The limitations of the benchmark E-DSGE framework and how these limitations restrict the ability of this framework to meaningfully capture the macroeconomics of the climate crisis.
Dynamic stochastic general equilibrium (DSGE) models have been heavily criticized since the Global Financial Crisis for being structurally incapable of analyzing real-world dynamics (Stiglitz 2018; Storm 2021). However, they continue to be the dominant macroeconomic modeling approach in the academic and policymaking world. More recently, DSGE models have been increasingly used to analyze the macroeconomic implications of the climate crisis and assess climate policies. In this INET Working Paper, we analyze why these environmental dynamic stochastic general equilibrium (E-DSGE) models are not fit for purpose.
We first focus on the benchmark E-DSGE modeling framework with finance and identify the following six limitations.
1. Banks are misrepresented as pure financial intermediaries and macro-financial feedback loops play a limited role. In the benchmark E-DSGE models, banks are not money-creating institutions but ‘pure intermediaries’, which mobilize savings to be able to provide loans. As a result, private sector savings drive investment. It is now well-accepted that this is a misrepresentation of how the banking system works. In reality, banks can create money endogenously by expanding their balance sheets with no need to rely on prior savings (McLeay et al., 2014; Deutsche Bundesbank, 2017). In addition, in the benchmark E-DSGE framework, banks’ borrowers do not default on their debt, and, therefore, the capital of banks is not affected by macro-financial conditions, restricting the role of feedback loops. Why do these assumptions matter from a climate perspective? First, climate policies, such as carbon taxes, green subsidies, and environmental regulation, are less likely to cause substantial macro-financial amplification effects (e.g. green credit booms and dirty credit busts) in DSGE models: banks rely on pre-existing savings and cannot, therefore, be an important driver of macroeconomic fluctuations by expanding and shrinking their balance sheets. Second, excluding the possibility of default of carbon-intensive or climate-vulnerable companies does not allow for the decline in the capital of banks (as a result of defaults) to reduce loan provision and generate macro-financial feedback loops. Therefore, E-DSGE models significantly under-estimate the emergence and implications of climate-related macro-financial instability contrary to what is the case in other modeling approaches (Dafermos, Nikolaidi, and Galanis 2018) and the work of the Network for Greening the Financial System (NGFS) (NGFS 2019).
2. Demand has a restricted impact on economic activity. In DSGE models, demand can only play a role in the short run when imperfect competition and nominal and real rigidities are present (Storm 2021). In the long run, demand is irrelevant since what is produced is assumed to be demanded (Say’s law). This is also the case in the environmental versions of DSGE models. Due to Say’s law, an increase in green investment spending is possible only if other components of aggregate demand go down so as to create savings for funding green investment. As a result, green investment cannot have expansionary effects, making the net zero transition look more costly from a GDP perspective compared to what might be the case in reality.
3. Disequilibrium phenomena cannot be adequately analyzed. The benchmark E-DSGE modeling framework adopts the rational expectations hypothesis (i.e. representative agents, on average, have perfect knowledge of the future) and, due to the lack of hysteresis, it assumes that long-run outcomes are independent of short-run developments. As a result of the rational expectations hypothesis, carbon pricing can have very unrealistic effects on inflation and GDP (it can increase inflation and decrease GDP contrary to what is expected in practice). Moreover, as a result of both assumptions, DSGE models cannot capture (i) green structural change driven by higher green spending in the short run and (ii) the economic implications of increasing climate damages that tend to create disequilibrium effects.
4. The substitutability assumptions in the production function and portfolio choices are unrealistic. In E-DSGE models there is a fixed substitutability between fossil and non-fossil energy. This makes it very difficult for environmental policies to achieve a phase-out of fossil-based energy through structural change (Yanovski and Lessmann 2021). Moreover, the assumption of perfect substitutability in the portfolio choice of households and banks implies that any changes in the returns on assets caused by monetary policy have no impact on investment and economic activity. Due to this, the introduction of any green quantitative easing program (or other green monetary policies that have an impact on financial markets) has either no economic and financial effects or only some short-run effects when portfolio adjustment costs are introduced. As a result, the environmental effects of green monetary policies are unrealistically negligible even if these policies are permanent.
5. The positive effects of expansionary fiscal policy on economic activity are limited. In DSGE models additional public spending can be funded (i) through the issuance of more government debt or (ii) by raising more taxes from the public. In the first case, the issuance of more debt can increase interest rates that have a negative impact on private investment spending. In the second case, households and firms that pay higher taxes end up spending less. The result is that green public investment has unrealistically high crowding-out effects in the E-DSGE framework and the positive complementarities between private and public investment in the context of the green transition are fully ignored. The benchmark E-DSGE framework also ignores the positive impact of green public investment on the reduction of climate economic damages in the long run.
6. Policies identified by E-DSGE models with frictions and rigidities are not Pareto-optimal. This is an internal consistency limitation. When climate damages (a negative environmental externality) are introduced in the benchmark E-DSGE framework where rigidities and financial frictions are already present, E-DSGE models deal with a ‘second-best’ world. The Theory of the Second-Best states that if all of the distortions in the economy cannot be eliminated, all bets are off: internalizing climate damage might raise welfare, but can just as easily reduce welfare (Rezai, Foley, and Taylor 2012). Therefore, the concept of an ‘optimal’ carbon tax is meaningless, and, more broadly, the social welfare implications of E-DSGE analyses are unclear.
Recent years have seen attempts from (E-)DSGE modelers to address some of these limitations. However, these attempts focus on dealing with some of these limitations in isolation. As we explain in the paper, this cannot resolve the problem: because of the interconnected nature of the limitations, a specific limitation cannot be addressed by simply relaxing a specific assumption and keeping the rest of the structural assumptions of DSGE models the same.
Here are two examples. First, consider an E-DSGE model that incorporates endogenous money, but keeps Say’s Law according to which supply creates its own demand. This cannot meaningfully address the misrepresentation of banks as financial intermediaries in E-DSGE models. In reality, an implication of endogenous money is that consumption, investment, and government spending (as sources of demand) can always increase as long as finance is available ─ this spending determines the level of output in the economy irrespective of pre-existing savings. Of course, supply constraints can lead to inflationary pressures, but this does not change the fact that newly-created money can lead to higher levels of economic activity.
Second, suppose that an E-DSGE model includes public capital in the production function (allowing public investment to have a direct impact on the steady-state output via the supply side of the economy), but keeps the rest of the benchmark E-DSGE model unchanged. Such a model would continue to downplay the positive impact of green expansionary fiscal policy on GDP since four other significant assumptions would remain the same: (i) private investment competes with public investment for pre-existing financial resources; (ii) demand is not a driver of economic activity in the long run and thus green fiscal spending does not have a direct impact on output as a source of demand; (iii) agents have rational expectations about the repayment of public debt and there are no hysteresis effects of government spending, and (iv) green public infrastructure cannot lead to a large-scale replacement of fossil energy with non-fossil energy and, thus, it cannot allow climate damages to decrease as a result of green public investment.
A reaction to this would be to try to develop an E-DSGE model that addresses all these limitations at the same time. However, doing so would be extremely difficult within the DSGE framework or would lead to models that would no longer satisfy the normative criteria that any DSGE model worth the name should satisfy.
We, therefore, call for a broader use of other macroeconomic models, such as ecological stock-flow consistent (E-SFC) models (see Dafermos, Nikolaidi, and Galanis 2017) and ecological agent-based (E-AB) models (Lamperti et al. 2018), that simultaneously address the limitations that characterize E-DSGE models. In these models, money is endogenous, demand plays an active role both in the short run and the long run, disequilibrium phenomena can be easily analyzed, there are no perfect substitutability assumptions in the production function and financial markets, and expansionary fiscal policy does not have unrealistic crowding-out effects.
These models are not without their own limitations, of course. For example, the calibration of parameter values is not sufficiently developed yet, there are no standard techniques to present sensitivity results to check the robustness of the results and the modeling of the behavioral equations of the models is sometimes considered to be relatively arbitrary and without forward-looking elements. However, further research can address most of these limitations and make these models useful tools for policymaking in the Anthropocene. E-DSGE models are, in any case, not fit for this purpose.
References
Dafermos, Y., Nikolaidi, M. and Galanis, G., 2017. A stock-flow-fund ecological macroeconomic model, Ecological Economics, 131, pp. 191-207.
Dafermos, Y., Nikolaidi, M. and Galanis, G., 2018. Climate change, financial stability and monetary policy, Ecological Economics, 152, pp. 219-234.
Deutsche Bundesbank, 2017. The role of banks, non-banks and the central bank in the money creation process, Deutsche Bundesbank Monthly Report April 2017.
Lamperti, F., Dosi, G., Napoletano, M., Roventini, A. and Sapio, A., 2018. Faraway, so close: Coupled climate and economic dynamics in an agent-based integrated assessment model, Ecological Economics, 150, pp. 315-339.
McLeay, M., Radia, A. and Thomas, R., 2014. Money creation in the modern economy, Bank of England Quarterly Bulletin, Q1.
NGFS, 2019. A call for action: Climate change as a source of financial risk, Network for Greening the Financial System, April.
Rezai, A., Foley, D.K. and Taylor, L., 2012. Global warming and economic externalities, Economic Theory 49, 329-351.
Stiglitz, J.E., 2018. Where modern macroeconomics went wrong, Oxford Review of Economic Policy, 34 (1-2), pp. 70-106.
Storm, S., 2021. Cordon of conformity: Why DSGE models are not the future of macroeconomics. International Journal of Political Economy, 50 (2), pp. 77-98.
Yanovski, B. and Lessmann, K., 2021. Financing the fossil fuel phase-out, SSRN Paper.
