🚗 Japan's new growth strategy makes a bold claim. Put ¥8.2 trillion (about $52 billion) into self-driving technology, it says, and the economy gets ¥187.3 trillion (about $1.19 trillion) in "economic ripple effects." Divide one by the other and you get roughly 23 to 1. The two numbers, though, do not measure the same thing.

Who did the division?

Both figures come from a public-private investment roadmap that Japan's Headquarters for Japan's Growth Strategy adopted on July 21, 2026, the same day the cabinet approved the broader Japan Growth Strategy. The roadmap picks 17 strategic fields and 62 key products and technologies, and puts cumulative public and private domestic investment across them at more than ¥370 trillion (about $2.35 trillion) through fiscal 2040.

For self-driving, it assumes ¥8.2 trillion of investment through FY2040: production lines for autonomous-ready vehicles, development of end-to-end AI (a single model that handles everything from perception to route decisions), and telecom networks. It puts the economic ripple effect over the same period at ¥187.3 trillion.

The ¥8.2 trillion is not a government budget. It combines public and private money and, according to the document, was built largely from ministries' interviews with major companies about their planned spending. Nor does the roadmap state any ratio. The "roughly 23 to 1" (22.8, to be exact) is what readers get when they divide, and the division compares two different kinds of numbers.

For comparison, Japan's auto-related industries ship about ¥72 trillion (about $458 billion) worth of goods in a single year, according to a 2023 figure the roadmap cites. Set against that, a total accumulated through FY2040 of ¥187.3 trillion is not wildly out of proportion.

The investment alone cannot produce 23x

According to a footnote, the calculation uses input-output analysis. This standard tool uses tables showing which industries buy from which. Feed in new demand, and it estimates how much extra production spreads to parts, materials and raw-material suppliers (the first-round effect). It then adds a second-round effect, in which wages from that extra production are spent and trigger still more production.

The key question is what went in at the top. The footnote says the inputs were the investment and the "production increase resulting from that investment." So the model received not only ¥8.2 trillion of spending but also the additional output Japan expects if the strategy succeeds. For self-driving, the stated goal is about 25% of global autonomous-vehicle sales in the 2030s, matching Japanese automakers' current share of global car sales.

Reading the footnote, the natural interpretation is that most of the 23-to-1 comes from those assumed sales, not from the investment itself. Yet the roadmap does not publish the production-increase figure for self-driving or for any other item. The number needed to check the result is missing.

There is a second catch: ¥187.3 trillion is not GDP. A guide to the same method from Ishikawa Prefecture's statistics office explains that what is usually called the economic ripple effect is total induced production, while the part corresponding to GDP is the smaller value-added figure. Gross production counts the same value each time it changes hands: steel becomes a part, the part becomes a car. The roadmap itself warns that effects overlap across items and cannot be added together.

That distinction cuts both ways in the debate. In a September 22, 2026 column for Toyo Keizai, Richard Katz, a New York-based contributing writer, described the claim as ¥8 trillion of investment producing 23 times as much GDP over the next 15 years, and argued that the plan's promised returns are exaggerated. He asks whether Japanese automakers, given the country's software weakness, can take on proven players such as Waymo, which is owned by Alphabet, and cites an IMD ranking that puts Japanese firms 56th of 58 countries on using big data. Doubting the 25% share assumption is reasonable. But by the government's own definition, ¥187.3 trillion is gross production, and the GDP-like portion is smaller.

Ratios also vary widely across items. Dual-use technology, meaning products with both civilian and military uses, comes out at about 27 to 1. Flying cars, which the government places in the same demonstration stage as self-driving, come out at about 5 to 1. Mineral refining ends up below its own investment; its investment figure includes overseas mines and smelters, which the footnote says are not counted in the effect. Change the assumptions, and the answer moves that far.

Big-looking numbers are not a Japanese invention

Results that swing with assumptions are common everywhere. When the United States passed the Inflation Reduction Act in 2022, its climate and energy law, the congressional Joint Committee on Taxation first estimated the energy tax credits at about $271 billion (2023 to 2031). Researchers at the Brookings Institution later put the cost anywhere between $244 billion and $1.1 trillion, depending on what they assumed about which projects qualify and about supply bottlenecks. That is a cost estimate rather than a benefit estimate, but the pattern is the same: the assumptions drive the headline.

Britain has turned the handling of such numbers into rules. The Treasury's appraisal guide, the Green Book 2026, says Keynesian multiplier effects, where new income is spent and generates more output, show up after almost any public spending, so they should not be booked as a social benefit when weighing one option against another. It also requires appraisers to correct for optimism bias, the habit of underestimating costs and overestimating benefits. Japan's figure is a policy indicator, not the kind of project appraisal the British guide covers. Still, the UK leaves out, in that kind of comparison, the very second-round effect Japan adds in.

The estimate still has a job to do

Even so, without some sense of scale there is no starting point for deciding where to bet public and private money. The strategy itself says the roadmap will go through a PDCA review cycle so it does not remain a plan on paper, and that budgets doing little to attract investment will be cut.

The harder question is whether anyone can check the result later. For the Recovery and Resilience Facility, the core of the EU's NextGenerationEU recovery fund, the European Court of Auditors reported in October 2023 that the monitoring system tracks progress but cannot capture overall performance.

In the documents reviewed for this article, there is no description of how anyone will measure, in FY2040, whether the ¥187.3 trillion actually materialized. If the production-increase figure and the share assumptions were published, that check could be run by people outside the government too.

Money is one question, rules are another

Even if the money flows exactly as projected, putting driverless cars on public roads requires clearing a separate hurdle: permits. We looked at what Tokyo's robotaxi plans are up against in a separate article, and at the ¥370 trillion framework as a whole in this piece.

How does your government estimate the payoff from the money it puts into industry? And does anyone go back later to see whether the numbers came true?

References