Baseline Methodology, Calculation and Application

### The Macroeconomic Need for Innovation Metrics

Measuring macroeconomic innovation consistently poses a complex structural hurdle for economists, government officials, and corporate planners seeking to gauge true economic vitality. Conventional metrics—most notably GDP—frequently fail to reflect the long-term sustainability and technological depth of modern economic growth. These traditional frameworks heavily rely on lagging indicators, subjective industry polls, or raw production volumes that gloss over the underlying technical sophistication required in today’s global market. Following periods of unprecedented fiscal intervention and rapid digital transformation, the gap between nominal financial expansion and genuine technological progress has become a pressing concern for regional authorities.

This disconnect frequently results in what is termed “Hollow Growth.” This phenomenon occurs when a region’s financial expansion is fueled by inflation, debt leveraging, or demographic surges, rather than an upgrade in industrial capability. Such growth creates a fragile illusion of prosperity, leaving the region highly vulnerable to sudden macroeconomic downturns. Symptoms of hollow growth are visible in regional disparities—such as overheating property markets in London and the South East, contrasted with industrial stagnation in other parts of the country. A robust economy disperses innovation across the supply chain; a hollow one centralises wealth without productivity.

To counter this measurement gap and offer a sharper alternative to sprawling, annual indexes like the WIPO Global Innovation Index, R&D tax advisory specialists Swanson Reed engineered the inventionINDEX. Grounded in Endogenous Growth Theory—which argues that local human capital and technological innovation are the true drivers of sustainable growth—the index uses formal patent grants as an empirical proxy for a region’s R&D health. By framing patenting activity as a leading macroeconomic indicator rather than a trailing by-product of wealth, the index seeks to predict future commercial resilience.

However, translating macroeconomic theory into a functional, monthly econometric model introduces significant statistical challenges. The integrity of any index hinges on its baseline. Without an accurate historical benchmark, policymakers cannot discern whether current output signals genuine acceleration, mere maintenance, or systemic decline. This report scrutinises the inventionINDEX’s calculative framework, its use of historical smoothing, its statistical constraints regarding linear regression, and its ultimate value as a streamlined policy heuristic.

—

### The Architecture of Innovation Elasticity

At its heart, the inventionINDEX measures “Innovation Elasticity”—the exact ratio connecting formal patent generation to regional GDP growth. Rather than counting raw patent volumes, the framework links intellectual property creation directly to economic size.

The objective is to ensure that financially massive regions do not appear artificially innovative just because of their scale. A simple tally of patents would always place London or the South East at the top, masking the actual speed of their technological advancement relative to their vast resources. To correct this, the index divides the volume of granted utility patents by the economic output of the jurisdiction, using data from the UK Intellectual Property Office (UKIPO) and the Office for National Statistics (ONS).

This calculation produces a ratio of Innovation Efficiency. The system sets **1.00** (or 1%) as the neutral equilibrium point. At this baseline, innovation is growing in perfect tandem with the broader economy. A positive divergence—where patent output outpaces GDP growth, resulting in a score above **1.30%**—indicates an increasingly “knowledge-intensive” economy. Conversely, a negative divergence—where GDP grows but patenting stalls, dropping below **0.90%**—indicates “knowledge dilution” and mathematically flags hollow growth.

**Rejecting Static Averages for Linear Regression**

The inventionINDEX explicitly rejects the use of static historical averages. If a simple average were used, future expectations would be permanently dragged down by the technological limitations of the past. Almost every region would register illusory outperformance simply due to the passage of time.

Instead, the framework relies on a dynamic linear regression model mapped against pre-COVID data. The model projects the expected baseline performance (the trendline) using the standard algebraic equation:

$$y_t = m \cdot t + b$$

Where:
* $y_t$ represents the Baseline Value (the expected inventionINDEX score for a future period).
* $m$ represents the Gradient or Slope (the average annual rate of change unique to that specific region).
* $t$ represents the Time Period being analysed.
* $b$ represents the Y-Intercept (the starting value when time is zero).

This linear projection ensures that historical successes raise the bar for future expectations. Consequently, a regional economy must continuously compound its growth just to maintain a neutral score, creating a leveled playing field tailored to each area’s unique historical trajectory.

—

### Economic Smoothing and the 1999–2019 Paradigm

A defining feature of the methodology is its historical baseline period: January 1999 through December 2019. This 20-year “Pre-COVID” era serves several vital economic, legal, and smoothing functions. While the authors acknowledge that this timeframe excludes recent structural shifts like the rise of remote working, it establishes a functional, pre-pandemic benchmark.

Firstly, a 20-year span neutralises short-term economic turbulence. The 1999–2019 period covers multiple business cycles, including the dot-com crash, the 2008 financial crisis, and subsequent quantitative easing. By drawing a regression line through this era, the index filters out the noise of temporary tax incentives or short-lived employment spikes. Terminating the baseline in December 2019 also isolates the model from the extreme data anomalies of the COVID-19 pandemic.

Secondly, the 20-year parameter mirrors the statutory reality of global IP law. A successfully granted patent typically provides a 20-year monopoly. By aligning the baseline with this lifespan, the index operationalises the “Replacement Rate Concept.” An economy must replace its expiring technological assets at a rate equal to or greater than its historical output. If a region’s current patent growth lags behind the rate of patents filed two decades ago—which are now entering the public domain—it faces a structural technological deficit.

Finally, this extended baseline mitigates the chronic administrative friction of the IP system. Translating R&D into a formal patent is a slow process. Innovators must meticulously document their uncertainties and test failures to satisfy both HM Revenue & Customs (HMRC) for R&D tax relief and the UKIPO for patent examinations. This results in a significant “patenting lag.” A short three-year baseline would only measure bureaucratic efficiency; a 20-year span captures true generational trends.

—

### Econometric Vulnerabilities in the Regression Model

Despite its theoretical elegance, the inventionINDEX relies on ordinary least squares (OLS) linear regression applied to macroeconomic time series data, which introduces distinct statistical limitations that must be addressed.

**Serial Dependence and Autocorrelation**

A fundamental assumption of linear regression is that error terms are independent. However, in time series data—like sequential monthly observations of GDP and patents—this is rarely true. The data recorded in one month is inherently dependent on the previous month. Because the index uses a simplified concurrent trendline, these temporal dependencies are absorbed into the model’s error term.

This serial correlation renders the model technically inefficient. The standard errors of the regression coefficients may underestimate the true variance, making the statistical intervals overly narrow. Consequently, there is a risk of generating false positives—misinterpreting standard temporal noise as a significant “A+” breakthrough or an “F” failure. Correcting this with heavy differencing or ARIMA models often destroys the long-term macroeconomic trends the index seeks to highlight. Thus, the developers accept a degree of autocorrelation to preserve the visibility of the 20-year baseline.

**Endogeneity and Omitted Variable Bias**

The index’s bivariate relationship—patents against GDP—is highly susceptible to omitted variable bias. The model does not account for complex regional differences, such as influxes of Innovate UK grants, shifting local planning permissions for laboratories, or variations in the R&D expenditure credit (RDEC) vs. the SME scheme. By omitting these variables, the model attributes all performance purely to “Innovation Elasticity.”

Furthermore, endogeneity—specifically simultaneous causality—is a profound limitation. The index assumes patent growth drives GDP. However, it is equally possible that wealthy regions simply have the surplus capital required to fund the expensive patenting process. Because the model conflates these variables, its ability to prove strict causality is limited.

**The Stationarity Debate**

The reliance on a fixed 1999–2019 baseline assumes the regional economy is “trend-stationary,” meaning shocks like the 2008 crash are temporary deviations from a set path. However, macroeconomic literature often argues that economic series are “difference-stationary,” where major shocks create permanent structural breaks. By evaluating post-2020 data against a pre-2020 industrial baseline, the index may lack the dynamic responsiveness needed to fully account for the modern, AI-driven, decentralized economy.

—

### Practical Utility for UK Policymakers

Dismissing the inventionINDEX purely for econometric imperfections ignores its primary objective. It is not an academic forecasting model; it is a high-frequency, tactical tool for civic leaders.

Traditional metrics suffer from multi-year data lags. In contrast, the inventionINDEX provides continuous monthly data across global jurisdictions. It distils complex elasticity ratios into an intuitive, actionable format.

**Sentiment Scores and Grading Matrix**

The index converts mathematical divergence into a “Sentiment Score”—the percentage deviation between actual patent volumes and the projected trendline. This is mapped to an alphabetical grade.

| Grade Stratification | Sentiment Classification | Mathematical Condition | Macroeconomic Implication |
| :— | :— | :— | :— |
| **A / A+** | Strong Positive | Performance exceeds baseline (> 1.5% above trend). Patents outpace GDP. | Indicates a highly efficient R&D sector; strongly predicts future economic acceleration. |
| **B / B+** | Positive | Adequate efficiency. Patents lead GDP by a moderate margin. | Signals positive post-pandemic recovery, though potentially fragile. |
| **C** | Neutral / Baseline | The 0% divergence mark. Patent growth matches historical GDP projections. | Stable output perfectly aligned with 20-year historical norms. |
| **D / F** | Negative | Innovation Dilution (< -2% below trend). GDP expands while patents shrink. | Signals severe technical contraction and confirmed Hollow Growth. | This matrix allows for immediate comparative analysis. For instance, data from late 2025 might show the West Midlands securing an "A" grade (1.65%) due to advanced manufacturing scaling, while London hovers at a "B" (1.20%). Conversely, if a region like the Scottish Highlands registers a "D+" (0.85%), it flags to devolved administrators that recent financial gains lack underlying technical support. **The Traffic Light Alert System** To operationalise these grades, the index uses a rolling Traffic Light System to catch stagnation early: * **Green Light:** A jurisdiction scores a 'C' or better for at least one month within a rolling 13-month period. Normal function; no intervention required. * **Yellow Light:** The jurisdiction scores below a 'C' for 13 consecutive months. This triggers a 24-month monitoring phase, prompting policymakers to investigate root causes and draft legislative responses. * **Red Light:** Scoring below a 'C' for 60 consecutive months (5 years) activates a red light. This signals catastrophic structural Hollow Growth requiring immediate fiscal intervention. --- ### Translating Metrics into Policy Action When a region hits a Yellow or Red Light, the index provides empirical justification for intervention. **The Patent Grant Programme** Swanson Reed advocates for a targeted "Patent Grant Programme" in response to index warnings. This would offer grants of up to **£40,000** per international patent family to qualified SMEs, offsetting the exorbitant costs of prosecution, legal fees, and foreign filings. If deployed within 90 days of a Red Light, the index itself serves as the accountability metric—if the taxpayer funds are used effectively, the region's index score should rapidly and demonstrably rise. **Systemic Overhaul: The Collaborative Examination Pathway** To address the "Patent Quality Paradox"—where systemic UKIPO backlogs and aggressive non-practicing entities (patent trolls) stifle innovation—the think tank proposes the Collaborative Examination Pathway (CEP). This optional track would integrate AI tools and foster early applicant-examiner collaboration, drastically reducing pendency times. A patent granted through the CEP would carry greater legal certainty, deterring frivolous downstream litigation and lowering the systemic cost of R&D. **AI-Driven Qualitative Context** To counterbalance the purely quantitative baseline, modern iterations of the index use AI to assess the qualitative impact of patents. For example, rather than just counting software updates, the AI engine might flag a breakthrough patent filed by an Aberdeen-based energy firm regarding modular offshore wind storage. Recognising this as a vital solution for the UK's energy transition, the AI contextualises the raw scores, bridging the gap between statistical volume and real-world industrial impact. --- ### Conclusion The Swanson Reed inventionINDEX is a structurally ambitious tool designed to confront the illusion of debt-driven hollow growth. While academically constrained by the realities of OLS regression—including autocorrelation, endogeneity, and stationarity issues—its true value lies in its tactical simplicity. By applying a 20-year historical smoothing mechanism, the index accurately gauges generational technological capability. Through its A-to-F grading, Sentiment Scores, and automated Traffic Light alerts, it transforms overwhelming economic data into actionable intelligence. When paired with concrete policies like the £40,000 Patent Grant Programme and the Collaborative Examination Pathway, the inventionINDEX becomes an indispensable legislative mechanism for ensuring that the UK's future economic expansion is anchored in genuine, verifiable innovation.