Discussion of Why Has Consumption Remained Moderate after the Great Recession? Federal Reserve Bank of Boston 60 th Economic Conference Karen Dynan Assistant Secretary for Economic Policy U.S. Treasury Department October 14, 2016 Notes for the slides can be found at the end of the presentation.
Economists have cited a number of factors that may have held back aggregate consumption growth during the recovery Slow income growth The drop in household wealth resulting from the crisis High household leverage/debt Credit constraints Greater desire for precautionary savings Lower expected future income growth Scarring, financial frictions, higher inequality and other factors that may have muted the response of consumption to positive innovations in its traditional determinants U.S. Treasury, Office of Economic Policy 1
The paper draws from a truly impressive array of different data sources to try to shed light on the roles played by different factors Current Population Survey (BLS) Home Mortgage Disclosure Act (FFIEC) Senior Loan Officer Opinion Survey (FRB) Survey of Consumer Finances (FRB) Survey of Consumer Expectations (FRB-NY) Income Concentration Statistics (Piketty and Saez) Cash-out Refinance Mortgages (Freddie Mac) Foreclosures (Zillow.com) Household Debt Service Ratio (FRB) U.S. Treasury, Office of Economic Policy Wharton Residential Land Use Regulation Index (Gyourko, Saiz, and Summers, 2008) National Income and Product Accounts (BEA) American Time Use Survey (BLS) Housing Affordability Index (California Association of Realtors) Surveys of Consumers (University of Michigan) Panel Study of Income Dynamics (University of Michigan) Homeownership (Census) Financial Accounts of the United States (FRB) 2
What is this paper asking? Has consumption growth been weak in some absolute sense? Has consumption growth been weak relative to what traditional macro consumption models would suggest? The paper explores both but this is the more interesting question from a policy perspective because of the reliance of policymakers on traditional macro models. U.S. Treasury, Office of Economic Policy 3
Summary of findings The paper offers a 2-part qualitative answer: Early recovery: consumption growth was held back by financial frictions (wealth and leverage effects). More recently: consumption growth has been held back by low consumer confidence (low expected income growth) and heightened uncertainty and perhaps higher income inequality. My view: The paper provides a useful tour of the candidate explanations and offers many tantalizing clues about the possible roles of different factors, but the evidence is far from conclusive. U.S. Treasury, Office of Economic Policy 4
Early Recovery: Agree that the lagged effects of the drop in wealth were important, but how big of an independent role did leverage/debt play? U.S. Treasury, Office of Economic Policy 5
Evidence from macro models: not clear that you need to appeal to factors beyond the weakness in the usual RHS variables (income, wealth) Simulations from Macro PCE Model Consumer Spending Growth 4-quarter real percent change 6 4 Model Personal Saving Rate Percent of disposable income, average in period 8 6 Model Actual 2 0 Actual 4-2 2-4 2007 2009 2011 2013 2015 0 2007 2008 2009-2010 2011-2012 2013 2014-2016 Traditional determinants alone appear to be able to explain the early recovery period. U.S. Treasury, Office of Economic Policy 6
Does regional data support the idea that leverage played a big role? Consumption growth was indeed weaker in states that had higher debt and leverage going into the crisis. But those states also saw larger home price declines and more job losses. So the regional correlation between leverage and consumption alone does not identify an independent role for leverage. U.S. Treasury, Office of Economic Policy 7
Does micro data support the idea that leverage played a big role? Evidence from this paper and Dynan (2012) supports the idea that leverage held back the consumption of some households. Median percent change in non-housing consumption, 2007-2009 0-5 -10-15 -20 Highly leveraged homeowners Other homeowners Source: Dynan (2012) based on PSID with sample restricted to housing boom states. Households with more mortgage leverage in 2007 saw a larger decline in consumption from 2007 to 2009. And the difference holds up even after controlling for income and wealth. U.S. Treasury, Office of Economic Policy 8
But the aggregate effect of leverage implied by micro data is fairly modest At the individual level, the Dynan (2012) estimates imply that an increase in a household s mortgage LTV from 1.0 to 1.1 would have reduced its consumption growth by ½ to ¾ percentage point between 2007 and 2009. Calibration of the aggregate effect based on these estimates + aggregate leverage data + information about share of households with mortgages: The debt overhang might have held back aggregate consumption growth by ¼ to ½ percentage point per year and presumably the effect waned as we entered the early recovery. U.S. Treasury, Office of Economic Policy 9
More recently: The paper s argument that people have revised down their expectations of future income growth and perceive income as more uncertain seems plausible. U.S. Treasury, Office of Economic Policy 10
The story is consistent with revisions to macro forecasts IMF World Real GDP Growth Forecast Percent change, year-over-year 5.5 4.5 Actual Growth Forecast Sep-2011 Forecast Oct-2012 Forecast Oct-13 Forecast Oct-14 Forecast Oct-15 3.5 Forecast Oct-16 2.5 2010 2012 2014 2016 2018 2020 U.S. Treasury, Office of Economic Policy 11
The paper shows that there is some evidence from surveys as well, but the data also raise questions Fraction expecting less comfortable retirement Average probability of job loss Have we really seen a lasting change? These measures seem to be recovering. Fraction who see >50% chance of a real income decline Standard deviation of permanent innovation in income process What do we make of the fact that these measures deteriorated right when job growth picked up? U.S. Treasury, Office of Economic Policy 12
More recently (continued): And, are financial frictions really behind us? U.S. Treasury, Office of Economic Policy 13
Don t agree with paper s conclusion that bank senior loan officers are saying credit conditions for mortgages and credit cards have recovered Paper integrates reported changes and takes out a linear trend Mortgage Loan Availability 150 Here s what you get if you don t take out linear trend Mortgage Loan Availability 50 100 50 0-50 -100-150 2002 2004 2006 2008 2010 2012 2014 2016 0-50 -100-150 -200-250 -300-350 2002 2004 2006 2008 2010 2012 2014 2016 Credit Card Availability 0.6 0.4 0.2 0-0.2-0.4-0.6-0.8 Credit Card Availability 1.2 1 0.8 0.6 0.4 0.2 0-0.2-0.4-0.6-1 2002 2004 2006 2008 2010 2012 2014 2016-0.8 2002 2004 2006 2008 2010 2012 2014 2016 U.S. Treasury, Office of Economic Policy 14
And there is other evidence that household credit remains tight for some people Standards remain at the tight end of the range for subprime borrowers at many banks. Mortgage Originations by Credit Score Billions of dollars 1,000 <620 620-659 660-719 720-779 780+ 800 600 400 Tighter than Average Lending Standards Relative to the Range since 2005 50 40 30 20 10 Tightest level Significantly tighter than midpoint Somewhat tighter than midpoint 200 0 '03 '04 '05 '06 '07 '08 '09 '10 '11 '12 '13 '14 '15 '16 0 Prime Borrowers Credit Cards Subprime Borrowers Prime Borrowers Auto Loans Subprime Borrowers Still almost no mortgages going to people with credit scores below 660. U.S. Treasury, Office of Economic Policy 15
More recently (continued): Perhaps relatedly, has the response of consumption to wealth increases declined? U.S. Treasury, Office of Economic Policy 16
There has been a considerable rebound in household wealth Household Wealth Relative to Disposable Personal Income Ratio 6.5 6.0 5.5 5.0 4.5 1990 1994 1998 2002 2006 2010 2014 U.S. Treasury, Office of Economic Policy 17
Wealth effects from the recovery in stock and home prices should have been considerable Back-of-the-Envelope Estimates of Wealth Effects Percentage point boost to real consumption growth 2 1.6 1.2 0.8 0.4 0 2013 2014 2015 Assumes a MPC of 0.035 and effects phased in over 3 years with 60 percent in the first year. U.S. Treasury, Office of Economic Policy 18
This graph from the paper suggests the MPC out of wealth has declined Change in MPCs Out of Housing Wealth Implied by State-level Regressions This fits with intuition: Greater difficulty extracting home equity to finance consumption. Wealth gains may seem smaller or less permanent post crisis. Wealth gains may be going to people with lower MPCs. U.S. Treasury, Office of Economic Policy 19
But the state-level analysis is problematic because we do not actually have state-level data on consumption BEA provides estimates: Annual estimates of consumption of goods and many services categories based on state-level retail wage data. Annual estimates of owner-occupied housing consumption based on value of owner-occupied housing stock, but that is also the key RHS variable. More work needed to figure out what the regressions are telling us; and work needed with other sources to confirm the result. U.S. Treasury, Office of Economic Policy 20
More recently (continued): What might be the role of growing income inequality? U.S. Treasury, Office of Economic Policy 21
Seems plausible that growing inequality is holding down consumption There has been a big increase in inequality over the last several decades. The rich have lower MPCs (at least when you are looking at our standard definition of consumption). Alichi et al estimates that growing inequality might have lowered consumption by 3.5 percent from 1998 to 2013. But it s a longer-term story Alichi estimates translate into a reduction in consumption growth of 0.1 to 0.2 percentage point per year. And it s not clear how much inequality has risen in recent years. U.S. Treasury, Office of Economic Policy 22
Conclusion U.S. Treasury, Office of Economic Policy 23
This paper makes impressive use of micro data to explore a number of theories about what has been driving consumption growth in recent years. The creative use of data and the questions raised by the results will spawn many further papers on these issues. But it only represents a first step: Much more work to be done to definitively identify and quantify the factors that have shaped the trajectory of consumption during the recovery: Uncertainty / income expectations Reduced wealth effects Credit constraints U.S. Treasury, Office of Economic Policy 24
Endnotes Slide 6: Source - Model similar to Davis and Palumbo (2001). A Primer on the Economics and Time Series Econometrics of Wealth Effects. FEDS Working Paper. Slides 8 and 9: Source Dynan (2012). Is a Household Debt Overhang Holding Back Consumption? Brookings Papers on Economic Activity. Highly leveraged households defined as those in the top quintile of the mortgage leverage distribution. Graph is based on analysis of data from the Panel Study of Income Dynamics, with the sample restricted to states that experienced a housing boom. Slide 11: Source International Monetary Fund World Economic Outlook. Slide 12: Source Pistaferri (2016). Slide 14: Source Federal Reserve Board, Senior Loan Officer Opinion Survey. Slide 15: Source Federal Reserve Bank of New York Consumer Credit Panel; Federal Reserve Board, July 2016 Senior Loan Officer Opinion Survey. Slide 17: Source Financial Accounts of the United States. Slide 18: Source Calculations based on data from the National Income and Product Accounts and the Financial Accounts of the United States. Estimates correspond to a Q4/Q4 percent change. Slide 19: Source Pistaferri (2016). Slide 22: Alichi, Kantenga, and Sole (2016). Income Polarization in the United States. IMF Working Paper WP/16/121. U.S. Treasury, Office of Economic Policy 25