Copy, Paste, Negotiate

Nikita Melashchenko · (18 January 2026) · opinion

There is a finding in the treaty design literature that I keep coming back to. Todd Allee and Manfred Elsig ran a text analysis over several hundred preferential trade agreements and found that most of them copy a sizable majority of their content word for word from an earlier agreement. At least one hundred PTAs take 80 per cent or more of their text from a single existing treaty, and many copy 95 per cent or more.^[1] The numbers climb higher still when you compare the substantive chapters that practitioners actually care about.

The first reaction to this is usually confusion. Are treaties not supposed to be individually crafted expressions of sovereign bargains? My reaction is the opposite. Having spent a good part of my research exploring datasets and coding FTA provisions into datasets, I can report that the copying is not a dirty secret of the trading system. It is the trading system.

Consider who copies, and why. Allee and Elsig find that replication is most prevalent at both ends of the power spectrum. Low-capacity governments lean on templates because drafting a two-hundred-page services chapter from scratch is expensive, and a treaty that resembles a known model is easier to sell at home and abroad. Powerful states copy for the opposite reason. A template is a distribution mechanism for their preferred rules. When the same source code disclosure clause appears in a dozen agreements, that is not laziness. That is somebody’s regulatory preference achieving escape velocity.

For a country like New Zealand, sitting firmly in the small-but-competent category, the template economy is mostly good news. We did not have to invent the CPTPP electronic commerce chapter to benefit from it, and the DEPA took the logic further by packaging digital trade rules into modules explicitly designed to be lifted. If treaty text is going to be copied anyway, a small state’s best move is to write the paragraph everyone else copies. That is influence disproportionate to market size, available at the price of drafting quality.

The methodological lesson matters just as much. If most of a new agreement is inherited text, then reading a treaty from page one is the wrong way to read it. The signal is in the diff. Where did the parties depart from the template? Which clause got quietly deleted, which exception got an extra limb, which “shall” became “shall endeavour”? Those deviations are the negotiation; everything else is the container it arrived in. This is also why coding treaties into structured data is not an accountant’s hobby but a reading strategy. You cannot spot the deviations across a few hundred agreements by eyeballing PDFs.

So the next time a freshly signed FTA lands, try treating it the way a developer treats a forked repository. Disregard the length. Run the diff, and spend your attention on the twenty pages that are actually new.