Making Sense of Conflicting Financial Inclusion Estimates

A policymaker eager to expand financial inclusion stands before two conflicting reports. One suggests that less than 30% of the adult population is financially included, based on surveys asking people directly about their account ownership. The other, derived from banking and mobile money administrative records, indicates that there are enough financial accounts to cover more than half of the population. Faced with these starkly different figures, the policymaker must decide how to allocate scarce public resources: should they fund new outreach campaigns or digital onboarding programs? Or has the job already been mostly done? Misinterpreting these numbers could mean either wasting funds on redundant initiatives–or worse, overlooking the millions who still remain outside the formal financial system.

This dilemma is not hypothetical. In numerous emerging markets, there are substantive gaps between estimates of financial inclusion, often between two or more leading demand-side surveys, and supply-side reports. In Pakistan, the most widely cited demand-side surveys – the World Bank’s Global Findex and the Karandaaz Financial Inclusion Survey (K-FIS) – paint diverging pictures. Findex reported a 20.9% inclusion rate in 2021, while the Financial Inclusion Insights (FII), later renamed the K-FIS, reported 21% in 2020 and 30% in 2022. To make matters more confusing, the State Bank of Pakistan’s supply-side reports diverge even further, showing enough active accounts for 43% of the population, and much larger total (active and inactive) account figures. 

These discrepancies are not just a technical issue for researchers. The divergence in financial inclusion numbers has attracted ongoing attention within Pakistan’s development and policy circles. They have pointed out that definitional inconsistency poses significant challenges for policymakers, as they may either underestimate success and misallocate resources, or overlook excluded populations whose needs remain unmet. 

So, what can explain these discrepancies? With Pakistan as our case study, we took a close look at the methodologies of each source (the two demand-side surveys and one supply-side report) and carried out analysis to determine to what extent methodological differences can account for the different numbers. 

Reconciling discrepancies between demand-side surveys

First, we compared the two demand-side surveys and their methodologies to see if we could reconcile their differing results. To begin, we had to address the fact that the Findex and FII/K-FIS surveys were collected at different points in time, so we took a linear time-weighted average of the FII/K-FIS estimates from Q1 2020 (21%) and Q3 2022 (30%) to match the timing of the Findex survey (Q4 2021). This exercise gives us an adjusted FII/K-FIS estimate of 27.3%, which still leaves a 6.4 percentage point gap in the demand-side estimates. 

What other factors can account for this remaining gap? We consider three explanations, first individually, and then collectively.

1. Accounting for differences in geographic sampling reduces the gap by 2.2 percentage points: The Findex and FII/K-FIS surveys don’t cover all the same regions. For example, Findex does not survey Azad Jammu and Kashmir (AJK) and Gilgit-Baltistan (population 5.5 million), while FII/K-FIS does.
2. Accounting for differences in census sampling and sample weights reduces the gap by 2.1 percentage points: The Findex and FII/K-FIS are sampled based on different Pakistani census rounds. They also re-weight their cluster samples differently to construct a nationally representative sample.
3. Accounting for differences in the definition of financial inclusion reduces the gap by up to 4.2 percentage points: Both demand-side surveys consider someone financially included if they have an active bank account or mobile money account. However, they differ in their approach to other financial institutions, and on whether they count only full-service accounts (credit plus savings and payments) or not. FII/K-FIS casts a wider net on the set of institutions they include, while Findex is slightly more inclusive on the type of account holding.
    

Successive effect of addressing discrepancies between FII/K-FIS and Findex methodology, as of Q4 2021. The last bar on the right is the final result after adjusting for geographic sampling, sampling method and differing definitions of financial inclusion. 

All three sources of discrepancy go in the same direction individually of narrowing the gap, collectively shifting the FII/K-FIS estimate down by 8.5 percentage points - reducing the 6.4 point gap to a 2.1 point gap in the other direction. However, this is likely an overcompensation due to ambiguity in the definition of financial inclusion between the two surveys. The remaining gap could also be due to random sampling variation, other methodological differences or errors in the assumptions behind our simple analysis.

In the end, methodological differences can largely explain the discrepancies between the demand-side surveys. But what about the supply-side numbers, which produced an even greater gap when compared with the demand-side figures?

Reconciling the demand-side and supply-side: Active mobile money accounts

The State Bank of Pakistan (SBP) compiles information on individual accounts based on reports from financial institutions. This data, which is shared quarterly, includes the total number of bank accounts (regardless of whether they’re active or not) and the total number of mobile money accounts, as well as the number of active mobile money accounts. Because the SBP data only differentiates between active and inactive accounts for mobile money, but not for bank accounts, we had to focus our analysis on mobile money in order to compare results with the demand-side surveys, which only include active accounts. 

In the K-FIS 2022, we estimate the rate of active mobile money account holding at 20.4%, which translates into at least 31.3 million active accounts (based on Pakistan’s over-15 population as of 2022). In contrast, the SBP reports 43.3 million active mobile money accounts as of Q3 2022–indicating a notable gap between the two reports.

The main way we reconcile the demand-side data with supply-side reports is to account for multiple account holding. For example, if someone holds active mobile money accounts with two different providers, they show up as two observations in the supply-side data, but only one financially included person in the demand-side surveys. 

The K-FIS demand-side data indicates that about 14% of account holders hold multiple active accounts. Taking this rate into account, we arrive at an adjusted estimate of 35.7 million active mobile money accounts based on the K-FIS data, compared to the 43.3 million reported by SBP as of Q3 2022. This adjustment narrows the gap significantly - by 37.6% – compared to the original figures.

To account for the remainder of the gap, we consider that our current approach with available data could still underestimate the number of accounts for at least three main reasons: 

1. Survey respondents might under-report account holdings in demand-side surveys (e.g., if they are unfamiliar with the brand identity of an account they hold, since the K-FIS survey asks about specific brands).
2. The K-FIS only allows for one account per provider per person, whereas it is possible for someone to hold more than one account with one provider (e.g., suppose they manage separate personal and business accounts, on two different phones).
3. The SBP considers a longer activity duration (180 days rather than 90 days as is typical in demand-side surveys) to define “active” accounts. This more permissive definition of active accounts would generate a larger total.
    

A call for methodological transparency and harmonized data collection

With the release of the Global Findex dataset for 2025 and new demand-side financial inclusion data, reconciling financial inclusion estimates remains as relevant as ever in Pakistan and in many markets globally. We have shown that demand-side figures can largely be reconciled once we account for some key methodological differences, and the gap between demand- and supply-side figures can be significantly narrowed. However, the discrepancies between different approaches highlight the need for methodological transparency and a more harmonized approach to data collection and reporting. 

Reducing confusion about the measurement of financial inclusion can help policymakers and financial institutions track progress and make more informed decisions - ultimately leading to more effective policies, better allocation of scarce resources and enhanced financial inclusion.