The Data That Doesn’t Exist: Sizing markets and measuring growth where the statistics are thin

There is a question we get asked, in one form or another, on almost every economic assignment in the Horn: what is the number? How large is this market, how many firms operate in this sector, what is the growth rate, what share is unserved. The client wants a figure they can act on. Reasonable.

The awkward part is that the figure often does not exist. The census is a decade stale or was never taken. The national accounts are estimates built on estimates. The sector the client cares about is mostly informal, which means it is mostly invisible to any register. And where a number does exist, two credible institutions publish two different ones, neither willing to defend theirs too hard in private.

This is the working reality of economic analysis in much of the region, and it separates the analysts who can operate here from the ones who cannot. The weak response is to find the least-bad official figure, cite it, and move on. The other weak response is to refuse to produce a number at all, hiding behind data limitations. Neither is useful to a client trying to make a decision. The actual skill is producing a defensible estimate in the absence of the data that would make it easy, and being honest about exactly how much weight it can bear.

Three ways the data fails

Thin-data environments fail in three distinct ways, and the fixes differ, so it is worth naming which one you are dealing with.

The first is absence. The number was never collected. No firm register, no household survey covering the variable, no administrative record. Here you are building an estimate from scratch out of adjacent evidence.

The second is staleness. The number exists but describes a country that has since changed shape, through conflict, displacement, currency movement, or growth. A ten-year-old figure in a stable economy ages gracefully. In a fragile one it can be actively misleading, and quoting it with a straight face is worse than admitting you do not know.

The third is contestation. Multiple numbers exist and they disagree, often because the institutions producing them have different mandates, methods, or incentives. The task here is not to pick a winner but to understand why they diverge, because the divergence usually tells you something real about the thing being measured.

Most assignments involve some mixture of all three. Recognising which failure dominates is the first analytical move, and skipping it is how people produce confident nonsense.

The discipline: triangulate, adjust, and show your uncertainty

When the direct measurement is missing, you reconstruct it from things that are measured, and you do it from more than one direction.

Take a concrete case: sizing the serviceable market for a small-ticket lending product in a regional economy with no reliable SME census. There is no register to count. So you triangulate.

Top-down, you start from something anchored, such as population and known economic activity in the target geography, and work down through plausible shares: what proportion are of working age, economically active, running or working in a micro or small enterprise, in the segment the product actually serves. Each step is a ratio, and each ratio is a judgment you can source, argue, and stress-test.

Bottom-up, you build from the ground: known operators already in the market, their reported or observable book sizes, branch footprints, transaction volumes where visible, and what that implies about total demand if you scale from the served to the addressable.

Then the two estimates have to meet. When top-down and bottom-up land far apart, that gap is not an embarrassment to be smoothed over. It is information. It usually means one of your assumptions is wrong, and finding out which one is where the real understanding of the market comes from. When they converge, you have something defensible, arrived at independently from two directions.

Around that, three further disciplines do the heavy lifting.

Proxy indicators, chosen carefully. Where the variable you want is unmeasured, find one that is measured and reliably tracks it. Mobile money activity, energy consumption, import volumes, and satellite-observed activity can all stand in for economic variables that no survey captures, provided you understand the relationship well enough to know where the proxy breaks.

Adjustment for informality. A number drawn from formal registers, in an economy that is mostly informal, is not a small underestimate. It can be off by a multiple. The adjustment factor is itself an estimate, and it deserves its own reasoning rather than a convenient round number pulled from the air.

Explicit confidence. Every serious estimate in this environment should travel with a statement of how much to trust it. A range, not a false-precision point. A note on which assumption the whole thing is most sensitive to, so the reader knows where it would break. An analyst who hands over a single confident number in a data-poor economy is either naive or hoping you will not ask.

The ethics of the estimate

Here is the part that gets skipped. The hardest discipline in thin-data work is not the arithmetic. It is knowing, and saying, what you do not know.

There is real pressure to launder uncertainty into precision, because precision is what gets a proposal funded and a recommendation accepted. A range makes a client uneasy; a single number feels like authority. Giving in to that pressure is how bad decisions get made on the strength of figures nobody should have believed, and it is how analysts quietly destroy their own credibility over time, because eventually the number gets checked against reality.

The credible move is to be precise about the imprecision. State the estimate. State the range. State what it rests on and where it would fall apart. A client can make a good decision under acknowledged uncertainty. What ruins them is a false certainty they were never warned about.

Why this is the work

Anyone can quote a figure someone else produced. The value in a data-poor economy is not access to numbers; it is the judgment to build one that holds, and the honesty to mark exactly how far it can be trusted.

That combination of method and candour is the intellectual signature of doing this work seriously. It is what a client is actually buying when the easy answer does not exist, which in this region is most of the time.

GeoAfrik Consulting Group provides research, data, and analytics for governments, investors, and development partners operating in the data-scarce economies of the Horn of Africa.