Path Dependence and the Enduring Efficacy of Customary Land Tenure in Africa — Epoche C1
The policy at issue Since the Swynnerton Plan of 1954 proposed converting African smallholdings in Kenya into individually registered freehold parcels, the replacement of customary land tenure by statutory title has been a recurring prescription for African agriculture. Two terms need fixing before anything else. Customary tenure means the systems, differing greatly between places, under which rights to land are allocated and enforced by lineage heads, chiefs or village councils, are recorded in memory and public witness rather than in documents, and typically comprise several overlapping entitlements to the same ground — one household's right to cultivate, another's to graze the stubble, a third's to harvest particular trees. Freehold title means a right vested in a named individual, recorded in a state register, transferable by written deed, and enforced by courts. The strongest statement of the case for conversion is Hernando de Soto's The Mystery of Capital (2000), which argues that assets held without documentary title are dead capital: they cannot be pledged to a bank, so the wealth they represent cannot be mobilised. The argument this note makes is not that de Soto is wrong about what titles do, but that the comparison usually made is not a comparison of costs. It assumes that formal institutions minimise transaction costs and informal ones do not, and it draws support from Douglass North's account of path-dependent institutional change. Both moves need scrutiny, and one of them — the second — will turn out to be an outright misuse of the theory it invokes, which must be corrected before the defence of customary tenure can be stated properly. Transaction costs, and the arithmetic of a registry A transaction cost , in the sense Ronald Coase gave the term and developed in 'The Problem of Social Cost' (1960), is any cost of arranging and enforcing an exchange that is not the price of the thing exchanged: finding out who holds the right, establishing that the seller really holds it, agreeing terms, and making the agreement stick. Coase's point was that where these costs are zero the initial allocation of rights does not affect what ends up being done with a resource, and where they are positive it does. Land is the case where they are highest, because the object cannot be moved, inspected or handed over, so every claim about it must be established by evidence external to the thing itself. Whether a registry lowers these costs is an arithmetical question rather than a doctrinal one, and it can be posed exactly. Let $F$ be the cost of bringing one parcel into a functioning register and keeping it there — survey, adjudication of competing claims, the registry's own overheads apportioned per parcel. Let $c_f$ be the cost of executing one transfer through the formal system, and $c_k$ the cost of executing one transfer through the customary system. Let $n$ be the number of transactions the parcel undergoes over the period in question. Formal registration is the cheaper arrangement precisely when $$n\,(c_k - c_f) > F, \qquad \text{that is, when} \qquad n > \frac{F}{\,c_k - c_f\,}.$$ Every term in that threshold works against formalisation in the smallholder case, and it is worth saying why in each instance. The numerator $F$ is large, and its largest component is not the survey but the adjudication: because a customary parcel typically carries several overlapping entitlements, issuing a single title requires deciding which of them survive and extinguishing or ignoring the rest, which is contentious, slow and frequently litigated. The denominator $c_k - c_f$ is small, because the customary procedure's chief input is the time of people who are assembled anyway, at gatherings that serve other purposes, so much of its apparent cost is jointly produced with activity that would occur regardless. And $n$ is small: a smallholding transferred by inheritance changes hands perhaps once in a generation. A threshold with a large numerator, a small denominator and a small realised $n$ is one that most parcels will not cross. This also explains a pattern the titling literature records repeatedly, and explains it as a mechanism rather than as fecklessness. Where the fee and travel cost of registering a transfer exceed the customary alternative, individual holders rationally transact customarily and leave the register unamended. But the register's value to any user depends on the proportion of transactions it captures: a register that records four transfers in five is still worth consulting, one that records one in five is not. Value therefore rises with use, and each defection lowers the return to everyone else's compliance. The system has two stable states — comprehensively used, or abandoned — and which one it settles into depends on the early fee structure rather than on the merits of registration. What path dependence licenses, and what it does not That last observation is a path-dependence argument, and it is worth being precise about the concept, because the earlier version of this note used it in a way its authors do not permit. The undergraduate proposition is that competition selects efficient arrangements: what survives, survives because it works. W. Brian Arthur's 'Competing Technologies, Increasing Returns, and Lock-In by Historical Events' (1989) showed the exception. Where returns increase with adoption — where a thing becomes more valuable the more it is used — the outcome of competition is determined by early, possibly accidental, events, and the arrangement that locks in need not be the one that would have been best. Douglass North built this into institutional history in Institutions, Institutional Change and Economic Performance (1990), where institutions are the rules of the game and the informal ones change far more slowly than the formal ones. Here is the correction. North's thesis is emphatically not that surviving institutions are efficient; it is the opposite. Path dependence is his