Jan Tinbergen on Decentralization

Jan Tinbergen was a Dutch economist who won the Nobel prize in economics. His brother, Nikolaas, won the Nobel Prize in physiology or medicine. Tinbergen was a mathematically oriented economist who wrote from a Keynesian perspective. In 1954 he wrote a monograph "Centralization and Decentralization in Economic Policy"* in which he outlines structural equations that might be useful in studying the effects of decentralization on social welfare. He defines "conformity" as the conformity of policy makers' assumptions to the real world. In the chapter on centralized economic policy he notes the importance of trial and error to policy making. He is talking about quantitative policy goals such as income and price levels (p. 16):

"Conformity will hardly ever exactly be realised in practice, if only because of our imperfect knowledge of the structural constants or because of changes in such 'constants'. This circumstance makes it necessary and more or less usual for policy-makers to follow a trial-and-error policy method. They will try a certain value of their instrument and observe the result. Dependent on this result they will adjust the instrument variable. In view of this practice it is sometimes doubted whether it is any use to make econometric estimates for the values of the instrument variables. The answer is that although it is possible by trial and error to find the correct values, this method nevertheless implies some waste...social utility is usually unfavourably influenced by the frequency of changes in political instrument values."

The problem of uncovering the correct policy course becomes even more severe in the case of qualitative policies:

"In the case of qualitative changes all we can do, as a rule, is just to calculate the outcome for the alternative cases and to compare them...the problem of comparing alternative forms of organisation of economic life constitutes the problem par excellence of economic science; its real raison d'etre."

Tinbergen notes that:

"If in particular a higher degree of centralization would have been possible and would have led to target variable values with a higher utility, the reason of the discrepancy may be said to be the very fact of decentralization. In the case of non-conformity it is by wrong insight into or knowledge of the objective equations that wrong instrument variables are chosen. Simple though this phenomenon and its remedy may seem, it is not so simple in practice. One reason is that exact knowledge on the mechanism of an economy does not exist and that there can, therefore, only be question of better or worse approximations. Another reason is that even experts in this field are not unanimous and that it will be very difficult, then, for non-experts to be. Even if there is unanimity among experts in some cases, it is not easy to convince the rank and file of large organizations of this expert opinion."

Although economies of scale make centralization more efficient (p. 59):

"(I)t is in the nature of centralization that it will be more costly as soon as purely technical economies of scale are out of the question. It generally requires a bigger administrative machinery and, before all, more interference with private freedom,"

and (p. 62):

"any application of an instrument in a centralized way will imply higher disutilities, or will lower utility more than will the same application in a decentralized situation would. There is, therefore, a general tendency in a centralized situation to make less use of all the instruments, which will be the stronger the higher the disutilities are."

However, in decentralized situations policies which help other units are used to a lesser degree than they would be in a centralized situation so if there are positive overflows, the decentralized approach may not be as good.

Where all the units have similar or identical perceptions of their own social welfare then centralized regimes are to be preferred. Moreover, where policy instruments conflict among the goals of subunits or where a policy would help all the subunits if applied across the board, then a centralized approach is better.
But instruments which do not affect other decentralized units are better applied at a decentralized level (p. 74):

"In plain language, the cost and trouble will be diminished, without changing the situation any further...Each decentralization means reduction of costs and disutility generally."

Thus, where each unit's social welfare is similar to the others, policies which are neutral or do not affect the other units are best pursued on a decentralized basis and policies that can have an across the board positive effect are best pursued on a centralized basis (such as general price levels).

It is difficult to know which policies have positive overflows. Many policies have negative effects. Economists of the 1940s believed that the Fed was an instrument for positive results. Today, even the Keynesian descendants of this group does not deny that Fed oversight of the banking system has resulted in endemic catastrophes and miscalculations leading to the need for a trillion dollar welfare infusion from the public.

Polices which have outright negative effects, which I would argue that the Fed and the American banking system has, are better implemented on a decentralized basis because the groupthink, ideological commitment to central banking and massive errors that policy makers commit can be limited to local units that have preferred to take such risks.

*J. Tinbergen, "Centralization and Decentralization in Economic Policy". Amsterdam: North-Holland Publishing Company, 1954.