Tracking the Economy through Firm Creation:Evidence from Real-Time Administrative Data

URL SCAN: Tracking the Economy through Firm Creation: Evidence from Real-Time Administrative Data
FIRST LINE: We introduce a novel real-time dataset, Companies House Real-Time (CHRT), that captures daily firm creation and dissolution activity for the full population of UK-registered companies.

THE DISSECTION

This is a data infrastructure paper. It does exactly what it claims: uses UK Companies House administrative data to build a real-time firm creation indicator that arrives months before official statistics. The SVAR analysis correctly identifies incorporation activity as a leading indicator of employment and output within the historical sample.

The paper is methodologically competent. It is also structurally irrelevant to the economy being diagnosed.

THE CORE FALLACY

The paper assumes the incorporation-to-employment-to-output pathway is structurally stable.

It documents that firm entry shocks produce persistent employment increases. This is fine as a backward-looking empirical regularity. Under the Discontinuity Thesis, the mechanism breaks in two ways simultaneously:

1. The numerator collapses. As AI achieves durable cost superiority across cognitive and physical tasks, the economic rationale for creating firms to employ humans evaporates. Incorporations may continue or even accelerate—not because they employ people, but because registering a firm costs nothing and AI can manage the legal/compliance overhead. The paper measures incorporation quantity, not incorporation purpose.

2. The denominator stabilizes at a floor that is not full employment. Even if some new firms hire humans, the employment signal from firm creation weakens because the marginal value of human labor to the firm converges toward zero. SVAR coefficients estimated on historical data reflect a world where new firms reliably needed human workers. That world ends when AI can replace the workers those firms would have hired.

The SVAR is not wrong about the past. It is structurally blind to the future.

HIDDEN ASSUMPTIONS

• Firms are created to employ humans. This is the operative assumption. Under AI-capable capital, firms are increasingly created to avoid employing humans—or to concentrate capital without requiring human labor at all.

• Business formation signals productive economic activity. Corporate registration activity has been growing in the UK for years while productivity growth stagnated. The paper uses incorporation counts as a proxy for economic health without interrogating whether incorporation quality (actual productive employment generated per firm) has collapsed.

• The leading indicator relationship is structural, not historical. The paper offers no theoretical mechanism for why incorporations lead employment—it simply documents the correlation in historical data. Correlations estimated during the post-WWII human labor era are not structural invariants.

• The unit of analysis (the firm) remains economically coherent. Under the DT framework, the firm as an employment-organizing institution is one of the entities under pressure. A world where AI capital operates without requiring formal corporate structures, or where human labor is organized through non-firm mechanisms (UBI-financed gig systems, cooperative structures, household production), renders firm creation data increasingly noisy as an economic signal.

SOCIAL FUNCTION

Status quo refinement masquerading as methodological progress. This paper does legitimate work: better data, faster release, cleaner identification. It serves economic monitoring infrastructure by making existing measurement more precise. This is useful within the paradigm of managing the current system.

Its social function is transition management: providing better tools for central banks and policymakers to monitor a system that will increasingly not work the way the tools assume it does. The paper contributes to the illusion that the post-WWII economic order can be monitored and managed into stability—without addressing the structural obsolescence underlying the monitoring problem itself.

THE VERDICT

Methodologically sound. Analytically blind. The CHRT dataset is a genuine improvement in real-time economic measurement. The SVAR results are empirically valid for the sample period. The paper does what good economics does.

It does not do what relevant economics must do: interrogate whether the leading indicator it discovered is measuring something that continues to exist. Under the DT framework, firm creation counts become increasingly decoupled from employment and output—not because the data is bad, but because the economy transitions away from the relationship the indicator is built to track.

Relevance score: High for current-cycle monitoring; low for diagnosing the terminal trajectory of the post-WWII order. The paper is a better thermometer for a fever while the patient is developing organ failure.