Macroeconomic policy models should track the different channels of monetary transmission, providing a framework for Monetary Policy Committees. They should also be useful for assessing risks to financial stability, including designing macroprudential stress tests and instrument settings in the new macroprudential toolkits. Current policy models, including ‘semi-structural’ non-DSGE econometric models such as FRB-US, are seriously deficient in these respects. The paper explains the implications of three sets of evidence for the channels of monetary transmission, and how these differ between countries. First, the micro evidence on the marginal propensity to consume out of current income is that it is highest for the asset-poor, intermediate for households poor in liquid assets but not in illiquid assets, and lowest for the doubly asset-rich. Second, the micro evidence is that changes in house prices have mainly a collateral effect on consumption. Finally, multi-country evidence confirms large differences in the investment response to house prices. This discussion also illuminates the role of real estate in the financial accelerator, prominent in the global financial crisis. A multi-purpose policy model needs to include a consumption function consistent with the micro-evidence, with equations for permanent income, for the balance sheet drivers, and for residential investment, in a household-housing sub-system. This needs to embed common credit conditions in the equations, and sub-system estimation to impose the cross-equation restrictions implied by these common factors.