Since the early 1950s, Japanese farmland rents have been regulated and a consensus emerged that rent control led to market failure. Hypothesising a rent-formation model where rents are determined by prices, this paper estimates a threshold autoregressive model which integrates three tests of market failure, namely, inefficiency, bias and asymmetry. There are four results. First, a long-run relationship exists between rents and prices, and the Japanese farmland rental market is efficient. Second, the rent-price elasticity is unity and the market is unbiased. Third, rents are Granger-caused by prices which supports the rent-formation model. Fourth, asymmetry exists where more rapid error-correction occurs immediately after policy reform when rent growth exceeds price growth by 3.6% or more, and rent control has benefitted tenants.