There is an interesting debate going on in Europe about the likely consequences of the TTIP (Transatlantic Trade and Investment Partnership). Much of the real debate is (or should be) about the proposed Investor-State dispute resolution (ISDS) and the desirability of regulatory harmonization when nations have different preferences about how these regulations should be designed. But there is also a fascinating numbers game going on, with alternative quantitative estimates deployed by pro- and anti-TTIP groups.
The studies used by the pro group tend to show positive, if small, GDP effects. Probably the best known among these is a study by Joseph Francois and his colleagues, according to which EU and US GDPs will rise by 0.5% and 0.4%, respectively, by 2027 (relative to the baseline scenario without TTIP). Francois et al use a standard computable general equilibrium model that assumes full employment and perfect competition (save for a few sectors where there are scale economies and monopolistic competition). Wisely, they stay away from some of the bells and whistles (e.g., induced learning and TFP gains) that have been used in the past to produce exaggerated benefits from trade agreements.
These results, however, have been challenged in a recent paper by Jeronim Capaldo. Capaldo uses a Keynesian model where output is demand-determined and finds that EU GDP would fall as a result of a decline in net exports. (But U.S. output would rise, since U.S. net exports increase.)
Now models are useful to the extent that they help us understand the consequences of different sets of assumptions and causal mechanisms. Numerical methods are useful to the extent that they give us a sense of ballpark estimates of likely quantitative magnitudes. I take the Francois et al. model to capture the standard comparative advantage intuition. The Capaldo model, by contrast, warns us to be careful about potential adverse effects in economies where there is considerable slack.
With that said, here are some pointers to how to think about these different sets of estimates.
1. The standard gains-from-trade argument is one about levels, not growth rates. Remove trade impediments, and the level of consumer welfare goes up. These gains rise generally with the tax of the trade barriers, so removing small barriers does not generate large gains. This is the main reason why the Francois et al. (and similar studies) yield relatively small numbers. Tariffs in manufactures are really low to begin with. And even though barriers in services are much higher, Francois et al. reasonably project (in their most aggressive scenario, from which the numbers above come) that at most 25% of those barriers will be done with.
My view has long been that numerical models that purport to show significant dynamic/growth effects are suspect. (See here for a discussion of one such example.) Dynamic effects in trade models tend to be highly fragile, and can be easily reversed by tweaking the assumptions appropriately. Not surprisingly, pro-trade pact models tend to choose assumptions on this score that magnify the economic gains.
As I mentioned, Francois et al. eschew these tactics, and they are to be commended for it. (This has left at least one European think tank unhappy. As they put it: “the [Francois et al.] estimate does not take account of known gains from trade that come through so-called dynamic effects. The model, for instance, does not estimate the positive effect on productivity that most trade economists hold to be an important effect of trade liberalisation.”)
2. There is one element in the Francois et al. model that introduces growth effect through the back door. Investment is endogenous and trade liberalization may induce capital accumulation and growth. I could not tell from the presentation in the report itself how big an issue this is. But I would worry if it is a large part of the projected additional increase in GDP by 2027.
The problem is that the change in GDP need not track change in economic welfare – especially in this case where investment requires sacrificing consumption. From an economic standpoint, the change in the capital stock that is induced by trade liberalization is not in itself a source of welfare gain and should not be counted as such. Otherwise, we wrap ourselves up in seeming paradoxes. (For example, in a capital-scarce economy, trade protection would be expected to raise the return to capital and lead to higher investment and GDP in the future; yet this would not be a reason to cheer the increase in GDP.)
For this reason, I would have preferred to see explicit welfare measures in the Francois et al. study (such as an equivalent variation measure). I don’t know why there isn’t one.
3. Just as the standard gains-from-trade model says nothing about growth, it is also silent about aggregate employment. Full employment is the maintained assumption, so at most we can say something about the structure of employment.
Studies that want to claim employment gains (or in the Capaldo case, losses) have to make additional assumptions. Because these tend to be highly specific to the macro context, they can be quite unreliable. For example, an early study by the (Peterson) Institute for International Economics on NAFTA projected large U.S. trade surpluses vis-a-vis Mexico and therefore significant U.S. employment gains. When the peso crisis struck in December 1994, the bilateral trade balance changed sign in short order. Opponents of NAFTA could now use the same methods to argue that the U.S. was a big loser!
Capeldo’s study seems to me to suffer from a similar problem, although of course he wants to argue against (not for) TTIP. Capeldo’s Keynesian model projects a reduction in EU net exports, and therefore losses in terms of GDP and employment, into the medium term. Even if I am willing to buy the Keynesian argument in the short term, I am not sure I want to take it as my baseline case for the next decade.
Moreover, the Keynesian perspective renders trade liberalization a purely zero-sum affair. If the EU loses, the U.S. must win – and this is what we see in Capeldo’s numbers. Again, this seems to me to be an inappropriate perspective for anything other than the short run.
4. One of the things that always bothers me about quantitative discussion of trade pacts are the claims about the increase in trade volumes that are bandied about. Standard numerical models can generate large changes in export and import volumes even if they yield small welfare gains. A frequent tactic by trade-pact proponents therefore is to focus attention on the trade effects rather than welfare (or GDP) effects.
But the volume of trade has no economic significance. One dollar of output that is exported is no more (or less) valuable to the home economy than one dollar of output that is consumed at home. So beware arguments that revolve around trade numbers.
5. My bottom line is that I am happy to accept the Francois et al. estimates as my baseline for the economic effects of TTIP a decade down the line (assuming my capital accumulation caveat turns out to be quantitatively not important). But I would put such large margins of error on either side of them that they would not be dispositive for policy purposes.
As I said at the outset, the real issues lie elsewhere – in the broader social/political consequences of regulatory harmonization and the appropriateness of an ISDS regime in the North Atlantic region. I have serious concerns in both areas, and perhaps I will get to write about them too at some point.