Research policy is usually argued about as a single number: how much a country spends. This study asks a different question — if the amount stays the same, does it matter what the money is spent on?
Of five compositional dimensions the literature suggests should matter, only one withstands systematic testing
Of the five compositional dimensions the literature suggests should matter for absorptive capacity, only the division between capital and current expenditure withstands systematic testing. The remaining four are not shown to be irrelevant; they are shown to be unsupported by this panel — a weaker and more honest claim.
The association between a higher capital share and faster convergence is concentrated in economies well behind the frontier and disappears near it. Any recommendation therefore has to be addressed to a particular kind of economy rather than to the Union as a whole.
The estimates hold total research intensity constant. They speak to how a given research budget is allocated, not to whether that budget is large enough, and they offer no support for reducing overall research effort.
Replacing the conventional accession-based grouping of member states with groupings built from measured distance to the frontier removed most of the apparently significant results. Studies relying on the historical grouping should be read with that fragility in mind.
It is the economy furthest from the frontier in the sample and the only one of the four with an implied effect distinguishable from zero, while also being the one whose research statistics have the thinnest historical coverage.
A credible source of exogenous variation in research composition would move these results from association to causation. The European Commission funding share was examined as a candidate instrument and rejected as too weak; firm- or project-level data on equipment purchases would be a more promising route.
The same total research budget, allocated five different ways. Only the last split moves the speed of convergence.
| The split | What it compares | Predicts catch-up? |
|---|---|---|
| Who does the work | Firms vs. universities | ✕ No |
| Who pays | Government vs. business | ✕ No |
| Type of research | Basic vs. applied | ✕ No |
| Field | Engineering vs. social sciences | ✕ No |
| What the money buys | Equipment vs. salaries | ✓ Yes |
Why a higher equipment share helps a laggard economy but not a leader
Equipment brings in technology invented elsewhere. Buying it and learning to run it is the quickest way to import knowledge the country has not produced itself.
There is nothing left to import. The same money does more when it goes to the people who create new knowledge rather than to the machines that carry existing knowledge.
A further move is well inside familiar territory
Projected association of a +10 percentage-point capital-expenditure share with productivity growth, at each country's average distance from the frontier
| Country | Frontier gap (avg.) | Tercile group | Implied effect of +10pp | 95% CI |
|---|---|---|---|---|
| Czechia | 0.62 | Middle | +0.22 pp | [−0.68, 1.12] |
| Hungary | 0.72 | Far-from-frontier | +0.61 pp | [−0.34, 1.59] |
| Poland | 0.74 | Far-from-frontier | +0.79 pp | [−0.25, 1.89] |
| Romania | 0.80 | Far-from-frontier | +1.46 pp | [0.04, 2.94] |
It is not an argument for spending less on researchers. The comparison holds the total research budget fixed. It is about the mix, not the size.
It is a pattern, not a proven cause. The evidence comes from what 27 countries actually did between 1995 and 2024, not from an experiment. The projected gains are scenarios, not guarantees.
Eight sections, from the catch-up literature to country-level implications
For a country at Romania's distance from the European productivity frontier, the composition of research spending is one of the few levers that can be moved inside a single budget cycle, without new money and without institutional reform. That makes it unusually cheap to try relative to most growth policy — and, on the evidence assembled here, the only compositional lever with a defensible claim to work.
Of five composition dimensions and multiple candidate variables examined, only the capital-versus-current expenditure split survives a robustness protocol designed to guard against four specific threats: statistically distorted outlier economies, reliance on a historical political grouping as a proxy for economic structure, sensitivity to the choice of lag length, and confounding from the two major crisis episodes of the sample period. Economies furthest from the EU productivity frontier — prominently including Romania, Poland and Hungary — appear to benefit disproportionately from directing R&D spending toward capital equipment and infrastructure rather than current expenditure, while this lever appears largely neutral for economies close to the frontier.
© CPAG — Consilium Policy Advisors Group. All rights reserved. This report may be quoted, reproduced or circulated for informational, educational, institutional or non-commercial purposes, provided CPAG is clearly credited as the source and the content is not altered or taken out of context. The material is intended to encourage better understanding of economic policy and financial markets and does not constitute an offer or solicitation to buy or sell any commodities, securities or investments. This document includes forward-looking statements based on current views and assumptions and subject to risks and uncertainties; actual results may differ significantly. Any person using this report's material does so solely at their own risk.