Yesterday, when commenting on the impact of Hurricanes Harvey and Irma, we noted that even before the two devastating storms were set to punish Texas, Florida and the broader economy, erasing at least 0.4% GDP from Q3 GDP according to BofA and costing hundreds of billions in damages (contrary to the best broken window fallacy, the lost invested capital more than offsets the "flow" benefits from new spending, which is why the US does not bomb itself every time there is a recession to "stimulate growth"), things were turning south for the US economy, which in turn prompted Deutsche Bank to point out that (adjusted) recession risk, at roughly 20%, is now the highest in the past decade, and that it was quite prudent for the Fed, which expects to hike rates at least once more in 2017, to pause its current tightening, especially since a period of both economic and market weakness is imminent.
It didn't take long for one of the most bullish on the US economy banks to follow in BofA's footsteps, and overnight in a note from Goldman's chief economist, Jan Hatzius, announced the he was slashing his Q3 GDP estimate by a whopping 30%, or 0.8%, to 2.0% annualized, to wit:
Given the potentially sizeable growth effects from Harvey—and with Irma risks now moving to center stage—we lowered our Q3 GDP tracking estimate by 0.8pp to +2.0%...
But fear not, because like all good Keynesian acolytes of the "broken window fallacy", Goldman is confident that the flawed perpetual engine of growth, namely destruction - after all, why else is the world's gearing for global war - will kick in, and more than offset the Q3 GDP loss, by boosting the next 3 quarters by a cumulative 1.1%:
... However, we expect this weakness to reverse over the subsequent three quarters, more than recouping the lost output. Accordingly, we are also increasing our respective quarterly growth forecasts by 0.4pp, 0.2pp and 0.4pp for Q4, Q1, and Q2, (to +2.7%, +2.5%, and +2.4%). We will revisit these estimates once reliable information about the toll from Irma becomes available. We stress that the overall impact of the hurricane on second-half growth is uncertain, as the negative effects are likely to be offset by an increase in business investment and construction activity once the storms have passed.
Some additional detail on what Goldman expects will happen in the coming months to the US economy:
We find that major natural disasters are associated with a temporary slowdown in most major growth indicators. We also find that costly and broad-based natural disasters are associated with particularly large declines in economic activity, but also sharper subsequent rebounds. Modeling these effects, we estimate that hurricane-related disruptions could reduce 3Q GDP growth by as much as 1 percentage point. We believe the main channels for these GDP effects are consumption, inventories, housing, and the energy sector.
We expect a meaningful drag on key growth indicators over the next two months (detailed herein), including a temporary drag on September payrolls growth of 20k—or as much as 100k if severe storm effects persist into next week (the payrolls reference period). We also expect a near-term boost to headline inflation (around 0.2pp on the yoy rate) due to higher gasoline prices, and a possible modest boost to core inflation (worth less than 0.05pp), due to the destruction of some of the automotive capital stock.
And here is how Goldman justifies the sharp economic surge that follows natural disasters:
In Exhibit 5, we summarize the historical growth experience around these 43 events using the “monthly GDP growth” measure developed by Stock and Watson (discussed here; we use a similar series from Macroeconomic Advisers after 2010. This measure interpolates the official quarterly GDP data using the same monthly source data used to construct it. In addition to the average evolution of monthly GDP across the 43 disasters, the graph shows averages across the top 10 costliest, longest, and most broad-based disasters (based on the earlier measures). On average, costly and broad-based natural disasters produce particularly large declines in economic activity, but also sharper subsequent rebounds. We find that long-lived disasters are not particularly notable (relative to the sample average), but note their subsequent growth rebounds are sometimes more muted.
Economic Data Are Particularly Sensitive to Costly Natural Disasters and Those that Affect a Large Share of the Population
The best news is that the above estimates only take Harvey into account: once the damage from Irma is added, the rebound will be even greater, potentially unleashing another Golden Age for the US economy... or something. Which of course, once again begs the rhetorical question: if the US is in dire need of growth, why not just nuke itself and end up with even more economic growth than prior to said nuking (please don't answer, as we have said previously, "contrary to the best broken window fallacy, the lost invested capital more than offsets the "flow" benefits from new spending, which is why the US does not bomb itself every time there is a recession to "stimulate growth", unfortunately few "economists" can grasp this simple logic.)
Sarcasm aside, here are some further observations from Goldman and its "handbook" attempt to quantify the Harvey damage:
We conclude with a “handbook” quantifying the potential impact of Hurricane Harvey on upcoming US economic data. We base these estimates on regression models as well as the average evolution of these indicators around past national disasters, with special emphasis given to Katrina observations. Exhibit 10 summarizes these potential impacts, including their “bottom-up” implications for Q3 and Q4 GDP growth (at-0.8pp and +1.1pp, respectively). The main channels for these effects are likely to be consumption, inventories, housing, and the energy sector, where we previously estimated that declining petroleum refining and energy output could by itself depress Q3 growth by as much as 0.2pp (these estimates remain valid, given continued outages in many areas).
Hurricane Handbook: Harvey Could Shift the Composition of Growth from 3Q into 4Q/1Q, Particularly if Irma fears are Realized
* * *
For those who wish to ignore Goldman's commentary and merely focus on its data interpolations and forecasts, here are some additional observations and, more importantly, charts. First, on Harvey's estimated damage:
Hurricane Harvey hit the Gulf Coast region of the United States on Friday, August 25, resulting in heavy rains, widespread flooding, and significant property damage. As shown in Exhibit 1, many estimates of Harvey damages rose sharply during the first week after landfall; however, most have now settled in the $70-100bn range (we assume $85bn in our analysis). The uncertainty around these figures remains high, but it seems clear that Harvey’s aftermath will be particularly severe.
Exhibit 1: Harvey Damage Estimates Rose Sharply and Have Settled in the $60-100bn Range
Next, on Harvey's damage in historical context: "Costly, Widespread, and Potentially Long-Lasting"
To place Harvey in historical context, we revisit some of our previous studies in order to construct a dataset of comparable natural disasters. We include the 35 largest hurricanes in the Billion-Dollar Weather Disasters dataset from the National Oceanic and Atmospheric Administration (NOAA). We then add major earthquakes, floods, and tornado events whose cost exceeded 0.05% of GDP (an additional 8 observations).
As shown in Exhibit 2, Harvey’s approximately $85bn in damages would represent 0.44% of GDP, which would make it the 2nd largest natural disaster since World War II in terms of domestic property damage. Hurricane-related losses also seem likely to rise further in coming weeks given possible damages associated with Hurricane Irma in Florida and other parts of the Southeast. Some estimates of insured losses for that storm range from $60bn to $180bn (implying total losses could exceed $100-200bn). Given that estimates for Irma remain tentative and particularly unreliable at this stage, we focus our attention on what we know about Harvey instead.
While widely cited, property losses are only one dimension of a disaster’s impact on the economy (and on society more generally). To complement this metric, we construct two complementary measures of severity, specifically, the duration and the societal breadth of a given storm’s disruptions.
* * *
For our measure of breadth, we calculate the share of the US population in counties with disaster declarations (directly related to the given storm event). As shown in Exhibit 3, Hurricane Harvey severely impacted nearly 5% of the US population based on this classification. This is a bit above the median (17th out of 44), just behind Katrina (16th).
For our measure of duration, we use the length of the federal “natural disaster declaration period” for each disaster, as shown in the left panel of Exhibit 4. Harvey made landfall only two weeks ago, so we cannot yet estimate the duration of the storm’s impact on this basis. But given the extent of the flooding and the continued weakness in electricity consumption (see right panel of Exhibit 4), we believe it is reasonable to expect Harvey’s duration could be longer than average (the median national disaster declaration in our dataset is 27 days). Electricity supplied also fell sharply in Louisiana after Hurricane Katrina struck that state in August 2005, with DOE data showing that electricity sales declined on a year-over-year basis for six consecutive months. If Harvey’s natural disaster declaration continues through the end of October, its duration would move to 6th out of 44 disasters (Katrina is 5th).
We find that temporary soft patches tend to be common for “hard” indicators such as nonfarm payrolls, retail sales, industrial production, and housing starts, as well as for soft indicators like the ISM Manufacturing Index and Conference Board Consumer Confidence. For payrolls, we find an average deceleration of 23k relative to recent averages but a wide range (that likely depends on the timing of the payrolls reference period). The trade balance also tends to widen in these episodes, as export growth tends to slow more quickly than import growth (and rebound more slowly).
One perhaps surprising aspect of the above relationships is the muted rebound in housing and in capital equipment in the months following major disasters (on average). We suspect this reflects the long lags typical of the rebuilding process. The experience following Katrina illustrates this point: it took 7 months for New Orleans residential building permits to return to their pre-Hurricane levels and another few quarters before they were materially higher. And while Congress has already allocated an additional $15 billion of hurricane relief funds, it’s important to remember that federal emergency spending on construction tends to ramp up gradually over years as opposed to months (see right panel).
Lastly, in terms of the inflation impact around natural disasters, we do not find a compelling historical pattern for either headline or core inflation (neither PCE nor CPI). In the case of Harvey in particular, we nonetheless expect a near-term boost to headline inflation from higher gasoline prices, themselves a result of disruptions to petroleum refining and the energy sector more broadly. We also note the possibility that the destruction of some of the automotive capital stock—our autos team estimates as many as 1.1mn cars destroyed by Harvey—could reduce downward pressure on new and used car prices. This and potential energy-price pass through (i.e. to airfares) could provide a modest boost to core inflation.
Putting it all together, here is Goldman's summary assessment of why a Hurricane may be precisely what the Keynesian Doctor ordered:
In Exhibit 8, we attempt to estimate the impact of natural disasters on monthly GDP growth. Given the relatively small sample size (43 major disasters) and the lack of geographic granularity for nearly all growth indicators, the timing and the magnitude of disaster effects are difficult to estimate empirically. The fact that we have three different measures of storm severity increases this difficulty.
We restrict the regression sample to periods around natural disasters and find statistically significant and economically meaningful storm effects. Our results are generally consistent with the message from Exhibit 5. Specifically, the share of the population affected and the amount of property losses are associated with larger declines in output in the month of and the month after the storm. We also find that longer-lived disasters may be associated with more muted growth rebounds, whereas broad-based storms are associated with fairly rapid rebounds.
Costly and Broad-Based Disasters Are Associated with Sharper GDP Growth Decelerations (but Also Sharper Rebounds)
In terms of the growth impact from Harvey, this “top-down” model would suggest a sharp drag on 3Q GDP growth of as much as 1.4 percentage points (qoq ar, relative to baseline)—reflecting Harvey’s huge property losses and the relatively broad-based societal footprint. Importantly though, the model would also suggest a rebound commencing in 4Q (+0.5pp impact) that we believe would likely continue into 1H18. As shown in Exhibit 9, higher-frequency real activity data in regions and sectors levered to the Houston area has already weakened considerably.
And the best news of all from this analysis on Harvey: just wait until the "economic boost" from the Irma devastation That should really unleash America's growth potential...