typeApproachMain methodsKey limitation
Capital-flow discrepancy Top-down / Macro Hot Money, Residual Method, Dooley, Trade misinvoicing Does not distinguish ML from other legitimate capital flight
Underground economy Top-down / Macro Currency Demand, GDP discrepancy, DYMIMIC Mixes ML with tax evasion and the informal economy
Predictive models (bottom-up) Bottom-up / Micro→Macro Walker Gravity Model (1995, 2009) Depends on low-quality criminal statistics
Money-laundering risk Composite / Indices Basel AML Index, Tax Justice Network, FATF-NRA Measures vulnerability, not the magnitude of the phenomenon
Anti-laundering system results Bottom-up / Operational STR/SAR analysis, analysis of convictions Underestimates real scale; detection bias

3. Conclusion

The absence of methodological consensus for measuring money laundering is not an anomaly, but a structural consequence of the nature of the phenomenon: unobservable, transnational, and deliberately camouflaged. Each family of methods described here contributes a partial and valid perspective —but insufficient on its own. Macroeconomic methods (flow discrepancies, underground economy) offer scale but little specificity; predictive models provide theoretical sophistication but depend on criminal data of heterogeneous quality; risk indices measure the environment but not the phenomenon; methods based on system results are the most reliable for typological analysis but underestimate the real magnitude.

The answer is not to choose a single method, but to build complementary approaches that combine these perspectives and that are calibrated to the particularities of each context.

References