-
After the « dead capital » of pensions and the « diverted » savings, the solution seems obvious: build a stock. This is where we must slow down.
-
For France already capitalises without knowing it — the ERAFP, employee share ownership, the PER — and its retirement savings place only 3.3% in the unlisted.
-
From the Netherlands to Canada, pension funds invest where the money pays, not at home. Reviving capital requires four conditions that no French report brings together.
At the end of December 2025, the French authorities authorised the sale of LMB Aerospace to the American group Loar Group. LMB Aerospace is an engine-maker based in the Corrèze. It supplies equipment for the Rafale, for the nuclear ballistic-missile submarines and for the aircraft carrier Charles de Gaulle.
The operation was authorised because there was no French buyer. It was this sale that prompted three members of parliament — Christophe Plassard, Charles Rodwell and Jean-Louis Thiériot — to submit, on 21 July 2026, a report on the economic security of France. « When we talk about economic security, we tend to talk only about foreign investment in France, » observes the first. « The weak link is small businesses. »
Their report proposes three things: that the State Shareholding Agency retain its dividends in order to direct them towards strategic sectors; that pension funds be created and steered towards companies deemed strategic; and that a dose of funded pensions come to feed them durably. National consultations on economic security are to be held in September 2026.
The conclusion of the two previous articles thus seems to find its answer: since capital is lacking, let us build it. This is where we must slow down.
France already capitalises, and does not know it
Before imagining a new scheme, one must look at those that already exist. They are more numerous than one thinks.
The ERAFP is a compulsory funded scheme. Created by the 2003 reform, operational since 2005, it levies 10% — half from the employee, half from the employer — on civil servants’ bonuses alone, themselves capped at 20% of index-linked salary. A tiny base, a modest rate. Twenty years later: €47.8 billion and 4.4 million contributors.
The pharmacists’ fund has been compulsorily funded for far longer. Its supplementary funded scheme holds €7.7 billion in assets, of which about €800 million is invested in French small and medium-sized enterprises.
Employee share ownership weighs €80.4 billion, out of €229.4 billion of employee savings and corporate retirement savings. It is, by far, the item most massively invested in the capital of French companies — because it does not leave a choice of geography.
The retirement savings plan (the PER), finally, reached €150.4 billion and 12.9 million holders at the end of 2025, seven years after the Pacte law.
The Pensions Advisory Council (Conseil d’orientation des retraites), for its part, explicitly excludes the ERAFP from the system it models. The note appears beneath each of its graphs: « excluding RAFP ». The main French compulsory funded scheme lies outside the scope of the official diagnosis on pensions — which says a great deal about the way the debate is framed.
First warning: the stock does not go where you think
Here is the decisive objection, and the demonstration of it is French.
French retirement savings plans hold about €5 billion of unlisted assets — 3.3% of their holdings. And this despite a regulatory obligation on the matter.
Read that sentence again. France already has €150 billion of retirement capital. A rule requires that part of it be placed in the unlisted. The result is 3.3%. Building the stock is therefore not enough to decide its use.
The rest of the world confirms this, and often more brutally.
The Netherlands has the largest funded system in Europe: €1,624 billion. Its leading fund, ABP, manages €533 billion for Dutch civil servants and invests about €20 billion of it in the Netherlands — less than 4%. At the end of 2024, Dutch funds held €293 billion of American companies against €97 billion of European Union companies. The year 2025 marked a turn — €30 billion of American securities sold, €23 billion of European securities bought — but the order of magnitude remains the same.
Canada is the model most readily cited in France. Its public pension fund placed 74% of its assets in Canada in 2005. It places 12% today, against 47% in the United States. The eight big Canadian funds manage 2,500 billion Canadian dollars; only three hold more assets at home than in the United States. The Caisse de dépôt du Québec, for its part, holds 25 billion Canadian dollars in France: Alstom, Keolis, Eurostar, Gecina. The model we admire invests in our country what it no longer invests in its own.
The United Kingdom is the limiting case. British insurers and pension funds held 52.1% of British listed shares in 1991. They hold 4.2% of them in 2022. Pension funds alone went from 32.4% in 1992 to 1.6%. The rest of the world today holds 58.8% of the British market, a historic record. The British Treasury itself acknowledges it: the fall exceeds what the mere decline of the country’s weight in stock indices would explain.
Across the seven largest pension markets in the world, the share of domestic equities in equity portfolios fell from 57.1% in 2004 to 34.1% in 2024.
What those who have not dispossessed themselves do
The picture is not uniform, and that is what makes the subject interesting.
Denmark holds one of the highest ratios of pension assets in the world — of the order of 200% of its gross domestic product — and invests a third of it at home. Sweden places 43% of its pension-fund assets on its own market, and more than half in equities. These are the two countries whose equity markets are the deepest in Europe relative to their size.
Above all, two countries have just decided that the domestic bias would not be left to chance.
The United Kingdom is legislating. The Mansion House accord, signed on 13 May 2025 by seventeen managers representing £252 billion, commits to placing at least 10% of default funds in private assets by 2030, of which at least 5% in British private assets. Then the Pension Schemes Act, enacted on 29 April 2026, creates a reserve power allowing the government to set investment targets by asset class — capped at 10% of assets and 5% in British assets, usable only once, with a programmed sunset in 2032.
Canada too. At the end of 2024, Ottawa lifted a regulatory constraint to facilitate domestic equity stakes, and made up to 45 billion Canadian dollars of public financing for data centres conditional on a contribution from Canadian pension funds in a two-to-one ratio.
« The domestic bias does not survive on its own. It is decreed, or it disappears. »
Second warning: the transition has to be paid for
The most solid objection is not that one. It was formulated by Nicholas Barr and Peter Diamond, and it is disarmingly simple: « if generation C contributes to its own accounts, the pension of generation B must be paid from another source. »
The gift received by the first generation of the pay-as-you-go system — the one that drew pensions without having contributed all its life — must be repaid by someone. There are only three possible sources of financing: debt, tax, or the repudiation of rights. No one has ever found the fourth.
The precedents are unambiguous.
Chile paid for its 1981 transition with budget surpluses accumulated before the reform, under an authoritarian regime — a cost of about 4% of GDP a year between 1985 and 2000.
Sweden, so often cited, did not pay for its transition: it already had the capital. The scheme created in 1960 had accumulated reserves equivalent to about 40% of GDP, five years of benefits. France has €91 billion of supplementary-pension reserves and €20.7 billion in the reserve fund. The Swedish analogy therefore only half holds.
Poland financed its own through debt: 14.4% of GDP in transfers over thirteen years, plus nearly seven points of additional debt-servicing. It reversed course in 2013, repatriating 120 billion zlotys to the public system.
Hungary settled the matter even more simply: in 2010 and 2011, it transferred to the state 2,950 billion forints of private pension savings. Out of a little over three million employees, 2.9 million — 96.8% — returned to the public scheme.
One lesson stands out, and it holds even when one argues for building the stock: a constituted pension capital can be seized. Hungary confiscated it, Poland repatriated it, Chile let a quarter of it be withdrawn in three withdrawal windows between 2020 and 2021 — $45.8 billion taken out, nearly 2.9 million accounts brought back to zero. A balance that can be read is a balance that can be spent.
Only one transition produced neither crisis nor reversal: Australia. It dismantled nothing. In 1992 it added a new compulsory contribution of 3%, raised progressively to 12% in 2025, alongside a public pillar that was maintained. Result: about 4,500 billion Australian dollars, nearly 160% of GDP. And a net effect measured by the Reserve Bank of Australia: about 38 cents of every dollar contributed is offset by a fall in other forms of saving — 62 cents constitute additional net saving. The gain is real; it is not one hundred per cent.
De Soto, a second time
We opened this dossier with Hernando de Soto and his dead capital. He must be summoned one last time, but for what has been objected to him.
For his prescription failed. In Peru, the COFOPRI programme distributed more than 1.2 million urban land titles between 1996 and 2002. Erica Field and Maximo Torero measured its effect: a rise of nine to ten points in the loan-approval rate at the public bank alone, for the purchase of building materials — and no effect on private lenders. A third of titled households remained entirely excluded from formal credit. De Soto’s own institute acknowledged it.
Three explanations have been advanced, and each transposes to our subject.
Christopher Woodruff observes that without a workable foreclosure procedure, a title is worth nothing as collateral. Without stable and enforceable rules, a stock of capital does not become financing.
Philippe Lavigne Delville, the leading French-language authority on the subject, points to a « staggering » absence in De Soto: that of the upkeep of the registers. Without updating of transfers, land registries end up delivering false information. Now a pension system is exactly that: a register of claims that must be kept. And ours does not even show its balance to those who feed it.
Timothy Besley and his co-authors show, finally, that the effect depends on competition between lenders: where it is weak, the lender captures the surplus and welfare can decline. If there are no companies to finance and no market to do it, the capital will go elsewhere.
De Soto was right on the diagnosis and wrong on the remedy. The title is not enough to bring the asset to life. The stock will not be enough to finance the country.
Four conditions
From all of the foregoing, four requirements emerge, which no French report formulates together.
A written domestic bias. Without an explicit and monitored obligation, a French fund will reproduce the Dutch case — the 3.3% of the retirement savings plans already prove it here. Denmark, Sweden, the United Kingdom and Canada have all ended up writing it down somewhere.
Governance beyond the reach of the budget. The Pension Reserve Fund did not lack returns: it has yielded 4.4% a year since 2010. It lacked protection. Created to reach €150 billion in 2020, it received €31 billion, its top-ups ceased in 2010, and since then it pays €2.1 billion a year to the social-debt amortisation fund. It has €20.7 billion left. A fund that can be tapped is not a fund: it is a budget reserve.
Contained fees. One point of annual fees eats about 20% out of a career’s accumulation — the calculation is Barr and Diamond’s. The gap between centralised and decentralised management runs to tens of billions over a generation. The Swedish public default fund charges 0.07% a year.
A credible lock-up period. See Hungary, see Chile, see our own reserve fund.
What this will not solve
Two objections deserve to be heard, and they come from serious people.
The first is Anne Lavigne’s, before the Pensions Advisory Council in December 2025. Her title was a question: « Financing pensions through capitalisation: why not, but with what capital? » Her argument turns the thesis on its head, from the same premises: French net saving being insufficient, building pension funds without new saving would amount to having French capital financed by foreign investors. In other words, poorly designed, a sovereignty pension fund would aggravate precisely the ill it claims to cure. She recalls, moreover, that return cannot be decreed: in 2024, pension funds yielded 4.29% in real terms in Canada, but 0.79% in the United States.
The second is an objection of pure logic, formulated by the Pensions Advisory Council itself: « the various advantages assigned to capitalisation cannot be expected simultaneously. » One cannot promise at the same time a high return, secure pensions and the patient financing of the national economy. One must choose which of the three one wants.
One must, finally, recall why France abandoned funded pensions. It was not an ideological decision. In 1941, the country restored pay-as-you-go because the funded system established in 1930 had been wiped out, and the accumulated contributions were assigned to paying current pensions. What has changed since is not the nature of the risk but its management: a 1930 portfolio was national, in bonds, exposed to domestic inflation; a 2026 portfolio is international, diversified, index-linked. The risk has not disappeared. It is no longer the same.
As for the orders of magnitude, they invite modesty. The most complete simulation available comes from the French Asset Management Association (Association française de la gestion financière) — the federation of management companies, that is, the first beneficiary of the scheme it proposes. It puts the result of a transfer of one to two and a half points of contribution, at unchanged overall levies and for employees under forty-one only, at about €200 billion after twenty years, and one additional point of the replacement rate. Even the most interested body promises no miracle. A French funded scheme would be an instrument of capital ownership in the medium term. It would not be a solution to the financing of pensions, and to confuse the two is the surest way to lose both debates.
Let us return to the pay slip with which we began.
A Swedish employee receives each year a statement of account — a balance, in kronor — on a scheme that, like ours, works on a pay-as-you-go basis. An Australian employee reads on his pay slip the amount paid in and the name of the fund that receives it. A French employee sees €783 a month leave under a risk heading, and waits fifty-five years to read an estimate devoid of legal value.
The line missing from our pay slips is not a matter of page layout. It is the title deed that De Soto sought in the shantytowns of Lima. It would not be enough — this dossier will have tried to show as much. But one must begin there, for one does not revive a capital whose amount no one knows.
Is economic sovereignty decreed, or bought? Perhaps neither. It is, first of all, counted.
Sources: report « La sécurité économique de la France », 21 July 2026; ERAFP; CAVP; AFG; France Assureurs; Conseil d’orientation des retraites (« De la répartition à la capitalisation », 18 December 2025; hearing of Anne Lavigne, 11 December 2025); Fonds de réserve pour les retraites, 2025 annual report; De Nederlandsche Bank; ABP; CPP Investments; La Caisse; Office for National Statistics; DWP and HM Treasury; Pension Schemes Act 2026; Reserve Bank of Australia; ASFA; Superintendencia de Pensiones; International Labour Organization (ESS Working Paper no. 68); Barr and Diamond, « The Economics of Pensions », Oxford Review of Economic Policy, 2006; Field and Torero on the COFOPRI programme; Woodruff, Journal of Economic Literature, 2001; Lavigne Delville, L’Économie politique, no. 28, 2005; Besley, Burchardi and Ghatak, Quarterly Journal of Economics, 2012.
