(1975), Clinton (1988), Taylor (1989), Peel and Taylor (2002) and Obstfeld and Taylor
(2004). Transfer points measure the minimum interest differentials that would induce
arbitrage (because of the existence of transaction costs, controls etc). These transfer
points have been estimated through data on triangular arbitrage, bid-ask spreads and
brokerage fees and endogenously through a Threshold Autoregressive (TAR) model in
Peel and Taylor (2002) and Obstfeld and Taylor (2004).
Clinton (1988) estimated, for 5 major currencies against the US$, a transfer band
of ±6 basis points in the mid-1980’s. He rejected the null of zero deviations from parity,
but found that any such deviations were small and short lived. Frenkel and Levich (1975)
estimate transaction costs in foreign exchange markets through triangular arbitrage
differentials and in securities markets through bid-ask spreads and find that these explain
85 per cent of the deviations from CIP. The rest could be explained to a large extent by
less than perfectly elastic demand and supply curves and by lags between observing a
profitable opportunity and arbitraging. Obstfeld and Taylor (2004) and Peel and Taylor
(2002) use a TAR methodology to estimate transfer bands of ± 19 basis points for New
York London and ± 35 basis points for London-Berlin transactions for the period 1880-
1914, and bands of about ± 50 basis points for the 1920’s. Deviations outside the bands
do occur in their samples, but these tend to be mean reverting. These suggest a much
higher degree of integration in the post-Bretton Woods era than what was attained in the
pre-1914 period of financial globalization. Moreover, Dooley and Isard (1980) showed
that much of the differentials could be accounted for by political risks, including capital
controls and risks of their imposition.
CIP has also been tested as H'0 : a = 0, b = 1 in: