The Comfort of Conformity: How Hegemony in Economic Modeling Precludes a Better World

By Jeronim Capaldo

Is the US economy running too hot, such that more spending is bound to drive up inflation with no benefit for growth and employment? Many experts think so. Would more trade be beneficial for economic growth and employment? On this, too, many agree.

In answering these and other macroeconomic questions, multiple sources appear to independently reaffirm the same conclusions, ostensibly demonstrating a strong consensus. However, most of these voices rely on the same economic model, merely repeating the same findings rather than truly confirming them.

Economic models are fairly mechanical contraptions that, based on certain assumptions on economic and social behavior, provide projections and simulations. Two of the most influential types are Dynamic Stochastic General Equilibrium (DSGE) models and Computable General Equilibrium (CGE) models. Both have dismal records in guiding actual policy choices, but have come to dominate policy debates: the former on fiscal and monetary policy, the latter on trade. With these tools at their disposal, it is not surprising that in too many countries – including the world’s richest economies – economic policy has followed the same strategy for decades, a strategy based on fiscal austerity, flexibility in the labor market and trade liberalization. Neither is it surprising that, after a year of upended conventions in policymaking, that same strategy is now resurfacing as a remedy against inflation and debt fear.

While problems with the prevailing trade models have long been recognized (and largely ignored), a systematic and realistic assessment of DSGE models was, until recently, missing. Servaas Storm changed that, with a sharp analysis bringing to light many of the models’ fatal flaws. His article was deemed so important that the International Journal of Political Economy (IJPE) devoted a special issue to it, with commentaries and responses.

Storm’s analysis – and the discussion it generated – clarified that one major cost of the DSGE hegemony is that it limits our understanding of the macroeconomy. The reliance on these models has crowded out alternative approaches that could do a better job at explaining critical aspects of reality such as employment growth, inequality and inflation. In academia, this is visible in undergraduate and graduate curricula, as well as in influential journals and (related) faculty hires.

But just as importantly, as I argue in my contribution to the special issue, DSGE hegemony also prevents a host of sensible policy discussions within states, international fora and civil society, at great peril for global economic stability. If these models were, say, a pharmaceutical product, their aggressive marketing – which includes strong-arming “clients” to buy into them – would earn them plenty of lawsuits. Instead, they are regarded as “state of the art” tools to inform policy decisions.

A closer look at the DSGE and CGE Models

The longest standing example of the influence of DSGE models on policy is the discussion on potential output and its effects on fiscal policy. If an economy stagnates long enough and its productive capacity is partially destroyed, as has been the case in many developing countries and in the European Union (EU) after 2008, the models will detect a small output gap and advise against fiscal expansion. It is with these tools that austerity has been enforced in the EU since 2011 and is now being discussed again. In general, rigid fiscal (as well as monetary) rules have emerged from DSGE-based research programs.

A similar state of affairs has emerged in trade policy and industrial policy with respect to Computable General Equilibrium models (CGEs). The prevailing type of CGE assumes away unemployment and inequality and, often, productivity growth – some of the most critical features of economic reality. Whereas DSGE models monopolize domestic macro policy, CGEs are virtually the only models used for global trade analysis. They preclude sensible discussions on trade policy, as the simulations they produce systematically point to GDP gains from trade liberalization, even when the latter is pursued through regulatory changes that are impossible to quantify. This can be done because unemployment and wage repression, which in reality have been major obstacles to aggregate demand growth, are ignored.

A case study: Italy’s six-year recovery plan

A recent example of the workings and the hegemony of these “evil twins” of economic modeling is offered by Italy’s six-year COVID-19 recovery plan. The document illustrating the plan includes macroeconomic projections from two different models: a DSGE and a CGE. The DSGE projects a somewhat linear expansion of GDP, while the CGE projects faster growth in the early years, followed by a stabilization at lower growth rates. The difference between the predictions matters greatly, as many things can happen after two or three years, when the difference between the growth rates projected by the two models is greatest. At that point, whether the economy is on the path projected by one or the other model, or neither, is critical. But the models are so influential that, even though they don’t support a clear policy action, they remain in use.

The disconnect between the DSGE world and reality greatly diminishes the relevance of the simulation. It is clear that, for the first three years, the fiscal multiplier is very small, possibly less than one (indicating that the costs of fiscal expansion will not be compensated with at least equal benefits in terms of GDP), despite two decades of austerity and stagnation that have been mostly marked by negative productivity and GDP growth, as well as the deepest recession on record in 2020. In similar conditions, a sudden injection of public investment is always strongly expansionary, but in the simulation, the multiplier only increases in the medium term. At first, this may seem consistent with the view (built into DSGE models) that growth mostly depends on supply side factors, such as infrastructure, which takes several years to build. However, the result is not only due to the model’s structure: in fact, the document adopts external estimates of the multipliers of public investment derived from European Commission and International Monetary Fund (IMF) simulations, which assume budget neutral boosts to investment.

In summary, the Italian plan entails deficit spending, but the DSGE simulation used to assess its macro impact uses parameters derived from a budget neutral exercise, which are bound to be much lower. In this confusion, one thing is clear: DSGE models and modelers have a hard time accounting for demand-driven growth.

The comfort of conformity

Macroeconomic policy is decided by governments and central banks, but experts armed with models have profound influence. When a challenge arises, both governments and central banks hear the opinions of other public officials and experts in intergovernmental fora, such as the G20, and read extended arguments in the publications of the IMF, the Organization for Economic Cooperation and Development (OECD), the World Bank or the United Nations. When a consensus emerges in these places, it understandably exerts influence on sovereign decisions. Because most governments must rely on other governments in a crisis, even if indirectly through the IMF or the European Commission, they have an incentive to conform to the prevailing opinion – a comfort in conformity. The trouble is that the prevailing opinion is informed by models that are clearly inadequate, such as DSGEs and CGEs.

Given the power that the existing approach wields, its influence on policy decisions can hardly be broken, even when officials know the models are inaccurate. Many ministry of finance officials must feel like the young Kenneth Arrow did when, serving in the US Army as a weather forecaster, he pointed out that the army’s models were doing no better than random guesses. He was reportedly told: “The commanding general is well aware that the forecasts are no good [but] he still needs them for planning purposes.”

Now, the most effective approach to breaking the influence of bad models on policy is to break their influence on prevailing expert opinion. Access to alternative models is critical to encourage policy discussions outside the path beaten by the DSGE orthodoxy. As the IJPE special issue shows, better options do exist.