Three most important questions

  • What is the true infection death rate?
    • If <0.1%, have overreacted.
    • If 1%, have reacted correctly. Likely see future shutdowns.
  • What is the endgame for the lockdowns and restart?
    • Show restarts unlikely before May.
    • Simulate % of time outside of lockdown given \(R0\) outside of lockdown
  • What is the expected fiscal cost?
    • Deficits of 15% of GDP
    • 2009-2012 deficits were financed by int'l portfolio flows
    • Not a deflationary trap like 2008. Different to QE.
    • 10Y BEIs underpriced at 0.95%. FV over 2%.

Disagreement large

  • Optimistic case (<0.1%), WSJ article, Dr Bendavid and Bhattacharya (Profs Medicine Stanford).
  • Relies primarily on DeCode Genetics (Iceland) and NBA example.

  • Both logics flawed.

  • Enough evidence to bound it between 0.25% and 1.25% under unconstrained medical resrouces.
  • Case death rates uninformative.

  • Deaths likely undercounted by factor of 1-2 times, cases undercounted by 2-10 times depending on country.

Determining true infection death rate

  • Case fatality rate purely measurement driven, infection death rate key
  • Due to lag from cases to deaths, increases as situation is actually improving
  • Reporting standards (e.g. Germany)

Determining true infection death rate

  • \(R^2\) between cases per capita and tests per capita is 74%
  • Deaths first to be tested (unlikely to be missed), mild cases missed.
  • Differences in death rates well explained by testing capacity.

Case fatality rate: Iceland experiment

  • Key is to determine underreporting of tests. Expert survey in US: median underreporting estimate of 8x. Positive test rate is circa 15%.
  • Iceland, 3% of population symptomatic and tested, deCode tested 2% of pop for antibodies:
  • Mar 13-21: 5600 nonsymptomatic tested for antibodies, 48 positive samples (1%).
  • Mar 13-21: 9600 symptomatic tested officially for virus, 473 positive samples (5%).
  • Mar 26: 2 deaths.
  • deCode implies 3100 asymptomatic cases nationally had virus before Mar 21, were never captured.
  • Using confirmed cases: 2/500, death rate 0.4%.
  • Using addt'l implied 3100: 2/3600, death rate <0.1%.

Case fatality rate: SK / AUS

  • To minimize counting problems only use countries with highest test positive rate (lowest undercount of cases).
  • In Aus and SK, % of symptomatic tests positive only 2% (in worst precincts of NY, Italy, 50%).

Case fatality rate: SK/ AUS case studies

  • Assume no symptomatic cases missed. Add asymptomatic to denominator.
  • Using 50% asymptomatic rate on Diamond Princess (Russell).
  • Method: Assume zero symptomatic cases are missed, but two thirds asymptomatic.
  • (Estimate from Aus): 1.8% x 1/3 = .6%
  • (Estimate from SK): 2.2% x 1/3 = .7%

Robust method: residual death

  • Deaths likely undercounted everywhere. No robust home death reporting.
  • Take Italian counties with highly granular + timely reporting.
  • Residual deaths v model small variance pre Feb 21, then spike.

Robust method: residual death

  • By age, cross-section of residual deaths match relative risks in med studies
  • Standard error bars incorporate annual variance and 30% to 80% infection rate

Robust method: residual death

  • Aggregate shares to population age distribution
  • Assuming 50% population infected, IFR of .75%.
  • Distribution of outcomes given counterfactual infection rate amongst early towns.

Implication: how far are we through this?

  • Implied percent who have contracted, IFR of 0.75%

Post-lockdown scenarios: 1-2months

  • Rapid decline in transmissions after lockdowns began.
  • Many countries now with \(R0 < 1\)
  • Likely stagger reopenings once \(R0\) stable below 0.5.

Post-lockdown scenarios: after reopening

  • Antibody testing also has low benefit while 95% uninfected.
    • Best US drug has specificity rate of 94%.
    • Specificity: \(P(T-|D-) = TN / (TN + FP)\)
    • I.e. suppose prior is 5% infected, 95% not infected.
  • Antibody test will return 1 false immunities for each 1 true immunities.
  • Helpful to estimate national % of immune, not individual cases.
  • "Clean sheet" strategy not tenable

Post-lockdown scenarios: after reop

  • R0 too high for herd immunity
  • 0.1% of the NY population has died (approx), implies 10% have had it.

Post-lockdown scenarios: after reop

  • Validated by random test of pregnant women in NY
  • NEJM published random sample of NY women admitted for delivery (April 13)

What will be the cost?

  • Percent of time in Q2/Q3/Q4 outside of lockdown depends on \(R0\) out of lockdown
  • \(R0\) in US states that have behavioral response but no lockdown around 1.15-1.35. Sweden has 1.2

R0 in countries without lockdown

  • Sweden in partial shutdown, \(R0\) of 1.3

Post-lockdown scenarios: after reopening

  • What's the strategy after reopening?
  • Optimal scenario steady state: keep new infections at exactly ventilator / ICU capacity until hitting 40%.
  • What would this look like?
  • Even in US, impossible.

Post-lockdown scenarios: after reopening

  • US has 35 icu beds per 100k persons. 50 ventilators per 100k. (170k total).

Post-lockdown scenarios: after reopening

  • New estimates of up to 10% of confirmed cases need intensive care (CDC).

Post-lockdown scenarios: after reopening

  • Optimistic scenario:
  • Peaked at 35k new cases a day. 10% were going into ICU. 2k deaths/day.
  • Even then, assume undercount means 200k getting immune each day.
  • If critical cases ave ~2 weeks of care, can sustain 10k new critical cases per day = (150k ICU beds / 14 day ave use).
  • To get to 40% of population immune, would be 20 months of deaths running at > 2k/day.

What will be the cost?

  • Spanish flu measures for top 17 economic cities. (1-6mo)

What will be the cost?

  • 41% of jobs (52% of national wages) can be performed from home, likely without wage cuts.
  • Assuming replacement rate at 30% for remainder, GDP would be at a steady state 34% below trend under total lockdown

Expected fiscal costs

  • Current announcements April 29th (excludes purchases of securities in secondary market)
  • Still woefully inadequate. All countries likely to have expense alone topping 15% of GDP.

Expected fiscal costs

  • UK budget deficit the only current realistic projection
  • Expect US and other major deficits to be 15% of GDP. New stimulus will come.
  • US Fed debt to GDP will surpass WWII peak.

Market far too sanguine on long term inflation (after virus)

  • Market too sanguine on debt deflation necessity

Market convinced this time is like 2008

  • 2008 did not require a deficit that couldn't be financed by private market (was international smoothing)
  • Larger the QE, the less the consolidated debt of government, the larger inflation must be