Who’d be an economic forecaster? The International Monetary Fund this week revised up Britain’s GDP growth in 2022 to 3.6 per cent and revised down 2023 growth to 0.3 per cent, but admitted its projections were already out of date. Given they were estimated before Kwasi Kwarteng’s tax-cutting mini-budget, the fund said growth was likely to be “somewhat” higher than suggested.

Forecasting a £2.2 trillion-plus, 69-million-person economy is perilous at the best of times. Add in today’s Ukraine war shocks, new tax policies, the Halloween fiscal statement, monetary tightening and the possibility of supply-side reform, and the IMF’s estimate was doomed to failure. “It’s tough to make predictions,” Yogi Berra supposedly said, “especially about the future.”

Unfortunately, “predicting the present” is hard enough. With GDP figures only published months after economic activity has occurred, policymakers are typically left searching for disparate clues as to our current economic health.

The IMF uses monthly purchasing managers’ indices, as well as consumer and business sentiment surveys, to infer that the UK’s economy has slowed down. When conditions are so volatile, though, these indicators are notoriously unreliable. Serving up duff information to ministers risks badly timed interventions and misguided economic narratives.

Thankfully, Google may be riding to the rescue. The Organisation for Economic Co-operation and Development’s Weekly Tracker of Economic Activity series, introduced in the pandemic, harnesses Google Trends data to provide weekly GDP estimates relative to a year earlier. With results published within a week, this is as close to a real-time GDP estimate as one could hope.

Google’s search data incorporates consumption (through searches for vehicles and appliances), housing (properties and mortgages), jobs (jobseeker’s allowance and vacancies), industrial activity (maritime transport or agricultural equipment), business (bankruptcy and venture capital), poverty (food banks), economic sentiment (searches for recession), and more.

The OECD collates this and applies machine learning techniques to construct a “timely picture of the economy”. It models how searches interact with historical GDP, and then uses that to estimate weekly GDP based on the past seven days’ search trends.

The results are pretty consistent with the eventual GDP outturn data (see chart). The OECD series perfectly captured the depth of the pandemic recession and the magnitude of the recovery, although recent revisions suggest it overestimated economic strength through late 2020. It has, though, has correctly identified nearly every major direction change in GDP growth since January 2020 and seems to be getting more accurate.

How does it “nowcast” our current plight? The last results for the week beginning September 25 suggest activity was 1.8 per cent higher than last year. This implies that the big picture stagnation seen in yesterday’s Office for National Statistics GDP data from January to August probably continued through September. Growth has stalled.

Intriguingly, this data is from the week after the mini-budget, yet one sees no major disruption from that. Despite gilt yields surging and the pound’s volatility, the Google estimate of activity relative to last year was stubbornly unchanged from the week before.

Given the unprecedented financial reaction to Kwarteng’s statement, perhaps the model just didn’t capture the right searches to measure the economic shock. Or maybe these effects will filter through shortly. There’s no evidence from Google search trends yet, though, of either a mini-budget stimulus or of Kwarteng “crashing the economy”.