We show that vector error correction models encompass different approaches to analysing market integration; we illustrate our method using English weekly wheat prices, 1770–1820. Price variation decomposes into: (i) magnitude of price shocks; (ii) correlation of price shocks; (iii) between-period arbitrage. Data frequency affects these components, but has the largest effect on between-period arbitrage, commonly measured by half-life. Since this measure has been generally employed, previous analyses should be interpreted with caution. We further show that estimated effects of better transport and communication depend on the model used to measure market integration. Notably, we observe market integration improvements, not in between-period arbitrage, but in the within-week behaviour of prices (i.e. over much shorter time periods). So transport impacted English market integration, but in a way not captured by half-lives.