The New Keynesian DSGE models that dominated the macroeconomic profession and central bank thinking for the last two decades were based on several principles. The first was formal derivation from micro-foundations, assuming optimising behaviour of consumers and firms with rational or ‘model-consistent’ expectations of future conditions. For such derivation to result in a tractable model, it was assumed that the behaviour of firms and of consumers corresponded to that of a ‘representative’ firm and a ‘representative’ consumer. In turn, this entailed the absence of necessarily heterogeneous credit or liquidity constraints. Another important assumption to obtain tractable solutions was that of a stable long-run equilibrium trend path for the economy. If the economy was never far from such a path, the role of uncertainty would necessarily be limited. Popular pre-financial crisis versions of the model excluded banking and finance, taking as given that finance and asset prices were merely a by-product of the real economy. Second, a competitive economy was assumed but with a number of distortions, including nominal rigidities – sluggish price adjustment – and monopolistic competition. This is what distinguished New Keynesian DSGE models from the general equilibrium real business cycle (RBC) models that preceded them. It extended the range of stochastic shocks that could disturb the economy from the productivity or taste shocks of the RBC model. Finally, while some models calibrated (assumed) values of the parameters, where the parameters were estimated, Bayesian system-wide estimation was used, imposing substantial amounts of prior constraints on parameter values deemed ‘reasonable’.
The ‘pretence of knowledge’
The centre-piece of Paul Romer’s scathing attack on these models is on the ‘pretence of knowledge’ (Romer 2016); echoing Caballero (2010), he is critical of the incredible identifying assumptions and ‘pretence of knowledge’ in both Bayesian estimation and the calibration of parameters in DSGE models.1 A further symptom of the ‘pretence of knowledge’ is the assumed ‘knowledge’ that these parameters are constant over time. A milder critique by Olivier Blanchard (2016) points to a number of failings of DSGE models and recommends greater openness to more eclectic approaches.
As explained in Muellbauer (2016), an even deeper problem, not seriously addressed by Romer or Blanchard, lies in the unrealistic micro-foundations for the behaviour of households embodied in the ‘rational expectations permanent income’ model of consumption, an integral component of these DSGE models. Consumption is fundamental to macroeconomics both in DSGE models and in the consumption functions of general equilibrium macro-econometric models such as the Federal Reserve’s FRB-US. At the core of representative agent DSGE models is the Euler equation for consumption, popularised in the highly influential paper by Hall (1978). It connects the present with the future, and is essential to the iterative forward solutions of these models. The equation is based on the assumption of inter-temporal optimising by consumers and that every consumer faces the same linear period-to-period budget constraint, linking income, wealth, and consumption. Maximising expected life-time utility subject to the constraint results in the optimality condition that links expected marginal utility in the different periods. Under approximate ‘certainty equivalence’, this translates into a simple relationship between consumption at time t and planned consumption at t+1 and in periods further into the future.
Under these simplifying assumptions, the rational expectations permanent income consumption function can be derived. In the basic form, consumption every period equals permanent non-property income plus permanent property income defined as the real interest rate times the stock of wealth held by consumers at the beginning of each period. Permanent non-property income converts the variable flow of labour and transfer incomes a consumer expects over a lifetime into an amount equally distributed over time.
However, consumers actually face idiosyncratic (household-specific) and uninsurable income uncertainty, and uncertainty interacts with credit or liquidity constraints. The asymmetric information revolution in economics in the 1970s for which Akerlof, Spence and Stiglitz shared the Nobel prize explains this economic environment. Research by Deaton (1991,1992),2 several papers by Carroll (1992, 2000, 2001, 2014), Ayigari (1994), and a new generation of heterogeneous agent models (e.g. Kaplan et al. 2016) imply that household horizons then tend to be both heterogeneous and shorter – with ‘hand-to-mouth’ behaviour even by quite wealthy households, contradicting the textbook permanent income model, and hence DSGE models. A second reason for the failure of these DSGE models is that aggregate behaviour does not follow that of a ‘representative agent’. Kaplan et al. (2016) show that, with these better micro-foundations, quite different implications follow for monetary policy than in the New Keynesian DSGE models. A third reason is that structural breaks, as shown by Hendry and Mizon (2014), and radical uncertainty further invalidate DSGE models, illustrated by the failure of the Bank of England’s DSGE model both during and after the 2008-9 crisis (Fawcett et al. 2015). The failure of the representative agent Euler equation to fit aggregate data3 is further empirical evidence against the assumptions underlying the DSGE models, while evidence on financial illiteracy (Lusardi 2016) is a problem for the assumption that all consumers optimise.
The evolving credit market architecture
Of the structural changes, the evolution and revolution of credit market architecture was the single most important. In the US, credit card ownership and instalment credit spread between the 1960s and the 2000s; the government-sponsored enterprises – Fannie Mae and Freddie Mac – were recast in the 1970s to underwrite mortgages; interest rate ceilings were lifted in the early 1980s; and falling IT costs transformed payment and credit screening systems in the 1980s and 1990s. More revolutionary was the expansion of sub-prime mortgages in the 2000s, driven by rise of private label securitisation backed by credit default obligations (CDOs) and swaps. The 2000 Commodity Futures Modernization Act (CFMA) made derivatives enforceable throughout the US with priority ahead of claims by others (e.g. workers) in bankruptcy. This permitted derivative enhancements for private label mortgage-backed securities (PMBS) so that they could be sold on as highly rated investment grade securities. A second regulatory change was the deregulation of banks and investment banks. In particular, the 2004 SEC decision to ease capital requirements on investment banks increased gearing to what turned out to be dangerous levels and further boosted PMBS, Duca et al (2016). Similar measures to lower required capital on investment grade PMBS increased leverage at commercial banks. These changes occurred in the political context of pressure to extend credit to poor.
The importance of debt
A fourth reason for the failure of the New Keynesian DSGE models, linking closely with the previous, is the omission of debt and household balance sheets more generally, which are crucial for understanding consumption and macroeconomic fluctuations. Some central banks did not abandon their large non-DSGE econometric policy models, but these were also defective in that they too relied on the representative agent permanent income hypothesis which ignored shifts in credit constraints and mistakenly lumped all elements of household balance sheets, debt, liquid assets, illiquid financial assets (including pension assets) and housing wealth into a single net worth measure of wealth.4 Because housing is a consumption good as well as an asset, consumption responds differently to a rise in housing wealth than to an increase in financial wealth (see Aron et al. 2012). Second, different assets have different degrees of ‘spendability’. It is indisputable that cash is more spendable than pension or stock market wealth, the latter being subject to asset price uncertainty and access restrictions or trading costs. This suggests estimating separate marginal propensities to spend out of liquid and illiquid financial assets. Third, the marginal effect of debt on spending is unlikely just to be minus that of either illiquid financial or housing wealth. The reason is that debt is not subject to price uncertainty and it has long-term servicing and default risk implications, with typically highly adverse consequences.
The importance of debt was highlighted in the debt-deflation theory of the Great Depression of Fisher (1933).5 Briefly summarised, his story is that when credit availability expands, it raises spending, debt, and asset prices; irrational exuberance raises prices to vulnerable levels, given leverage; negative shocks can then cause falls in asset prices, increased bad debt, a credit crunch, and a rise in unemployment. In the 1980s and early 1990s, boom-busts in Norway, Finland, Sweden, and the UK followed this pattern. In the financial accelerator feedback loops that operated in the US sub-prime crisis, falls in house prices increased bad loans and impaired the ability of banks to extend credit. As a result, household spending and residential investment fell, increasing unemployment and reducing incomes, feeding back further into lower asset prices and credit supply. The transmission mechanism that operated via consumption was poorly represented by the Federal Reserve’s FRB-US model and similar models elsewhere. A more relevant consumption function for modelling the financial accelerator is needed, modifying the permanent income model with shorter time horizons,6 incorporating important shifts in credit lending conditions, and disaggregating household balance sheets into liquid and illiquid elements, debt and housing wealth.
Implications for macroeconomic policy models
To take into account all the feedbacks, a macroeconomic policy model needs to explain asset prices and the main components of household balance sheets, including debt and liquid assets. This is best done in a system of equations including consumption, in which shifts in credit conditions – which have system-wide consequences, sometimes interacting with other variables such as housing wealth – are extracted as a latent variable.7 The availability of home equity loans, which varies over time and between countries – hardly available in the US of the 1970s or in contemporary Germany, France or Japan – and the also the variable size of down-payments needed to obtain a mortgage, determine whether increases in house prices increase (US and UK) or reduce (Germany and Japan) aggregate consumer spending. This is one of the findings of research I review in Muellbauer (2016). Another important finding is that a rise in interest rates has different effects on aggregate consumer spending depending on the nature of household balance sheets. Japan and Germany differ radically from the US and the UK, with far higher bank and saving deposits and lower household debt levels so that lower interest rates reduce consumer spending. A crucial implication of these two findings is that monetary policy transmission via the household sector differs radically between countries – it is far more effective in the US and UK, and even counterproductive in Japan (see Muellbauer and Murata 2011).
Such models, building in disaggregated balance sheets and the shifting, interactive role of credit conditions, have many benefits: better interpretations of data on credit growth and asset prices helpful for developing early warning indicators of financial crises; better understandings of long-run trends in saving rates and asset prices; and insights into transmission for monetary and macro-prudential policy. Approximate consistency with good theory following the information economics revolution of the 1970s is better than the exact consistency of the New Keynesian DSGE model with bad theory that makes incredible assumptions about agents’ behaviour and the economy. Repairing central bank policy models to make them more relevant and more consistent with the qualitative conclusions of the better micro-foundations outlined above is now an urgent task.
Aron, J., J. Duca, J. Muellbauer, K. Murata, and A. Murphy (2012), “Credit, Housing Collateral and Consumption in the UK, US, and Japan.” Review of Income and Wealth 58 (3): 397-423.
Aiyagari, S. R (1994), “Uninsured Idiosyncratic Risk and Aggregate Saving.” The Quarterly Journal of Economics 109 (3): 659-684.
Blanchard, O. (2016), “Do DSGE Models have a Future?” Policy Brief 16-11: 1-4. Washington: Peterson Institute for International Economics.
Brayton, F., E. Mauskopf, D. Reifschneider, P. Tinsley, and J. Williams (1997), “The Role of Expectations in the FRB/US Macroeconomic Model.” Federal Reserve Bulletin, April: 227-245.
Caballero, R. J. (2010), “Macroeconomics after the Crisis: Time to Deal with the Pretense-of-Knowledge Syndrome.” National Bureau of Economic Research Working Paper 16429.
Campbell, J. and G. Mankiw (1989), “Consumption, Income and Interest Rates; Reinterpreting the Time Series Evidence.” In NBER Macroeconomics Annual 1989, edited by O. Blanchard and S. Fischer, 185 – 246. Cambridge, Mass: MIT Press.
Campbell, J., and G. Mankiw (1991), “The Response of Consumption to Income: A Cross-Country Investigation.” European Economic Review 35 (4): 723-56.
Carroll, C. D. (1992), “The Buffer-Stock Theory of Saving: Some Macroeconomic Evidence.” Brookings Papers on Economic Activity 23 (2): 61-135.
Carroll, C. D. (2000), “Requiem for the Representative Consumer? Aggregate Implications of Microeconomic Consumption Behavior.” American Economic Review Papers and Proceedings 90 (2): 110-115.
Carroll, C. D. (2001), “A Theory of the Consumption Function, With and Without Liquidity Constraints.” Journal of Economic Perspectives 15 (3): 23-45.
Carroll, C. D. (2014), ”Representing Consumption and Saving Without a Representative Consumer.” in Measuring Economic Sustainability and Progress. NBER Studies in Income and Wealth. Cambridge, Mass: National Bureau of Economic Research.
Deaton, A. (1991), “Saving and Liquidity Constraints.” Econometrica 59 (5): 1221-48.
Deaton, A. (1992), Understanding Consumption. Oxford: Oxford University Press.
Duca, J., Muellbauer, J., and A. Murphy (2016), “How Mortgage Finance Reform Could Affect Housing.” American Economic Review Papers & Proceedings 10 (5): 620–624.
Fawcett, N., R. Masolo, L. Koerber, and M. Waldron (2015), “How did the Bank’s Forecasts Perform Before, During and After the Crisis?” Bank Underground, Bank of England.
Fisher, I. (1933), “The Debt-Deflation Theory of Great Depressions.” Econometrica 1 (4): 337-57.
Hall, R. E. (1978), “Stochastic Implications of the Life Cycle-Permanent Income Hypothesis: Theory and Evidence.” Journal of Political Economy 86 (6): 971–87.
Hendry, D. F. and G. E. Mizon (2014), "Why DSGEs crash during crises", VoxEU.org, 18 June.
Jordà, O., M. Schularick, and A. M. Taylor (2016), "The Great Mortgaging: Housing Finance, Crises and Business Cycles." Economic Policy 31 (85): 107-152.
Kaplan, G., B. Moll, and G.I. Violante (2016), “Monetary Policy According to HANK.” NBER Working Paper 21897.
Laubach, T., and J. C. Williams (2003), “Measuring the Natural Rate of Interest.” The Review of Economics and Statistics 85 (4): 1063-1070.
Lusardi. A. (2016), “Financial Literacy and Ignorance.” Blog.
Mian, A., and A. Sufi (2014), House of Debt. Chicago: University of Chicago Press.
Muellbauer, J. (2010), “Household Decisions, Credit Markets and the Macroeconomy: Implications for the Design of Central Bank Models.” Bank for International Settlements Discussion Paper 306, March.
Muellbauer, J., and K. Murata (2011b), “Mistaken Monetary Policy Lessons from Japan?”, VoxEU.org, 21 August.
Muellbauer, J. (2016), “Macroeconomics and Consumption" CEPR DP 11588, and Oxford University, Department of Economics working paper 811,
Romer, P. (2016), “The Trouble with Macroeconomics.” Mimeo, Stern School of Business, New York University, 14th September, 2016. (Forthcoming in The American Economist)
Shea, John (1995), "Union Contracts and the Life-Cycle/Permanent-Income Hypothesis." American Economic Review 85(1): 186-200.
Turner, A. (2015), Between Debt and the Devil: Money, Credit, and Fixing Global Finance. Princeton, NJ: Princeton University Press.
 Part of the problem of identification is that the DSGE models throw away long-run information. They do this by removing long-run trends with the Hodrick-Prescott filter, or linear time trends specific to each variable. Identification, which rests on available information, then becomes more difficult, and necessitates ‘incredible assumptions’. Often, impulse response functions tracing out the dynamic response of the modelled economy to shocks are highly sensitive to the way the data have been pre-filtered.
 This important research was highly praised in Angus Deaton’s 2015 Nobel prize citation: http://www.nobelprize.org/nobel_prizes/economic-sciences/laureates/2015/advanced.html
 See Campbell and Mankiw (1989, 1990) and for even more powerful evidence from the UK, US and Japan; Muellbauer (2010); and micro-evidence from Shea (1995).
 Net worth is defined as liquid assets minus mortgage and non-mortgage debt plus illiquid financial assets plus housing assets, and this assumes that the coefficients are all the same.
 In recent years, several empirical contributions have recognised the importance of the mechanisms described by Fisher (1933). Mian and Sufi (2014) have provided extensive micro-economic evidence for the role of credit shifts in the US sub-prime crisis and the constraining effect of high household debt levels. Focusing on macro-data, Turner (2015) has analysed the role of debt internationally with more general mechanisms, as well as in explaining the poor recovery from the global financial crisis. Jorda et al. (2016) have drawn attention to the increasing role of real estate collateral in bank lending in most advanced countries and in financial crises.
 The FRB-US model does build in shorter average horizons than text-book permanent income. It also has a commendable flexible treatment of expectations, Brayton et al (1997).
 The use of latent variables in macroeconomic modelling has a long vintage. Potential output, and the “natural rate” of unemployment or of interest are often treated as latent variables, for example in the FRB-US model and in Laubach and Williams (2003), and latent variables are often modelled using state space methods. Flexible spline functions can achieve similar estimates. Interaction effects of latent with other variables seem not to have been considered, however. We use the term ‘latent interactive variable equation system’ (LIVES) to describe the resulting format.