Housing Is The Business Cycle (Leamer, 2007)
Not a book, but figured I’d publish my notes anyway -- I wanted to type this all out to really make it stick in my head, because I found it fascinating. Original paper is available here and worth a close read. Takeaways are pretty simple: the cycle is about consumer spend rather than business spend, housing is the best leading indicator, and housing is influenceable via monetary policy. (Recommended by: Elliot Turner)
Housing is an obvious way that monetary policy affects the economy but is not studied or even acknowledged in most macroeconomic literature.
Advocating for a modified Taylor Rule that depends on a LT measure of inflation but uses housing starts and change in housing starts in place of the output gap. Starts are the best leading indicator he’s aware of. Doing this would create preemptive anti-inflation in the middle of expansions making anti-inflation policy less necessary at the ends of cycles. Idea is to make recessions less frequent & less severe.
This is empirical data -- temporal, not causal since it can’t be measured in experiments.
“Economics... is a self-consciously interventionist discipline. We think we are designing the best way for our governments to influence the outcomes.”
Of 10 recessions since WWII, 8 have been “consumer recessions” and 2 have not been - namely the “DoD downturn” in 1953 (end of Korean War led to massive contraction of defense spend) and “Internet comeuppance” of 2000-2001 driven by collapse in biz investment (software/equipment) as Internet profitability disappointed.
Sectors with most volatile employment are construction and manufacturing. If you exclude these from total employment, employment flattens out but doesn’t decline perceptibly in recessions. Manufacturing always recovered in a “V” out of a recession except in 1990 when it was a “U” and after 2001 it didn’t recover (”L”).
From 1970 onward, economy has grown at a surprisingly steady 3% real rate, almost always within a +/-3% band (”3-3 rule”). Shows policy has not had a demonstrable effect on LT growth. Shows us policy should be focused on "ironing out” the cycle and keeping real GDP growth within the corridor.
Resi investment is a very small part of LT growth, like 4% or 13 bps of 3.10% -- consumer services leads, followed by nondurables and durables. Equipment & software has become more important, and inventories/structures hardly contribute at all. Interestingly, inventory contributes volatility but less volatility in inventories has been a big contributor to increased stability of GDP growth post-1984. Hard to know if that’s improved inventory management or greater stability of sales. Volatility of every component declines after 1985 -- esp resi investment, durables and nondurables, which are all inventory-intensive.
Leamer extracts abnormal contribution of each component relative to its smoothed normal contribution (complicated process but good methodology given) and measures that around recessions, setting value at cycle peak = 0.
Because these lines show cumulative abnormal contribution to GDP growth:
flat line = contribution each Q is normal
declining line = contribution less than normal, contributing to weakness
rising line = contribution greater than normal, contributing to strength
Resi investment subtracts from GDP growth before recessions but contributes > normal soon after recession starts (2nd or 3rd quarter). Equipment & software is very different, less and less consistent negative contribution prior to recessions (though seemingly growing over time) and a clear contribution to weakness *after* recession has begun. Biz capex clearly contributed more than normal to weakness before and during recession in 2001 because of its abnormal nature (truly a biz recession). As a rule: weakness in housing precedes recessions; weakness in equipment/software is coincident. The cycle is a consumer cycle, not a business cycle.
On average, housing contributes 20-25% of the weakness in growth prior to a recession and is the biggest and most consistent contributor. Only two recessions where it wasn’t a significant contributor were, again, 1953 & 2001
Equipment & software is the largest contributor of weakness during a recession, averaging ~20% of the weakness contributed. It’s only the 6th-largest when measuring prior to recessions.
Timing the path of a recession goes: homes, durables, nondurables, services. Average paths of biggest components are shown in Figs 8 & 9 (consumer recessions only)
Significant portion of weakness in demand is exported via weaker imports, which makes imports’ abnormal contribution positive during recessions (as opposed to its normal drag on GDP). This is offset by substantial negative abnormal contribution of exports -- possibly because foreign GDP is weaker due to lower US imports.
When has a recession happened without housing predicting it (ie housing has given a false negative signal)? The years already covered, 1953 & 2001 (though ‘01 had some weakening in home values).
When has housing predicted a recession but it hasn’t occurred (ie housing has given a false positive signal)? 1951-2 and 1966-7. Lack of recession attributable to ramp of defense spending on Korean War and Vietnam War, respectively.
If we include defense spending’s abnormal contribution, we get 90% of the cycle story of the last 60 years.
Uses multivariate approach to proving this out, doing a regression of GDP growth vs. one-quarter-lagged contribution. Here, the t-stats measure independent contribution of each GDP component, controlling for all others. Top 4 contributions are resi investment, consumer services, consumer nondurables and consumer durables.
Coefficient on resi investment implied by this exercise is 2.0, meaning an abnormal contribution to GDP growth from housing implies twice as much contribution the next quarter. Consumer services at 1.7 is the only other component with coefficient >1, implying a multiplicative effect. Coefficient on consumer durables is negative -- controlling for strength in the economy (as this exercise does), an abnormal surge in sales of durables in one quarter steals from the next.
Housing is a fairly unique asset in that prices are inflexible downwards, which means volumes are crushed when in a downturn because bid-ask widens. The impact on GDP is from volume, not price (simple appreciation does not impact “production”, which is what GDP measures) -- especially when you consider jobs in construction, finance, real estate.
Another way to think about price not moving GDP is that any appreciation in land value that’s booked as an asset for homeowners needs to be booked with an exactly offsetting liability for future homebuyers
Price weakness in housing makes the effective interesting rate high even at low nominal interest rates --> it’s highly leveraged and thus small price declines can kill off building. Makes it hard for stimulative rate cut to move the needle once volumes are declining.
Normal sales volumes in houses happen when buyers are confident that home prices should increase at normal rates or at least not decline. Otherwise, buyer waits to get a better deal. If prices were able to recalibrate quickly when the cycle turned down, normal prospective appreciation and thus normal sales volumes could reappear quickly. But sluggish downward price adjustments leads to a more extreme volume cycle. This is the core of *why* housing is so important to recessions. There are empirical examples like in LA in the 90s.
Builders are more motivate sellers than homeowners (who can just decide to stay put), but this is also observe in new home sales.
Argues that new home volumes best observed via +/-20% bands vs. prices which normal fit within +/-10% bands when considered on a real basis.
Basically, homeowners who won’t sell into a weak market sit on a house while the price erodes on a real basis due to general inflation. So while nominal price stays steady, house becomes cheaper with time. This is to say that price cycle is even less pronounced in nominal terms than in real terms and supports argument that it’s a volume cycle.
Trend in real price by geography shows (negative) correlation between new building (LT trend of new home sales growth) and real appreciation rate. E.g., South (lots of building) real appreciation is +1.1%, Northeast (little building) +2.0%. Where buildable land is plenty, response to increased demand is more building, but where there is little, the response is price increase which discourages buyers and reequilibrates supply & demand.
Several reasons for seller unwillingness to adjust prices down.
1) Ego -- we love our houses, don’t want to sell our loved ones for less than we think they’re worth (different from a stock), especially not if we’re anchoring off what a neighbor got for their house
2) Sellers look backward (what they paid), buyers look forward (what house might be worth in near future). In a rising market, this causes bid-ask to cross (buyer sees higher future value and seller sees lower past value) and a transaction occurs, but in a falling market it’s the opposite, bid-ask spread remains wide and no transaction occurs
3) Aversion to crystallizing the loss
Makes an argument for why we need to attenuate the cycle anyway -- unemployment and foreclosure does damage to working people’s ability to budget for their lifetime, and unemployment and foreclosure disproportionately impacts the poor and the young.
Hypothesizes that financial cycle is experienced disproportionately by low-income first-time homebuyers and that idiosyncratic risk is greatest for homes bought at the top of the cycle, usually by those same buyers. Finds greatest appreciation in LA from 2003-2005 was in low-priced zip codes and for smaller homes. Meanwhile, idiosyncratic risk hypothesis was not right -- buying/selling acumen explains the idio risk, which is surprisingly large with a stdev of 30%
Paul McCully (2007) applied Hyman Minsky’s 1986 theory of financial cycle to housing. Three types of loans
hedge finance -- supports acquisition of assets with current profits sufficient to cover interest and amortization
Speculative finance -- supports acquisition of assets with current profits sufficient to cover only interest and relies on future appreciation or income growth to pay down the debt
Ponzi finance -- current profits won’t even cover interest.
Through the cycle, progress through these as lending standards ease and tighten
Studies price appreciation between homes priced in the 10th percentile and 90th percentile. 10th percentile had higher appreciation when underwriting standards were relaxed (2004-2005). This differs from 1988-1989 when 90th percentile homes saw the most appreciation.
early warning signs of recession are weakness in homes and consumer durables;
most of the job loss in US recessions comes in construction and durables manufacturing.
Special features of durable manufacturing and residential contstruction
Previous product (new homes, cars) creates a stock of existing assets that compete with current production -- long periods of unsustainably high levels of sales/production that increases stock beyond equilibrium levels gives rise to long periods of low sales/production to reequilibrate. Conversely, recessions create pent-up demand that is met by high levels of sales/production after recessions. I.e., it’s cyclical.
Services flow from existing stock is very elastic and ability to postpone acquisition of new car or home is great. I.e., you can always keep driving it or not move for another year or two
Price of durability = real rate of interest. When real interest is low, equilibrium stock of homes/cars is high. If this is a permanent shift in real interest, fine. If it’s impermanent, creates problems due to longevity of the assets. I.e., equilibrium demand is rate-sensitive.
Asset prices of homes/new cars suffer from downward rigidity. This causes production to stay low even if rental market is strong. Deflation in asset prices of existing durable stocks in the fact of strong rental markets is a real problem.
Due to longevity, creates an intertemporal control problem -- i.e. stimulus today steals from the future. In 2001 recession, sales of homes and durable held up well. There were no lost-sales to transfer forward in time, so Fed easing in 2002-2004 instead transferred sales backward in time (i.e. stole from the future, 2006-2009). That’s why cutting rates won’t help.
Uses correlograms (moving correlation) to study inflation vs. housing:
Inflation is persistent -- correlation of CPI inflation with CPI inflation prior month is 0.63. After 4 years, it’s still 0.2. Once it gets going it’s hard to stop, this is why monetary policy takes it so seriously.
For housing, it’s the cycle that’s persistent -- correlation of housing starts to housing starts prior month is 0.93. That goes to 0 after 24 months, and is significantly negative (-0.2) within 36 months.
Of all the GDP components, resi investment has the largest correlogram (i.e. it’s the most cyclical). Other 3 that are significant are defense, business structures and business equipment & software.
Defense has a long string of negatives at 8-9 years, implying buildups are followed by cutbacks or vice-versa.
Studies where and how housing conflicts with explicit Fed targets of inflation and unemployment (e.g. weak housing calling for a rate cut while high inflation is calling for a hike). Conflict between housing and UE are not great, conflict with inflation is more significant.
Regresses a bunch of indicators to see which most explains changes in Fed funds rate --> 10Y Tsys win (on t-stat and coefficient). Housing starts is last, implying it’s least-watched by Fed.
Studies whether historical policy has amplified or attenuated housing cycle, even though this is not what Fed was targeting -- does this across a bunch of cycle periods with interesting results. He uses slope of yield curve (10s - FFR)/10s to measure monetary tightness. Goes through predictive (market signal) and causal (credit crunch) narratives for yield curve inversion.
Not much info available on Great Depression, but housing starts turned down in 1925, 3+ years ahead of industrial production (July’29) and DJIA peak (Oct’29). In total, starts went from 900k at peak to <100k at 1933 trough.
There is potentially a relationship between durables and housing -- they are complementary (you buy dishwashers, furniture, etc to fill your house), housing wealth may help finance durables spending and the same interest rates may drive both cycles. However, most of the amplitude in durables comes from autos, not furniture, which hurts the complementarity argument. Given that prices adjust slowly and thus the wealth effect seems like a weak argument, he argues it’s interest rates & employment that are the common drivers between the housing & durables cycles.
Negative wealth effect (price-driven) is different from recessions (volume-driven) -- it’s an argument for sluggish growth, not contraction.
Conclusion: Housing is a small contributor to normal economic growth, but while unimportant in normal periods, it’s critical in US recessions: the first thing to soften and the first to turn back up.
His personal monetary to-do list -- he believe Fed watching housing can contribute to all:
Smooth business cycle -- attenuate collective unwanted idleness of recessions. We are better off if recessions are less frequent and less severe.
Keep us working productively -- limit speculative bubbles that absorb labor time and divert savings into low-yielding investments.
Limit re-distribution of wealth caused by financial disruptions
Keep consumer balance sheets accurately reflecting reality -- improves ability to plan for retirement realistically which informs how and how much we work. “We want our measured asset values to increase when our investments and discoveries make us confident that future GDP will be greater than we had originally thought. We do not want a monetary system that allows us to put phantom assets on our balance sheets and that signals to us that hard work and savings are not needed to prepare for our retirements.”
