The 300-Year Business: Government budgets for long-term results. A summary by the author
1. Three to six years against three hundred
I come from finance, and I begin where finance begins: how long does the business last? Politicians run a business of three to six years, the length of a term. The society they govern runs one of three hundred years, which is ten generations of thirty. This is a business fact, not an accusation. The rule of my book fits in one sentence: a politician is worth as much as the most distant generation they take into account when they decide. Three hundred years is the frame for weighing a decision, not a payment date. The payment I propose falls in twenty years.
2. Does planning pay?
I think it does, and I give the answer before the evidence. I compared seventeen countries over 62 years, 1960 to 2022, on four measures: income per capita, years of schooling, years of life and extreme poverty.
- The poorest of the countries I classify as planners stands at 22,741 dollars of income per capita. The richest of those that did not plan stands at 18,292 (author's calculation from the Maddison Project Database 2023, 2011 international dollars, retrieved September 2026).
- The planners grew between 0.6 and 3.4 percentage points a year faster than their comparable countries (same source). One point a year, held for 62 years, multiplies income by 1.85 (author's calculation from the Maddison Project Database 2023).
- Among countries that were already rich, planning bought something else: not falling.
The classification is my reading, not a formula, and seventeen countries chosen by an author illustrate a thesis without proving a cause. The strongest objection is that the fastest planners are those that cannot be voted out. I concede its force. I also show its costs: nobody can stop a mistake, and a rule lasts only as long as its leader. My answer is a rule that survives elections.
3. How long does it take?
Longer than a term. Across 777 dated decisions, a good decision took a median of fourteen years to show its full effect, and about nineteen on average. A bad one took a median of nine and about fourteen on average (author's calculation from the dated decisions of Chapter 3, version of October 2026). About one bad decision in four showed its damage in under five years, against about one good decision in ten. Voters can punish the fast error. Nobody collects for the slow good decision, and nobody pays for the slow bad one. The gap is plain:
years the effect takes minus years the office lasts = years the benefit is still on its way after the term has ended.
Twenty years covers just under two thirds of the good decisions and about four in five of the bad ones (same source). That is why I propose twenty.
4. How much does it pay, and how often does it fail?
I do not know the exact rate of return, and I say so. The growth gaps above, when applied to a laggard, imply it would have had between 1.4 and 7.5 times its actual income had it grown like its neighbour (author's calculation from the Maddison Project Database 2023, 2011 international dollars, retrieved September 2026). These are models, not losses anyone measured.
Planning also fails, and often. All seventeen countries appear on both the list of good decisions and the list of costly ones. Nine of them made the same error of opening credit faster than supervision. What separated them was the years taken to admit it:
cost of an error = size of the error × years it takes to be recognised.
Some corrected within one to three years, and one took seven. The failure rate itself is missing from the book, and I mark it as missing.
5. Compared with what?
Planning competes with handing out the same money today. For the politician, handing out wins. In an election year the fiscal balance worsens by about 1.3 percentage points of output in the average developing country, and by 0.1 in the average developed one (Shi and Svensson, 2006; 85 countries, 1975 to 1995).
For the country, investing usually wins. In my water illustration, a new plant costs 280 units over twenty-five years and renewing the pipes costs 180 (author's illustration with composite figures). Only the first 100 of the plant's cost falls on the government that chose it.
Economists will object that the future is discounted. At 3 per cent, a euro received in a hundred years is worth 5.2 cents today (author's calculation, September 2026, with the HM Treasury Green Book formula). I do not ask anyone to value the future more. I ask that politicians be paid when it arrives.
6. The proposed contract
Every 100 euros of a politician's salary is split three ways: 30 for presence, 30 for the year's work and 40 deferred. The deferred 40 goes into a trust outside the budget, so a later majority would have to fetch money it no longer holds.
- Three quarters is released when what was decided enters into operation, and the rest in twenty yearly instalments. Inaugurations and announcements do not count.
- At a real rate of 4 per cent, 120 deferred units on this calendar are worth 98 today, against 55 if paid in one sum at year twenty (author's calculation, September 2026, illustrative three-year term).
- The deferred balance is also a surety. A final conviction for corruption costs the whole of it.
- Under the rule of the fox, the party that proves a rival's corruption collects a prize of ten, fifteen or twenty times the amount proven, paid out of the corrupt party's public money.
- Part of each party's public money follows results, rising from 2 to 40 per cent over twenty years, with the pool not growing. In my illustration a small party that delivers goes from 30 to 78 units, 2.6 times as much (author's calculation, October 2026, illustrative figures).
None of the pieces is new. Deferred pay exists in banking rules, and results-based pay exists in the public sector. The novelty is joining them and applying them to those who decide the budget and to their parties.
7. What can go wrong
It goes wrong in one way: the person who decides does not live with the consequences. I describe three patterns, in public works, purchases through intermediaries and subsidies without an audited register. Others have measured part of them. A study of 258 transport projects found an average cost overrun of 28 per cent (Flyvbjerg, Holm and Buhl, 2002). Bribes in foreign-bribery cases averaged 10.9 per cent of the contract (OECD, 2014). My own estimates, that 38 to 40 euros in every 100 are lost, are composites and orders of magnitude, not measurements (author's estimate, September 2026).
The contract has limits, and I state them. It does not compete with illegal income. It needs a minimum of rule of law. It favours those who can wait twenty years. People will also try to manipulate whatever is measured. It begins with a pilot of three municipalities over three years, with a stopping rule fixed before the first payment. I have also written down what would prove me wrong.
8. The invitation
Four of my seven questions are answered by the evidence and three only in part, and I have left those gaps in plain view. Whoever completes one of them, whether how much planning pays, how often it fails or what it should be compared against, has a dinner with me and credit in the next edition. Nobody needs to agree with the book to accept. Being wrong is allowed; not planning is forbidden. "You are mad, but I wish it were so" is the answer I hope for. I would rather be checked than believed.
The full book is coming to Kindle.