Developed Countries

Special Report

Work from home policies, originally designed as emergency measures in the early phase of the COVID-19 pandemic, are likely to be “sticky” in a post-pandemic world. This will negatively impact the labor market in central business districts, via reduced spending on services by office workers. The potential impact of working from home is often cited as an example of what is likely to be a lasting and negative effect on jobs growth, but we find that it is not likely to be a barrier to the labor market returning to the Fed’s assessment of “maximum employment.” The size of the impact depends importantly on whether employee preferences or employer plans for WFH prevail, but our sense is that the latter is more likely. A weaker pace of structures investment in response to elevated office vacancy rates will likely have an even smaller impact on growth than the effect of reduced central business district services employment. The contribution to growth from structures investment has been small over the past few decades, office building construction is a small portion of overall nonresidential structures, and there are compelling arguments that the net stock of office structures will stay flat, rather than decline. Our analysis suggests that job growth over the coming year could be even stronger than the Fed and investors expect, possibly resulting in a first rate hike by the middle of next year. This would be earlier than we currently anticipate, but it underscores that fixed-income investors should remain short duration on a 6-12 month time horizon, and that equity investors should favor value over growth positions beyond the coming 3-4 months. The outlook for US monetary policy over the next 12 to 18 months depends almost entirely on the outlook for employment. Many investors are focused on the potential for elevated inflation to force the Fed to raise interest rates earlier than it currently anticipates, but it is the progress in returning to “maximum employment” that will determine the timing of the first Fed rate hike – and potentially the speed at which interest rates rise once policy begins to tighten. In this report, we estimate the extent to which the “stickiness” of working from home (WFH) policies and practices could leave a lasting negative impact on the US labor market. We noted in last month's report that a large portion of the employment gap relative to pre-pandemic levels can be traced to the leisure & hospitality and professional and business services industries, both of which – along with retail employment – stand to be permanently impaired if the office worker footprint is much lower in a post-COVID world.1 Using employee surveys and a Monte Carlo approach, we present a range of estimates for the permanent impact of WFH policies on the unemployment rate, and separately examine the potential for lower construction of office properties to weigh on growth. We find that the impact of reduced office building construction is likely to be minimal, and that WFH policies may structurally raise the unemployment rate by 0.3 to 0.4%. While non-trivial, when compared with a pre-pandemic unemployment rate of 3.5%, WFH policies alone are not likely to cause a long-term deviation from the Fed’s maximum employment objective. Relative to the Fed’s expectations of a strong, lasting impact on the labor market from the pandemic, this suggests that job growth over the coming year could be even stronger than the Fed and investors expect, possibly resulting in a first rate hike by the middle of next year. This would be earlier than we currently anticipate, but it underscores that fixed-income investors should remain short duration on a 6-12 month time horizon, and that equity investors should favor value over growth positions beyond the coming 3-4 months (a period that may see outperformance of the latter). Quantifying The Labor Market Impact Of The New Normal For Work In a January paper, Barrero, Bloom, and Davis (“BBD”) presented evidence arguing why working from home will “stick.” The authors surveyed 22,500 working-age Americans across several survey “waves” between May and December 2020, and asked about both their preferences and their employer’s plans about working from home after the pandemic. Chart II-1 highlights that the desired amount of paid work from home days (among workers who can work from home) reported by the survey respondents is to approximately 55% of a work week, suggesting that a dramatic reduction in office presence would likely occur if post-pandemic WFH policies were set fully in accordance with worker preferences. Chart II-1Employee Preferences Imply A Dramatic Reduction In Post-COVID Office Presence However, Table II-1 highlights that employer plans for work from home policies are meaningfully different than those of employees. The table highlights that employers plan for employees to work from home for roughly 22% of paid days post-pandemic, which essentially translates to one day per week on average.2 BBD noted that CEOs and managers have cited the need to support innovation, employee motivation, and company culture as reasons for employees’ physical presence. Managers believe physical interactions are important for these reasons, but employees need only be on premises for about three to four days a week to achieve this. Table II-1 also shows that employers plan to allow higher-income employees more flexibility in terms of working from home, and less flexibility to employees whose earnings are between $20-50k per year. Table II-1Employer Plans, However, Imply Less Working From Home Than Employees Prefer Based on the survey results, BBD forecast that expenditure in major cities such as Manhattan and San Francisco will fall on the order of 5 to 10%. In order to understand the national labor market impact of work from home policies and what implications this may have on monetary policy, we scale up BBD’s calculations using a Monte Carlo approach that incorporates estimate ranges for several factors: The percent of paid days now working from home for office workers The amount of money spent per week by office workers in central business districts (“CBDs”) The number of total jobs in CBDs The percent of CBD jobs in industries likely to be negatively impacted by reduced office worker expenditure The average weekly earnings of affected CBD workers The average share of business revenue not attributable to strictly variable expenses The percent of affected jobs likely to be recovered outside of CBDs Our approach is as follows. First, we calculate the likely reduction in nationwide CBD spending from reduced office worker presence by multiplying the likely percent of paid days now permanently working from home by the number of total jobs in CBDs and the average weekly spending of office workers. This figure is then increased due to the estimated acceleration in net move outs from principal urban centers in 2020 (Chart II-2); we assume a 5% savings rate and an average annual salary of $50k for these resident workers, and assume that all of their spending occurred within CBDs. We also assume that roughly 50% of jobs connected to this spending are recovered. Chart II-2Fewer Residents Will Also Lower Spending In Central Business Districts Then, we calculate the gross number of jobs lost in leisure & hospitality, retail trade, and other services by multiplying this estimate of lost spending by an estimate of non-variable costs as a share of revenue for affected industries, and dividing the result by average weekly earnings of affected employees. For affected CBD employees in the administrative and waste services industry, we simply assume that the share of jobs lost matches the percent of paid days now permanently working from home. Finally, we adjust the number of jobs lost by multiplying by 1 minus an assumed “recovery” rate, given that some of the reduction in spending in CBDs will simply be shifted to areas near remote workers’ residences. We assume a slightly lower recovery rate for lost jobs in the administrative and waste services industry. Table II-2 highlights the range of outcomes for each variable used in our simulation, and Charts II-3 and II-4 present the results. The charts highlight that the distribution of outcomes based on employer WFH intensions suggest high odds that nationwide job losses in CBDs due to reduced office worker presence will not exceed 400k. Based on average employee preferences, that number rises to roughly 800-900k. Table II-2The Factors Affecting Permanent Central Business District Job Losses Chart II-3The Probability Distribution Of CBD Jobs Lost… Chart II-4…Based On Our Monte Carlo Approach   This raises the question of whether employer plans or employee preferences for WFH arrangements will prevail. Our sense is that it will be closer to the former, given that we noted above that employer WFH plans are the least flexible for employees whose earnings are between $20-50k per year (who are presumably employees who have less ability to influence the policy of firms). Chart II-5 re-presents the projected job losses shown in Chart II-4 as a share of the February 2020 labor force, along with a probability-weighted path that assumes a 75% chance that employer WFH plans will prevail. The chart highlights that WFH arrangements would have the effect of raising the unemployment rate by approximately 0.35%. However, relative to a pre-pandemic starting point of 3.5%, this would raise the unemployment rate to a level that would still be within the Fed’s NAIRU estimates (Chart II-6). Therefore, the “stickiness” of WFH arrangements alone do not seem to be a barrier to the labor market returning to the Fed’s assessment of “maximum employment,” suggesting that the conditions for liftoff may be met earlier than currently anticipated by investors. Chart II-5CBD Job Losses Will Not Be Trivial, But They Will Not Be Enormous Chart II-6Sticky WFH Policies Will Not Prevent A Return To Maximum Employment The Impact Of Lower Office Building Construction A permanently reduced office footprint could also conceivably impact the US economy through reduced nonresidential structures investment, as builders of commercial real estate cease to construct new office towers in response to expectations of a long-lasting glut. However, several points highlight that the negative impact on growth from US office tower construction will be even smaller than the CBD employment impact of reduced office worker presence that we noted above. First, Chart II-7 highlights the overall muted impact that nonresidential building investment has had on real GDP growth by removing the contribution to growth from nonresidential structures and for overall nonresidential investment. The chart clearly highlights that the historically positive contribution to real US output from capital expenditures over the past four decades has come from investment in equipment and intellectual property products, not from structures. Chart II-8 echoes this point, by highlighting that US real investment in nonresidential structures has in fact been flat since the early-1980s, contributing positively and negatively to growth only on a cyclical basis (not on a structural basis). Chart II-7Structures Have Not Contributed Significantly To US Growth For Some Time Chart II-8Nonresidential Structures Investment Has Been Flat For Four Decades Second, Table II-3 highlights that office properties make up a small portion of investment in private nonresidential structures. In 2019, nominal investment in office structures amounted to $85 billion, compared with $630 billion in overall structures investment, meaning that office properties amounted to just 13% of structures investment. Table II-3Office Structures Investment Is A Small Share Of Total Structures Investment Table II-4Conceivably, Vacant Office Properties Could Be Converted To Luxury Residential Units Third, it is true that investment is a flow and not a stock variable, meaning that, if the net stock of office buildings were to fall as a result from WFH policies, then the US economy would see a potentially persistently negative rate of growth from nonresidential structures (which would constitute a drag on growth). But if the net stock were instead to remain flat, then gross office property investment should equal the depreciation of those structures. The second column of Table II-3 highlights that current-cost depreciation of office structures was $53 billion in 2019 (versus nominal gross investment of $85 billion). Had office property investment been ~$30 billion lower in 2019, it would have reduced nominal GDP by a mere 14 basis points (resulting in an annual growth rate of 3.84%, rather than 3.98%). Fourth, there is good reason to believe that the net stock of office properties will stay flat, as the economics of converting offices to luxury housing units (whose demand is not substantially affected by factors such as commuting) – either fully or partially into mixed-use buildings – appear to be plausible. Table II-4 highlights that the average annual asking rent for office space per square foot in Manhattan was $73.23 in Q1 2021, and that the recent median listing home price per square foot is roughly $1,400. In a frictionless world where office space could be instantly and effortlessly sold as residential property, existing prices would imply a healthy (gross) rental yield of 5.2%. Thoughts On The Future Of Office Properties Of course, reality is far from frictionless. There are several barriers that will slow office-to-residential conversion as well as construction costs, which will meaningfully lower the net value of existing office real estate in large central business districts such as Manhattan. In a recent article in the Washington Post, Roger K. Lewis, retired architect and Professor Emeritus of Architecture at the University of Maryland, College Park, detailed several of these technical barriers (which we summarize below).3 Office buildings are typically much wider than residential buildings, the latter usually being 60 to 65 feet in width in order to enable windows and natural light in living/dining rooms and bedrooms. This suggests that office-to-residential conversion might require modifying the basic structure of office buildings, including cutting open parts of roof and floor plates on upper building levels to bring natural light into habitable and interior rooms, and other costly structural modifications to address the additional plumbing and infrastructure that will be needed. Lewis noted that floor-to-floor dimensions are typically larger in office buildings, which is beneficial for office-to-residential conversion because increased room heights augments the sense of space and openness, while allowing natural light to penetrate farther into the apartment. It also allows for extra space to place needed additional building infrastructure, such as sprinkler pipes, electrical conduits, light fixtures, and air ducts. But unique apartment layouts are often needed to use available floor space effectively in an office-to-residential conversion, which will increase design costs and raise the risk that nonstandard layouts may result in unforeseen quality-of-living problems that will necessitate additional future construction to correct. Zoning regulations and building code constraints will likely add another layer of costs to office-to-housing conversions, as these rules are written for conventional buildings, meaning that special exceptions or even regulatory changes are likely to be required. So it is clear that the process of converting office space to residential property will be a costly endeavor for office tower owners, which will likely reduce the net present value of these properties relative to pre-pandemic levels. But; this process appears to be feasible and, when faced with the alternative of persistently high vacancy rates and lost revenue, our sense is that office tower owners will choose this route – thus significantly reducing the likelihood that the growth in national gross investment in office properties will fall below the rate of depreciation. In addition, the trend in suburban and CBD office property prices suggests that there are two other possible alternatives to widespread office-to-residential conversion that would also argue against a significant and long-lasting decline in office structures investment. Chart II-9 highlights that the average asking rent has already fallen significantly in most Manhattan submarkets, and Chart II-10 highlights that suburban office prices are accelerating and rising at the strongest pace relative to CBD office prices over the past two decades, possibly in response to increased demand for workspace that is closer to home for many workers who previously commuted to CBDs. Chart II-9Working From The Office Is Getting Cheaper Chart II-10Suburban Offices Are Getting More Expensive Thus, the first alternative outcome to CBD office-to-residential conversion is that an increase in suburban office construction offsets the negative impact of outright reductions in CBD office investment if residential conversions prove to be too costly or too technically challenging. The second alternative is that owners of CBD office properties “clear the market” by dramatically cutting rental rates even further, to alter the cost/benefit calculation for firms planning permissive WFH policies. We doubt that existing rents reflect the extent of vacancies in large cities such as Manhattan, so we would expect further CBD office price declines in this scenario. But if owners of centrally-located office properties face significant conversion costs and a decline in the net present value of these buildings is unavoidable and its magnitude uncertain, owners may choose to cut prices drastically as the simpler solution. Investment Conclusions Holding all else equal, the fact that owners of CBD office properties are likely to experience some permanent decline in the value of these real estate assets is not a positive development for economic activity. But these losses will be experienced by firms, investors, and ultra-high net worth individuals with strong marginal propensities to save, suggesting that the economic impact from this shock will be minimal. And as we highlighted above, a decline in the pace of gross office building investment to the depreciation rate will have a minimal impact on the overall economy. This leaves the likely impact on CBD employment as the main channel by which WFH policies are likely to affect monetary policy. As we noted above and as discussed in Section 1 of our report, the Fed is now focused entirely on the return of the labor market to maximum employment, which we interpret as an unemployment rate within the range of the Fed’s NAIRU estimates (3.5% - 4.5%) and a return to a pre-pandemic labor force participation rate. Chart II-11On A One-Year Time Horizon, Favor Value Over Growth Our analysis indicates that WFH policies may structurally raise the unemployment rate by 0.3 to 0.4%. While non-trivial, when compared with a pre-pandemic unemployment rate of 3.5%, this suggests that WFH policies alone are not likely to cause a long-term deviation from the Fed’s maximum employment objective. The implication is that job growth over the coming year could be even stronger than the Fed and investors expect, which could mean that the Fed may begin lifting rates by the middle of next year barring a major disruption in the ongoing transition to a post-pandemic world. This is earlier than we currently expect, but the fact that it would also be earlier than what is currently priced into the OIS curve underscores that fixed-income investors should remain short duration on a 6-12 month time horizon. In addition, as noted in Section 1 of our report, while value stocks may underperform growth stocks over the coming 3-4 months,4 rising bond yields over the coming year will ultimately favor value stocks and will likely weigh on elevated tech sector valuations. Chart II-11 highlights that the relative valuation of growth stocks remains above its pre-pandemic starting point (Chart II-11), suggesting that investors should continue to favor MSCI-benchmarked value over growth positions over a 6-12 month time horizon. Finally, as also noted in Section 1 of our report, we do not expect rising bond yields to prevent stock prices from grinding higher over the coming year, unless investor expectations for the terminal fed funds rate move sharply higher – an event that seems unlikely, although not impossible, before monetary policy actually begins to tighten. Jonathan LaBerge, CFA Vice President The Bank Credit Analyst Footnotes 1 Please see The Bank Credit Analyst "June 2021," dated May 27, 2021, available at bca.bcaresearch.com 2 Readers should note that the desired share of paid work from home days post-COVID among employees is shown to be lower in Table II-1 than what is implied by Chart II-1 on a weighted-average basis. This is due to the fact that Table II-1 excludes responses from the May 2020 survey wave, because the authors did not ask about employer intensions during that wave. This underscores that the average desired number of paid days working from home declined somewhat over time, and thus argues for the value shown in Table II-1 as the best estimate for employee preferences. 3 Roger K. Lewis, “Following pandemic, converting office buildings into housing may become new ‘normal,’ Washington Post, April 3, 2021. 4 Please see US Equity Strategy "Rotate Into Growth Stocks, Be Granular In The Selection Of Cyclicals," dated June 14, 2021, available at uses.bcaresearch.com

Highlights The ongoing transition to a post-pandemic state and fiscal policy are either positive or net-neutral for risky asset prices. Fiscal thrust will turn to fiscal drag over the coming year, but the negative impact this will have on goods spending will likely be offset by a significant improvement in services spending, and thus is not likely to cause a concerning slowdown in overall economic activity. A modestly hawkish shift in the outlook for monetary policy is likely over the coming year, potentially occurring over the late summer or early fall in response to outsized jobs growth. However, such a shift is not likely to become a negative driver for risky asset prices over the coming 6-12 months, barring a major rise in market expectations for the neutral rate of interest. This may very well occur once the Fed begins to raise interest rates, but not likely before. Investors should overweight risky assets within a multi-asset portfolio, and fixed-income investors should maintain a below-benchmark duration position. We continue to favor value over growth on a 6-12 month time horizon, although growth may outperform in the near term. A bias toward value over the coming year supports an overweight stance toward global ex-US equities, and an overall pro-risk stance favors bearish US dollar bets. Feature Three factors continue to drive our global macroeconomic outlook and our cyclical investment recommendations. The first factor is our assessment of the global progress that is being made on the path to a post-pandemic state, and the return to pre-COVID economic conditions; the second is the likely contribution to growth from fiscal policy over the coming year; and the third is the outlook for monetary policy and whether or not monetary conditions will remain stimulative for both economic activity and financial markets. If the world continues to progress meaningfully on the path to a post-pandemic state, and if the impact of fiscal and monetary policy remains in line with market expectations, then we see no reason to alter our recommended investment stance. Equity market returns will be modest over the coming 6 to 12 months in this scenario given how significantly stocks have rebounded from their low last year, but we would still expect stocks to outperform bonds and would generally be pro-cyclically positioned. We present below our assessment of these three factors and their potential to deviate from consensus expectations over the coming year, to determine their likely impact on economic activity and financial markets. The Ongoing Transition To A Post-Pandemic World Chart I-1Enormous Progress Has Been Made In The Fight Against COVID-19 Chart I-1 highlights that meaningful progress continues to be made in vaccinating the world's population against COVID-19. North America and Europe continue to lead the rest of the world based on the share of people who have received at least one dose, but South America continues to make significant gains, and recent data updates highlight that Asia and Oceania are also making meaningful progress. Africa is the clear laggard in the war against SARS-COV-2 and its variants, but progress there has been delayed, at least in part, by India’s export restrictions of the Oxford-AstraZeneca/COVISHIELD vaccine. This suggests that, while Africa will continue to lag, the share of Africans provided with a first dose of vaccine will begin to rise once India resumes its exports and deliveries to African countries under the COVAX program continue. If variants of the disease were not a source of concern, Chart I-1 would highlight that the full transition to a post-pandemic economy over the next several months would be near certain. However, as evidenced by the recent decision in the UK to postpone the lifting of COVID-19 restrictions by 4 weeks due to the spreading of the Delta variant, the global economy is not entirely out of the woods yet. Encouragingly, the delay in the UK genuinely appears to be temporary. Chart I-2 highlights that while the number of confirmed UK COVID-19 cases has been rising over the past month, the uptick in hospitalizations and fatalities has so far been quite muted. Importantly, the rise in hospitalizations appears to be occurring among those who have not yet been fully vaccinated, underscoring that variants of the disease are only truly concerning if they are vaccine-resistant. The evidence so far is that the Delta variant is more transmissible and may increase the risk of hospitalization, but that two doses of COVID-19 vaccine offer high protection. Of course, vaccines only offer protection if you get them, and evidence of vaccination hesitancy in the US is thus a somewhat worrying sign. Chart I-3 shows that the daily pace of vaccinations in the US has slowed significantly from mid-April levels, resulting in a slower rise in the share of the population that has received at least one dose (second panel). On this metric, the US has recently been outpaced by Canada, and the gap between the UK and the US is now widening. Germany and France are close behind the US and may surpass it soon. Chart I-2The UK Delay In Removing Restrictions Seems Genuinely Temporary Chart I-3Recent Vaccination Progress In The US Has Been Underwhelming   Sadly, Chart I-4 highlights that there is a political dimension to vaccine hesitancy in the US. The chart shows that state by state vaccination rates as a share of the population are strongly predicted by the share of the popular vote for Donald Trump in the 2020 US presidential election. Admittedly, part of this relationship may also be capturing an urban/rural divide, with residents in less-dense rural areas (which typically support Republican presidential candidates) perhaps feeling a lower sense of urgency to become vaccinated against the disease. Chart I-4The US Politicization Of Vaccines Raises The Risk From COVID-19 Variants But given the clear politicization that has already occurred over some pandemic control measures, such as the wearing of masks, Chart I-4 makes it difficult to avoid the conclusion that the same thing has occurred for vaccines. This is unfortunate, and seemingly raises the risk that the Delta variant may spread widely in red states over the coming several months, potentially delaying economic reopening, or risking the reintroduction of pandemic control measures. However, there are two counterarguments to this concern. First, non-vaccine immunity is probably higher in red than blue states, and CDC data suggest that this effect could be large. While this figure is still preliminary and subject to change (and likely will), the CDC estimates that only 1 out of 4.3 cases of COVID-19 were reported from February 2020 to March 2021. Taken at face value, this implies that there were approximately 115 million infections during that period, compared with under 30 million reported cases. That gap accounts for 25% of the US population, and given that red states were slower to implement pandemic control measures last year and their residents often more resistant to the measures, it stands to reason that a disproportionate share of unreported cases occurred in these states. Second, as noted above, the evidence thus far suggests that the Delta variant is not vaccine resistant, at least for those who are fully vaccinated. This is significant because if Delta were to spread widely in red states over the coming several months, the resulting increase in hospitalizations would likely convince many vaccine hesitant Americans to become vaccinated out of fear and self-interest – two powerfully motivating factors. Thus, the Delta variant may become a problem for the US in the fall, but if that occurs a solution is not far from sight. And, in other developed countries where vaccine hesitancy rates appear to be lower, it would seem that a new, vaccine-resistant variant of the disease would likely be required in order to cause a major disruption in the transition to a post-pandemic state. Such a variant could emerge, but we have seen no evidence thus far that one will before vaccination rates reach levels that would slash the odds of further widespread mutation. Fiscal Policy: Passing The Baton To Services Spending Chart I-5 highlights that US fiscal policy is set to detract from growth over the coming 6-12 months, reflecting the one-off nature of some of the fiscal response to the pandemic. This is true outside of the US as well, as Chart I-6 highlights that the IMF is forecasting a two percentage point increase in the Euro Area’s cyclically-adjusted primary budget balance, representing a significant amount of fiscal drag relative to the past two decades. Chart I-5Fiscal Thrust Will Eventually Turn To Fiscal Drag In The US… Should investors be concerned about the impact of fiscal drag on advanced economies over the coming year? In our view, the answer is no. The reason is that much of the fiscal response in the US and Europe has been aimed at supporting income that has been lost due to a drastic reduction in services spending, which will continue to recover over the coming months as the effect of the pandemic continues to ebb. Chart I-7 underscores this point by highlighting the “gap” in US consumer goods and services spending relative to its pre-pandemic trend. The chart highlights that US goods spending is running well above what would be expected, whereas there is a sizeable gap in services spending (which accounts for approximately 70% of US personal consumption expenditures). Goods spending will likely slow as fiscal thrust turns to fiscal drag, but services spending will improve meaningfully – aided not just by a post-pandemic normalization in economic activity, but also by the sizeable amount of excess savings that US households have accumulated over the past year (Chart I-7, panel 2). Chart I-6... And In Europe Chart I-7But Reduced Transfers Will Only Impact Spending On Goods, Not Services While some of these savings have already been deployed to pay down debt and some may be permanently saved in anticipation of higher future taxes, the key point for investors is that the negative impact on goods spending from reduced fiscal thrust will be offset by a significant improvement in services spending, and thus is not likely to cause a concerning slowdown in overall economic activity. Monetary Policy: A Modestly Hawkish Shift Is Likely This leaves us with the question of whether or not monetary policy will become a negative driver for risky asset prices over the coming 6-12 months, which is especially relevant following last week’s FOMC meeting. The updated “dot plot” following the meeting shows that 7 of the 18 FOMC participants anticipate a rate hike in 2022, and the majority (13 members) expect at least one rate hike before the end of 2023, raising the median forecast for the Fed funds rate to 0.6% by the end of that year. Chart I-8 highlights that while 10-year Treasury yields remains mostly unchanged following the meeting, yields moved higher at the short-end and middle of the curve. Chart I-8The FOMC Meeting Resulted In Higher Short- And Mid-Term Yields Investor fears that the Fed may shift in a significantly hawkish direction at some point over the next year have been far too focused on inflation, and far too little focused on employment. It is not a coincidence that the Fed’s guidance was updated following the May jobs report, which saw a stronger pace of jobs growth relative to April. Table I-1 updates our US Bond Strategy service’s calculations showing the average monthly nonfarm payroll growth that will be required for the unemployment rate to reach 3.5-4.5% assuming a full recovery in the participation rate, which is the range of the Fed’s NAIRU estimates. May’s payroll growth number of 560k implies that the Fed’s maximum employment criterion will be met sometime between June and September next year, if monthly payroll growth continues at that pace. Table I-1Calculating The Distance To Maximum Employment Chart I-9Lighter Restrictions In Blue States Will Push Down The Unemployment Rate It is currently difficult to assess with great confidence what average payroll growth will prevail over the coming year, but we noted in last month’s report that there were compelling arguments in favor of outsized jobs growth this fall.1 In addition to those points, we note the following: Blue states have generally been slower to reopen their economies, and Chart I-9 highlights that these states have consequently been slower to return to their pre-pandemic unemployment rate. Among blue states, California and New York are the largest by population, and it is notable that both states only lifted most COVID-19 restrictions on June 15 – including the wearing of masks in most settings. This implies that services jobs are likely to grow significantly in these states over the coming few months. Both consensus private forecasts as well as the Fed’s expectation for real GDP growth imply that the output gap will be closed by Q4 of this year (Chart I-10). These expectations appear to be reasonable, given the substantial amount of excess savings that have been accumulated by US households and the fact that monetary policy remains extremely stimulative. When the output gap turned positive during the last economic cycle, the unemployment rate was approximately 4% – well within the Fed’s NAIRU range. Chart I-10 also shows that the Fed’s 7% real GDP growth forecast for this year would put the output gap above its pre-pandemic level, when the unemployment rate stood at 3.5%. In fact, it is possible that annualized Q2 real GDP growth will disappoint current consensus expectations of 10%, due to the scarcity of labor supply (scarcity that will be eased by labor day when supplemental unemployment insurance benefit programs end). Were Q2 GDP to disappoint due to supply-side limitations, it would strengthen the view that job gains will be very strong this fall ceteris paribus, as it would highlight that real output per worker cannot rise meaningfully further in the short-term and that stronger growth later in the year will necessitate very large job gains. Chart I-11 highlights that US air travel and New York City subway ridership have already returned close to 75% and 50% of their pre-pandemic levels, respectively. Based on the trend over the past three months, the chart implies that air travel will return to its pre-pandemic levels by mid-October of this year, and New York City subway ridership by June 2022. This underscores that travel-related services employment will recover significantly in the fall, and that jobs in downtown cores will rebound as office workers progressively return to work. Chart I-10Expectations For Growth This Year Suggest A Rapid Decline In The Unemployment Rate Chart I-11Services Employment Will Recover In The Fall   On the latter point, one major outstanding question affecting the outlook for monetary policy is the magnitude of the likely permanent impact of work from home policies on employment in central business districts. Fewer office workers commuting to downtown office locations suggests that some jobs in the leisure & hospitality, retail trade, professional & business services, and other services industries will never return or will be very slow to do so, arguing for a longer return to maximum employment (and the Fed’s liftoff date). We examine this question in depth in Section 2 of this month’s report, and find that the “stickiness” of work from home policies will likely cause permanent central business job losses on the order of 575k (or 0.35% of the February 2020 labor force). While this would be non-trivial, when compared with a pre-pandemic unemployment rate of 3.5%, WFH policies alone are not likely to cause a long-term deviation from the Fed’s maximum employment objective. Outsized jobs growth this fall, at a pace that quickly reduces the unemployment rate, argues for a first Fed rate hike that is even earlier than the market expects. Chart I-12 presents The Bank Credit Analyst service’s current assessment of the cumulative odds of the Fed’s liftoff date by quarter; we believe that it is likely that the Fed will have raised rates by Q3 of next year, and that a rate hike in the first half of 2022 is a possibility. These odds are slightly more aggressive than those presented by our fixed-income strategists in a recent Special Report,2 but are consistent with their view that the Fed will raise interest rates by the end of next year. Chart I-12The Bank Credit Analyst’s Assessment Of The Odds Of The First Rate Hike The odds presented in Chart I-12 are also more hawkish than the Fed funds rate path currently implied by the OIS curve, meaning that we expect investors to be somewhat surprised by a shifting monetary policy outlook at some point over the coming year, potentially over the next 3-6 months. Payroll growth during the late summer and early fall will be a major test for the employment outlook, and is the most likely point for a hawkish shift in the market’s view of monetary policy. Is this likely to become a negative driver for risky asset prices over the coming 6-12 months? In our view, the answer is “probably not.” While investors tend to focus heavily on the timing of the first rate hike as monetary policy begins to tighten, the reality is that it is the least relevant factor driving the fair value of 10-year Treasury yields. Investor expectations for the pace of tightening and especially for the terminal Fed funds rate are far more important, and, while it is quite possible that expectations for the neutral rate of interest will eventually rise, it seems unlikely that this will occur before the Fed actually begins to raise interest rates given that most investors accept the secular stagnation narrative and the view that “R-star” is well below trend rates of growth (we disagree).3 Chart I-13 highlights the fair value path of 10-year Treasury yields until the end of next year, assuming a 2.5% terminal Fed funds rate, no term premium, and a rate hike pace of 1% per year. The chart highlights that while government bond yields are set to move higher over the coming 6-12 months, they are likely to remain between 2-2.5%. This would drop the equity risk premium to a post-2008 low (Chart I-14), which would further reduce the attractiveness of stocks relative to bonds. But we doubt that this would be enough of a decline to cause a selloff, and it would still imply a stimulative level of interest rates for households and firms. Chart I-1310-Year Yields Will Rise Over The Coming Year, But Not Sharply Chart I-14Rising Yields Will Cause An Unwelcome But Contained Decline In The ERP   Investment Conclusions Among the three factors driving our global macroeconomic outlook and our cyclical investment recommendations, continued progress on the path toward a post-pandemic state and fiscal policy remain either positive or mostly neutral for risky assets. A potentially hawkish shift in the outlook for monetary policy this fall remains the chief risk, but we expect the rise in bond yields over the coming year to remain well-contained barring a sea change in investor expectations for the terminal Fed funds rate – which we believe is unlikely to occur before the Fed begins to raise interest rates. Consequently, we continue to recommend that investors should overweight risky assets within a multi-asset portfolio, and that fixed-income investors should maintain a below-benchmark duration position. We expect modest absolute returns from global equities, but even mid-single digit returns are likely to beat those from long-dated government bonds and cash positions. While value stocks may underperform growth stocks over the coming 3-4 months,4 rising bond yields over the coming year will ultimately favor value stocks and will likely weigh on elevated tech sector (and therefore growth stock) valuations (Chart I-15). Chart I-16 highlights that the attractiveness of US value versus growth is meaningfully less compelling for the S&P 500 Citigroup indexes, suggesting that investors should continue to favor MSCI-benchmarked value over growth positions over a 6-12 month time horizon.5 Chart I-15Value Is Extremely Cheap Chart I-16Value Vs. Growth: The Benchmark Matters   The likely outperformance of value versus growth also has implications for regional allocation within a global equity portfolio. The US is significantly overweight broadly-defined technology relative to global ex-US stocks, and financials – which are overrepresented in value indexes – have already meaningfully outperformed in the US this year compared with their global peers and are now rolling over (Chart I-17). This underscores that investors should favor ex-US stocks over the coming year, skewed in favor of DM ex-US given that China’s credit impulse continues to slow (Chart I-18). Chart I-17Favor Global Ex-US Stocks Over The Coming Year Chart I-18Concentrate Global Ex-US Exposure In Developed Markets   Finally, global ex-US stocks also tend to outperform when the US dollar is falling, and we would recommend that investors maintain a short dollar position on a 6-12 month time horizon despite the recent bounce in the greenback. Chart I-19 highlights that the dollar remains strongly negatively correlated with global equity returns, and that the dollar’s performance over the past year has been almost exactly in line with what one would have expected given this relationship. Thus, a bullish view toward global stocks implies both US dollar weakness and global ex-US outperformance over the coming year. Chart I-19A Bullish View Towards Global Stocks Implies A Dollar Bear Market Jonathan LaBerge, CFA Vice President The Bank Credit Analyst June 24, 2021 Next Report: July 29, 2021   II. Work From Home “Stickiness” And The Outlook For Monetary Policy Work from home policies, originally designed as emergency measures in the early phase of the COVID-19 pandemic, are likely to be “sticky” in a post-pandemic world. This will negatively impact the labor market in central business districts, via reduced spending on services by office workers. The potential impact of working from home is often cited as an example of what is likely to be a lasting and negative effect on jobs growth, but we find that it is not likely to be a barrier to the labor market returning to the Fed’s assessment of “maximum employment.” The size of the impact depends importantly on whether employee preferences or employer plans for WFH prevail, but our sense is that the latter is more likely. A weaker pace of structures investment in response to elevated office vacancy rates will likely have an even smaller impact on growth than the effect of reduced central business district services employment. The contribution to growth from structures investment has been small over the past few decades, office building construction is a small portion of overall nonresidential structures, and there are compelling arguments that the net stock of office structures will stay flat, rather than decline. Our analysis suggests that job growth over the coming year could be even stronger than the Fed and investors expect, possibly resulting in a first rate hike by the middle of next year. This would be earlier than we currently anticipate, but it underscores that fixed-income investors should remain short duration on a 6-12 month time horizon, and that equity investors should favor value over growth positions beyond the coming 3-4 months. The outlook for US monetary policy over the next 12 to 18 months depends almost entirely on the outlook for employment. Many investors are focused on the potential for elevated inflation to force the Fed to raise interest rates earlier than it currently anticipates, but it is the progress in returning to “maximum employment” that will determine the timing of the first Fed rate hike – and potentially the speed at which interest rates rise once policy begins to tighten. In this report, we estimate the extent to which the “stickiness” of working from home (WFH) policies and practices could leave a lasting negative impact on the US labor market. We noted in last month's report that a large portion of the employment gap relative to pre-pandemic levels can be traced to the leisure & hospitality and professional and business services industries, both of which – along with retail employment – stand to be permanently impaired if the office worker footprint is much lower in a post-COVID world.6 Using employee surveys and a Monte Carlo approach, we present a range of estimates for the permanent impact of WFH policies on the unemployment rate, and separately examine the potential for lower construction of office properties to weigh on growth. We find that the impact of reduced office building construction is likely to be minimal, and that WFH policies may structurally raise the unemployment rate by 0.3 to 0.4%. While non-trivial, when compared with a pre-pandemic unemployment rate of 3.5%, WFH policies alone are not likely to cause a long-term deviation from the Fed’s maximum employment objective. Relative to the Fed’s expectations of a strong, lasting impact on the labor market from the pandemic, this suggests that job growth over the coming year could be even stronger than the Fed and investors expect, possibly resulting in a first rate hike by the middle of next year. This would be earlier than we currently anticipate, but it underscores that fixed-income investors should remain short duration on a 6-12 month time horizon, and that equity investors should favor value over growth positions beyond the coming 3-4 months (a period that may see outperformance of the latter). Quantifying The Labor Market Impact Of The New Normal For Work In a January paper, Barrero, Bloom, and Davis (“BBD”) presented evidence arguing why working from home will “stick.” The authors surveyed 22,500 working-age Americans across several survey “waves” between May and December 2020, and asked about both their preferences and their employer’s plans about working from home after the pandemic. Chart II-1 highlights that the desired amount of paid work from home days (among workers who can work from home) reported by the survey respondents is to approximately 55% of a work week, suggesting that a dramatic reduction in office presence would likely occur if post-pandemic WFH policies were set fully in accordance with worker preferences. Chart II-1Employee Preferences Imply A Dramatic Reduction In Post-COVID Office Presence However, Table II-1 highlights that employer plans for work from home policies are meaningfully different than those of employees. The table highlights that employers plan for employees to work from home for roughly 22% of paid days post-pandemic, which essentially translates to one day per week on average.7 BBD noted that CEOs and managers have cited the need to support innovation, employee motivation, and company culture as reasons for employees’ physical presence. Managers believe physical interactions are important for these reasons, but employees need only be on premises for about three to four days a week to achieve this. Table II-1 also shows that employers plan to allow higher-income employees more flexibility in terms of working from home, and less flexibility to employees whose earnings are between $20-50k per year. Table II-1Employer Plans, However, Imply Less Working From Home Than Employees Prefer Based on the survey results, BBD forecast that expenditure in major cities such as Manhattan and San Francisco will fall on the order of 5 to 10%. In order to understand the national labor market impact of work from home policies and what implications this may have on monetary policy, we scale up BBD’s calculations using a Monte Carlo approach that incorporates estimate ranges for several factors: The percent of paid days now working from home for office workers The amount of money spent per week by office workers in central business districts (“CBDs”) The number of total jobs in CBDs The percent of CBD jobs in industries likely to be negatively impacted by reduced office worker expenditure The average weekly earnings of affected CBD workers The average share of business revenue not attributable to strictly variable expenses The percent of affected jobs likely to be recovered outside of CBDs Our approach is as follows. First, we calculate the likely reduction in nationwide CBD spending from reduced office worker presence by multiplying the likely percent of paid days now permanently working from home by the number of total jobs in CBDs and the average weekly spending of office workers. This figure is then increased due to the estimated acceleration in net move outs from principal urban centers in 2020 (Chart II-2); we assume a 5% savings rate and an average annual salary of $50k for these resident workers, and assume that all of their spending occurred within CBDs. We also assume that roughly 50% of jobs connected to this spending are recovered. Chart II-2Fewer Residents Will Also Lower Spending In Central Business Districts Then, we calculate the gross number of jobs lost in leisure & hospitality, retail trade, and other services by multiplying this estimate of lost spending by an estimate of non-variable costs as a share of revenue for affected industries, and dividing the result by average weekly earnings of affected employees. For affected CBD employees in the administrative and waste services industry, we simply assume that the share of jobs lost matches the percent of paid days now permanently working from home. Finally, we adjust the number of jobs lost by multiplying by 1 minus an assumed “recovery” rate, given that some of the reduction in spending in CBDs will simply be shifted to areas near remote workers’ residences. We assume a slightly lower recovery rate for lost jobs in the administrative and waste services industry. Table II-2 highlights the range of outcomes for each variable used in our simulation, and Charts II-3 and II-4 present the results. The charts highlight that the distribution of outcomes based on employer WFH intensions suggest high odds that nationwide job losses in CBDs due to reduced office worker presence will not exceed 400k. Based on average employee preferences, that number rises to roughly 800-900k. Table II-2The Factors Affecting Permanent Central Business District Job Losses Chart II-3The Probability Distribution Of CBD Jobs Lost… Chart II-4…Based On Our Monte Carlo Approach   This raises the question of whether employer plans or employee preferences for WFH arrangements will prevail. Our sense is that it will be closer to the former, given that we noted above that employer WFH plans are the least flexible for employees whose earnings are between $20-50k per year (who are presumably employees who have less ability to influence the policy of firms). Chart II-5 re-presents the projected job losses shown in Chart II-4 as a share of the February 2020 labor force, along with a probability-weighted path that assumes a 75% chance that employer WFH plans will prevail. The chart highlights that WFH arrangements would have the effect of raising the unemployment rate by approximately 0.35%. However, relative to a pre-pandemic starting point of 3.5%, this would raise the unemployment rate to a level that would still be within the Fed’s NAIRU estimates (Chart II-6). Therefore, the “stickiness” of WFH arrangements alone do not seem to be a barrier to the labor market returning to the Fed’s assessment of “maximum employment,” suggesting that the conditions for liftoff may be met earlier than currently anticipated by investors. Chart II-5CBD Job Losses Will Not Be Trivial, But They Will Not Be Enormous Chart II-6Sticky WFH Policies Will Not Prevent A Return To Maximum Employment The Impact Of Lower Office Building Construction A permanently reduced office footprint could also conceivably impact the US economy through reduced nonresidential structures investment, as builders of commercial real estate cease to construct new office towers in response to expectations of a long-lasting glut. However, several points highlight that the negative impact on growth from US office tower construction will be even smaller than the CBD employment impact of reduced office worker presence that we noted above. First, Chart II-7 highlights the overall muted impact that nonresidential building investment has had on real GDP growth by removing the contribution to growth from nonresidential structures and for overall nonresidential investment. The chart clearly highlights that the historically positive contribution to real US output from capital expenditures over the past four decades has come from investment in equipment and intellectual property products, not from structures. Chart II-8 echoes this point, by highlighting that US real investment in nonresidential structures has in fact been flat since the early-1980s, contributing positively and negatively to growth only on a cyclical basis (not on a structural basis). Chart II-7Structures Have Not Contributed Significantly To US Growth For Some Time Chart II-8Nonresidential Structures Investment Has Been Flat For Four Decades Second, Table II-3 highlights that office properties make up a small portion of investment in private nonresidential structures. In 2019, nominal investment in office structures amounted to $85 billion, compared with $630 billion in overall structures investment, meaning that office properties amounted to just 13% of structures investment. Table II-3Office Structures Investment Is A Small Share Of Total Structures Investment Table II-4Conceivably, Vacant Office Properties Could Be Converted To Luxury Residential Units Third, it is true that investment is a flow and not a stock variable, meaning that, if the net stock of office buildings were to fall as a result from WFH policies, then the US economy would see a potentially persistently negative rate of growth from nonresidential structures (which would constitute a drag on growth). But if the net stock were instead to remain flat, then gross office property investment should equal the depreciation of those structures. The second column of Table II-3 highlights that current-cost depreciation of office structures was $53 billion in 2019 (versus nominal gross investment of $85 billion). Had office property investment been ~$30 billion lower in 2019, it would have reduced nominal GDP by a mere 14 basis points (resulting in an annual growth rate of 3.84%, rather than 3.98%). Fourth, there is good reason to believe that the net stock of office properties will stay flat, as the economics of converting offices to luxury housing units (whose demand is not substantially affected by factors such as commuting) – either fully or partially into mixed-use buildings – appear to be plausible. Table II-4 highlights that the average annual asking rent for office space per square foot in Manhattan was $73.23 in Q1 2021, and that the recent median listing home price per square foot is roughly $1,400. In a frictionless world where office space could be instantly and effortlessly sold as residential property, existing prices would imply a healthy (gross) rental yield of 5.2%. Thoughts On The Future Of Office Properties Of course, reality is far from frictionless. There are several barriers that will slow office-to-residential conversion as well as construction costs, which will meaningfully lower the net value of existing office real estate in large central business districts such as Manhattan. In a recent article in the Washington Post, Roger K. Lewis, retired architect and Professor Emeritus of Architecture at the University of Maryland, College Park, detailed several of these technical barriers (which we summarize below).8 Office buildings are typically much wider than residential buildings, the latter usually being 60 to 65 feet in width in order to enable windows and natural light in living/dining rooms and bedrooms. This suggests that office-to-residential conversion might require modifying the basic structure of office buildings, including cutting open parts of roof and floor plates on upper building levels to bring natural light into habitable and interior rooms, and other costly structural modifications to address the additional plumbing and infrastructure that will be needed. Lewis noted that floor-to-floor dimensions are typically larger in office buildings, which is beneficial for office-to-residential conversion because increased room heights augments the sense of space and openness, while allowing natural light to penetrate farther into the apartment. It also allows for extra space to place needed additional building infrastructure, such as sprinkler pipes, electrical conduits, light fixtures, and air ducts. But unique apartment layouts are often needed to use available floor space effectively in an office-to-residential conversion, which will increase design costs and raise the risk that nonstandard layouts may result in unforeseen quality-of-living problems that will necessitate additional future construction to correct. Zoning regulations and building code constraints will likely add another layer of costs to office-to-housing conversions, as these rules are written for conventional buildings, meaning that special exceptions or even regulatory changes are likely to be required. So it is clear that the process of converting office space to residential property will be a costly endeavor for office tower owners, which will likely reduce the net present value of these properties relative to pre-pandemic levels. But; this process appears to be feasible and, when faced with the alternative of persistently high vacancy rates and lost revenue, our sense is that office tower owners will choose this route – thus significantly reducing the likelihood that the growth in national gross investment in office properties will fall below the rate of depreciation. In addition, the trend in suburban and CBD office property prices suggests that there are two other possible alternatives to widespread office-to-residential conversion that would also argue against a significant and long-lasting decline in office structures investment. Chart II-9 highlights that the average asking rent has already fallen significantly in most Manhattan submarkets, and Chart II-10 highlights that suburban office prices are accelerating and rising at the strongest pace relative to CBD office prices over the past two decades, possibly in response to increased demand for workspace that is closer to home for many workers who previously commuted to CBDs. Chart II-9Working From The Office Is Getting Cheaper Chart II-10Suburban Offices Are Getting More Expensive Thus, the first alternative outcome to CBD office-to-residential conversion is that an increase in suburban office construction offsets the negative impact of outright reductions in CBD office investment if residential conversions prove to be too costly or too technically challenging. The second alternative is that owners of CBD office properties “clear the market” by dramatically cutting rental rates even further, to alter the cost/benefit calculation for firms planning permissive WFH policies. We doubt that existing rents reflect the extent of vacancies in large cities such as Manhattan, so we would expect further CBD office price declines in this scenario. But if owners of centrally-located office properties face significant conversion costs and a decline in the net present value of these buildings is unavoidable and its magnitude uncertain, owners may choose to cut prices drastically as the simpler solution. Investment Conclusions Holding all else equal, the fact that owners of CBD office properties are likely to experience some permanent decline in the value of these real estate assets is not a positive development for economic activity. But these losses will be experienced by firms, investors, and ultra-high net worth individuals with strong marginal propensities to save, suggesting that the economic impact from this shock will be minimal. And as we highlighted above, a decline in the pace of gross office building investment to the depreciation rate will have a minimal impact on the overall economy. This leaves the likely impact on CBD employment as the main channel by which WFH policies are likely to affect monetary policy. As we noted above and as discussed in Section 1 of our report, the Fed is now focused entirely on the return of the labor market to maximum employment, which we interpret as an unemployment rate within the range of the Fed’s NAIRU estimates (3.5% - 4.5%) and a return to a pre-pandemic labor force participation rate. Chart II-11On A One-Year Time Horizon, Favor Value Over Growth Our analysis indicates that WFH policies may structurally raise the unemployment rate by 0.3 to 0.4%. While non-trivial, when compared with a pre-pandemic unemployment rate of 3.5%, this suggests that WFH policies alone are not likely to cause a long-term deviation from the Fed’s maximum employment objective. The implication is that job growth over the coming year could be even stronger than the Fed and investors expect, which could mean that the Fed may begin lifting rates by the middle of next year barring a major disruption in the ongoing transition to a post-pandemic world. This is earlier than we currently expect, but the fact that it would also be earlier than what is currently priced into the OIS curve underscores that fixed-income investors should remain short duration on a 6-12 month time horizon. In addition, as noted in Section 1 of our report, while value stocks may underperform growth stocks over the coming 3-4 months,9 rising bond yields over the coming year will ultimately favor value stocks and will likely weigh on elevated tech sector valuations. Chart II-11 highlights that the relative valuation of growth stocks remains above its pre-pandemic starting point (Chart II-11), suggesting that investors should continue to favor MSCI-benchmarked value over growth positions over a 6-12 month time horizon. Finally, as also noted in Section 1 of our report, we do not expect rising bond yields to prevent stock prices from grinding higher over the coming year, unless investor expectations for the terminal fed funds rate move sharply higher – an event that seems unlikely, although not impossible, before monetary policy actually begins to tighten. Jonathan LaBerge, CFA Vice President The Bank Credit Analyst III. Indicators And Reference Charts BCA’s equity indicators highlight that the “easy” money from expectations of an eventual end to the pandemic have already been made. Our technical, valuation, and sentiment indicators are very extended, highlighting that investors should expect positive but more modest returns from stocks over the coming 6-12 months. Our monetary indicator has aggressively retreated from its high last year, reflecting a meaningful recovery in government bond yields since last August. The indicator still remains above the boom/bust line, however, highlighting that monetary policy remains supportive for risky asset prices. Forward equity earnings already price in a complete earnings recovery, but for now there is no meaningful sign of waning forward earnings momentum. Net revisions remain very strong, and positive earnings surprises have risen to their highest levels on record. Within a global equity portfolio, there has been a modest tick down in global ex-US equity performance, driven by a rally in growth stocks (which may persist for a few months). EM stocks had previously dragged down global ex-US performance, and they continue to languish. A bias towards value stocks on a 1-year time horizon means that investors should still favor ex-US stocks over the coming year, skewed in favor of DM ex-US given that China’s credit impulse continues to slow. The US 10-Year Treasury yield has trended modestly lower since mid-March, after having risen to levels that were extremely technically stretched. Despite this pause, our valuation index highlights that bonds are still expensive, and we expect that yields will move higher over the cyclical investment horizon if employment growth in Q3/Q4 implies a faster return to maximum employment than currently projected by the Fed. We expect the rise to be more modest than our valuation index would imply, but we would still recommend a short duration stance within a fixed-income portfolio. The extreme rise in some commodity prices over the past several months is beginning to ease. Lumber prices have fallen close to 50% from their recent high, whereas industrial metals and agricultural prices are down roughly 5% and 17%, respectively. We had previously argued that a breather in commodity prices was likely at some point over the coming several months, and we would expect further declines as supply chains normalize, labor supply recovers, and Chinese demand for metals slows. US and global LEIs remain in a solid uptrend, and global manufacturing PMIs are strong. Our global LEI diffusion index has declined significantly, but this likely reflects the outsized impact of a few emerging market countries (whose vaccination progress is still lagging). Strong leading and coincident indicators underscore that the global demand for goods is robust, and that output is below pre-pandemic levels in most economies because of very weak services spending. The latter will recover significantly later this year, as social distancing and other pandemic control measures disappear. EQUITIES: Chart III-1US Equity Indicators Chart III-2Willingness To Pay For Risk Chart III-3US Equity Sentiment Indicators   Chart III-4US Stock Market Breadth Chart III-5US Stock Market Valuation Chart III-6US Earnings Chart III-7Global Stock Market And Earnings: Relative Performance Chart III-8Global Stock Market And Earnings: Relative Performance   FIXED INCOME: Chart III-9US Treasurys And Valuations Chart III-10Yield Curve Slopes Chart III-11Selected US Bond Yields Chart III-1210-Year Treasury Yield ComponentsChart III-13US Corporate Bonds And Health Monitor Chart III-14Global Bonds: Developed Markets Chart III-15Global Bonds: Emerging Markets   CURRENCIES: Chart III-16US Dollar And PPP Chart III-17US Dollar And Indicator Chart III-18US Dollar Fundamentals Chart III-19Japanese Yen Technicals Chart III-20Euro Technicals Chart III-21Euro/Yen Technicals Chart III-22Euro/Pound Technicals   COMMODITIES: Chart III-23Broad Commodity Indicators Chart III-24Commodity Prices Chart III-25Commodity Prices Chart III-26Commodity Sentiment Chart III-27Speculative Positioning   ECONOMY: Chart III-28US And Global Macro Backdrop Chart III-29US Macro Snapshot Chart III-30US Growth Outlook Chart III-31US Cyclical Spending Chart III-32US Labor Market Chart III-33US Consumption Chart III-34US Housing Chart III-35US Debt And Deleveraging   Chart III-36US Financial Conditions Chart III-37Global Economic Snapshot: Europe Chart III-38Global Economic Snapshot: China   Jonathan LaBerge, CFA Vice President The Bank Credit Analyst Footnotes 1 Please see The Bank Credit Analyst "June 2021," dated May 27, 2021, available at bca.bcaresearch.com 2 Please see US Bond Strategy/Global Fixed Income Strategy Special Report "A Central Bank Timeline For The Next Two Years," dated June 1, 2021, available at usbs.bcaresearch.com 3 Please see The Bank Credit Analyst Special Report "R-star, And The Structural Risk To Stocks," dated March 31, 2021, available at bca.bcaresearch.com 4 Please see US Equity Strategy "Rotate Into Growth Stocks, Be Granular In The Selection Of Cyclicals," dated June 14, 2021, available at uses.bcaresearch.com 5 For a discussion of the differences in value and growth benchmarks, please see Global Asset Allocation Special Report “Value? Growth? It Really Depends!” dated September 19, 2019, available at gaa.bcaresearch.com 6 Please see The Bank Credit Analyst "June 2021," dated May 27, 2021, available at bca.bcaresearch.com 7 Readers should note that the desired share of paid work from home days post-COVID among employees is shown to be lower in Table II-1 than what is implied by Chart II-1 on a weighted-average basis. This is due to the fact that Table II-1 excludes responses from the May 2020 survey wave, because the authors did not ask about employer intensions during that wave. This underscores that the average desired number of paid days working from home declined somewhat over time, and thus argues for the value shown in Table II-1 as the best estimate for employee preferences. 8 Roger K. Lewis, “Following pandemic, converting office buildings into housing may become new ‘normal,’ Washington Post, April 3, 2021. 9 Please see US Equity Strategy "Rotate Into Growth Stocks, Be Granular In The Selection Of Cyclicals," dated June 14, 2021, available at uses.bcaresearch.com

Overweight The juggernaut trend in the US software & services industry is as strong as ever, and today we are reiterating our overweight call for this large sector. First, within the context of our recent recommendation to rotate into growth, software & services stocks are quintessential growth companies that outperform during periods of a growth slowdown and benefit from rate stabilization. Second, the US private fixed investment in software is going to the moon with the latest print making a 20-year high (top panel). There is no doubt that all this capex will boost both top-line and bottom-line growth. Finally, software & services earnings growth expectation data is also revealing. Sell-side analysts have completely thrown in the towel on software companies with relative forward earnings probing dotcom and GFC era Mariana Trenches (bottom panel). Bottom Line: Secular software & services growth story remains intact and we reiterate our overweight recommendation for this key sector.  

Special Report

Highlights Now that the dust has settled on the hotly contested 2020 election, we introduce our revised and updated quantitative presidential election model. We will periodically update the model as a gauge of President Biden’s political capital as well as the Democratic Party’s evolving odds of keeping the White House in 2024. The model measures the probability of the ruling party’s winning the Electoral College vote for each of the 50 states. As of now, the Democrats only have a 53% chance. Granting that Republicans have a good chance of retaking at least one chamber of Congress in the 2022 midterm election, investors likely face a return to gridlock. Gridlock would mean neither too much nor too little spending and zero tax hikes. The Democratic Party’s success on its current legislative agenda in 2021-22 is highly significant as it will set US fiscal policy for the foreseeable future. Democrats are still highly likely to pass an infrastructure bill by year’s end that will hike corporate taxes and mark peak stimulus for this cycle. Stay long the BCA Infrastructure Basket. Feature The 2020 US Presidential Election has come and gone. Joe Biden defeated Donald Trump with a margin of 74 Electoral College votes to become the 46th president of the United States of America. 57 of these votes came from states where Biden’s margin of victory over Trump hovered around one percentage point or less, highlighting how close the race for the White House was. In this report – for your Independence Day reading pleasure – we introduce the US Political Strategy quantitative presidential election model. Sadly it is never too soon to gear up for the next US presidential election. Our election model is a state-by-state model that uses both economic and political variables to predict the probability of the incumbent party winning the Electoral College votes in each of the 50 states.1 We favor predicting the Electoral College vote over the popular vote since the winner of the presidential election is determined by the Electoral College. There have been five cases in history where the nationwide popular vote did not determine the outcome and two in recent history (George W. Bush in 2000 and Donald Trump in 2016). The college imposes a significant (and deliberate) constraint on majority opinion if it is not shared across America’s geographic regions. The model’s sample size includes ten presidential elections, from 1984-2020, across 50 states, netting 500 observations. The model incorporates the lessons of the narrow 2020 election which took place amid extreme political polarization and an economic recession. The Four Variables Our election model is based off a Probit regression that produces the probability that each state will remain under the control of the incumbent party. The dependent variable (classified as “elected”) is stated as follows: 1 = Incumbent party wins Electoral College votes in state; or, 0 = Incumbent party loses the Electoral College votes in state. This method allows us to measure the probability that a state with certain characteristics will fall into one of these two categories. We can then predict the probability of the incumbent party winning all the Electoral College votes in each of the 50 states. The model has four independent variables, or predictors: State economic health. Specifically, we use the Federal Reserve Bank of Philadelphia State Coincident Index for each of the 50 states. The coincident index combines four of a given state’s economic indicators to summarize current economic conditions in a single statistic. The four indicators are nonfarm payroll employment; average hours worked in manufacturing by production workers; the unemployment rate; and wage and salary disbursements plus proprietors' income deflated by the consumer price index (US city average). In other words, it captures job growth, manufacturing wages, joblessness, and real household income. Margin of victory in previous election. Specifically, we use the incumbent party’s margin of victory in the previous presidential election in each state. A “time for change” variable. This is a categorical variable indicating whether the incumbent party has occupied the White House for one or more terms. Since Biden is serving his first term as president this variable will have no impact on our model’s predictions for the 2024 election. If the Democratic Party were to win the 2024 election and hold the White House for a second term, this variable would then have a negative impact on the party’s odds of winning a third straight term in 2028. Presidential approval rating. Namely, we use the average approval level of the incumbent president in July of an election year. Biden Would Still Win The Election Today Our election model gives us an early look into the 2024 presidential race. We can also look back to see if Biden would win the 2020 presidential election if it were held again today. As it stands, Biden would still win with 308 Electoral College votes (Chart 1), two more than the official account of last year’s election. The two additional votes are a result of the model suggesting Florida (29 votes) would turn Democratic, while Arizona (11 votes) and Georgia (16 votes) would turn Republican, opposite to the 2020 election outcome. Chart 1Quant Model Gives Democrats Only 53% Chance Of Retaining The White House Biden’s overall probability of an election win lies at 53%, in line with early market predictions (Chart 2). These odds reinforce the fact that the 2020 election was closely fought, that the US public remains nearly evenly divided, and that national economic conditions contribute to this division. While it is still early days in the 2024 election cycle, there are some interesting takeaways from our model’s latest prediction. For starters, Florida remains a toss-up state but leans toward the Democrats. Philadelphia and Wisconsin, which were hotly contested in 2020, are only just favored to remain Democratic. Another interesting prediction concerns Arizona and Georgia. Both states were highly contested battlegrounds. For Arizona, it was the first time since the 1996 presidential election that the state turned Democratic; for Georgia it was the first time since 1992. Both states saw larger turnouts for Democrats than in recent elections. However, both states would flip back to Republican control if the election were held today, according to our model, by a more than 10 percentage point change in probability. This is an interesting prediction given that only seven months have passed since the 2020 election. Chart 2Market Has Democrats Ahead Of Republicans The stage was largely set for a Trump loss in 2020. Recessions are catastrophic for presidents running for reelection, especially if they take place during the election year. Coupled with a nationwide health pandemic, Trump was highly likely to lose. In fact his race with Biden proved a lot closer than many commentators expected – in large part due to his unwavering base of support, as reflected in the unprecedentedly small range of his approval rating. This is what prompted us to upgrade his odds from 35% to 45% in a BCA Geopolitical Strategy report on October 26, 2020 (for further discussion see Statistical Appendix). By contrast, Democrats are heavily favored to keep the White House in the 2024 cycle as they will ride the coattails of a recovering US economy, an increasingly vaccinated population, and a (likely) divided Republican opposition. US Still At Peak Polarization Our model produces a novel measure of US political polarization: it shows how many states will be won or lost with extreme certainty (less than 5% or greater than 95%). These are states that are not really competitive because of overwhelming partisan favoritism among their voting populations. Results of in-sample predictions from our model show a slight uptick in the degree of polarization in 2024, which is now above both 2012 and 2020 levels (Chart 3, Top Panel). This change is intuitive coming off the back of one of the most highly contested US elections in history. However, polarization should not rise much higher in the 2024 presidential election cycle. In better economic times, polarization tends to fall, as wider prosperity tends to blanket nationwide social grievances. If Trump wins the Republican nomination in 2024 then one would assume that polarization will remain near peak levels. But if the economy has improved substantially, as we expect, then Trump’s populist platform will have less appeal for voters and the Republican Party will remain divided. This would lead to a higher level of Republican approval of the Democratic candidate, i.e. falling polarization (Chart 3, bottom panel). Chart 3Still At Peak Polarization   Over the next five-to-ten years, we hold the contrarian view that polarization will fall. Generational change in the US will produce more domestic policy consensus, specifically on government spending and taxes, while geopolitical struggle with China will unify the nation against a common enemy for the first time since the Cold War. But our quantitative model pushes against this view at present. Accuracy In Back Tests Our model performs well during in-sample back-testing when comparing it to actual Electoral College vote outcomes for each election since 1984. The model correctly predicts all presidential election outcomes over our sample period (Chart 4), including last year’s narrow result. Chart 4Our Model Predicts All Election Outcomes In Our Sample … The model performs well in out-of-sample back testing too, with prediction accuracy of states at 92%. All election outcomes from 2000-2020 are correctly predicted (Chart 5).  Chart 5… And During Out-Of-Sample Back Testing What Now? We are still a long way from the next presidential election, but the cycle has begun. This means we can begin to form an early view of what is to come over the next three and a half years. The model also gives us a look into what the election backdrop looks like just seven months after the 2020 election. Right now, the Democratic Party holds a decent margin over whoever the Republican competitor may be in 2024. Our model suggests the Democrats would win 308 Electoral College votes if a presidential election were run today, as mentioned. Overall, they have a 53% chance of victory. From a qualitative point of view, our model may be understating the Democrats’ odds in 2024, as things stand today. First, the surest rule of thumb in US politics is that voters will ask themselves whether they are better off than they were four years ago. It is unlikely that voters will be worse off in November 2024 than they were amid the pandemic, recession, and nationwide racial and social unrest of November 2020. Second, the split within the Republican Party over President Trump’s populism, symbolized by marginal Republican votes to convict him of incitement of insurrection over the January 6 riot on Capitol Hill, is likely to produce a closely fought Republican primary election or even a third party candidate, dividing the Republican vote. That’s not to say Republicans have zero chance. Republicans are likely to retake the House of Representatives in 2022, which will give them a base to mount a challenge over the succeeding two years. President Biden will be about to turn 82 years old when the 2024 vote is held – he may choose or be forced to hand the reins to Vice President Kamala Harris, who did not perform well in the 2020 Democratic primary election. Exogenous shocks could take the world by surprise and undermine the “return to normalcy” that the Democrats are trying to project. There are also some interesting toss-up states in 2024, but these will change as we continue to update our model with the latest data. If Biden has to step down, and the Republicans reunify, then the US could see another closely fought election. But Republican reunification is a stretch as things stand today. For now, Biden’s reelection bid will benefit from the recovery and Republican divisions. Investment Takeaways Our quantitative election model gauges the probability that the incumbent political party will retain the White House in the Electoral College vote. The model is based on state-level economic health, the president’s job approval rating, and the strength of his margin of victory in each state, plus an “incumbent advantage” for parties that have only held the White House for one term. The model currently shows that the Democratic Party would win if the 2024 election were held today, albeit with only a 53% probability – an indication of how nearly evenly divided the states remain after the hotly contested election of 2020. However, the model is likely underrating the Democrats as the economy will improve substantially between now and 2024. This will increase the odds of Democrats retaining critical swing states. It will also prolong Republican divisions by depriving them of an economic message around which to rally. But of course anything can happen over three and a half years. The Democrats are favored in 2024 notwithstanding the subjective 75% chance that Republicans retake the House of Representatives in the 2022 midterm elections. A new party in the White House almost always loses seats in Congress at its first midterm. While 2022 could be an exception, we still favor Republicans to regain the House. The takeaway from all of the above is that while 2022 will produce gridlock, nevertheless the 2024 election is unlikely to resolve it. Hence the US will see no drastic domestic legislative changes after 2021-22 period – fiscal policy will be frozen. This provides certainty for investors as it means neither excessive spending, nor austerity, nor tax hikes. Yet midterm elections that produce gridlock exhibit a “buy the rumor, sell the news” profile and are not more bullish for markets than those that produce single-party rule (Chart 6). Monetary policy will probably tighten in 2023 so everything will depend on where the market stands before the election. Incidentally, the model suggests that US political polarization, which hit extreme levels in 2020, will increase further in the 2024 cycle. But this result may not pan out. Over the long run as generational change and geopolitical conflict will force Americans to gather around a new consensus on key policies, namely government spending and foreign and trade policy. Still, we recognize that this reduction in polarization may not occur substantially by 2024 – and on a deeper level that US politics will always be very partisan, as they have been since the presidential election of 1800. Investors should stay constructive on the bull market in the second half of the year as President Biden’s infrastructure bill and/or American Jobs Plan is likely to pass Congress. However, passage in the Senate will mark the top of this cycle’s fiscal stimulus and investors should no longer underweight defensive sectors and growth stocks going forward. Chart 6Gridlock 2022 Will Give Investors Fiscal Certainty   Guy Russell Research Analyst guyr@bcaresearch.com   Statistical Appendix Some clients may be curious as to how our US Political Strategy election model differs from our Geopolitical Strategy model used in the 2020 elections, and where it has made improvements in its efficiency and predictive accuracy. We discuss these improvements herein. Changes To The Geopolitical Strategy Presidential Election Model The last update to the BCA Geopolitical Strategy presidential election model was published at the end of October 2020. We correctly forecast that Biden would win the election in March 2020 and maintained this view throughout the year. By October, however, our quantitative model gave President Trump a 51% chance of winning, predicting that he would gain 279 electoral college votes. We read the model as “too close to call” and stuck with our subjective judgement in favor of Biden for the final prediction, a testament to the need for both quantitative and qualitative analysis. The model missed four states: Arizona, Georgia, Michigan, and New Hampshire. The popular margin of victory in these states was 0.3%, 0.2%, 2.8%, and 8.4% respectively. We knew our model might be over-generous to Trump because we chose to use the range rather than the level of his popular approval rating as a key variable in the model. We did this to counteract the effect of “shy Trump voters,” which distorted traditional public opinion polling.2 Methodology And Variables For the most part, we retain the methodology and suite of economic and political variables used in previous versions of the model. For long-time clients and those who are new to the US Political Strategy and Geopolitical Strategy service, the original version of our model can be found here while the updated 2020 version can be found here.3 The one and only economic variable is now transformed by a six-month change to each state’s coincident index, capturing the improvement or deterioration of the state’s economy. The six-month change results in the best statistical fit for the overall model this time round. In the 2020 model, we transformed the variable by a three-month change. A fast-changing economic environment coupled with a then-higher statistical impact in our model led us to this decision. We still weight the transformation of our economic variable in the same manner as we did in last year’s updated model. We take a weighted average of the six-month change of all the monthly state coincident indices in the presidential term preceding the election. Later months are weighted heavier than earlier months as the most recent context will have a greater impact on voter opinion in the election. In terms of our political variables, the margin of victory is simply measured as the incumbent party’s share of the popular vote minus the non-incumbent party’s vote share. This has not changed from previous versions of our model. For the 2024 model, we have switched back to including the average job approval level instead of range. We use the level as of July of the election year.4 July job approval data shows the highest correlation with the popular and Electoral College vote. October is marginally higher but not enough higher to justify losing three-months of data lead time in our estimation (Chart A1). Obviously whenever we update the model for predictive purposes ahead of November 2024, the latest month’s approval rating serves as a proxy for the final July 2024 reading. Chart A1July Job Approval Highly Correlated With Election Outcome Model Performance Predicted Error The 2024 model has made noteworthy improvements in predictive accuracy across recent elections when compared to the 2020 model. Most noticeable is the large difference in error (Chart A2). The 2020 model failed by a small margin to predict the election outcome. The 2024 model accurately predicts last year’s outcome, although it overpredicts the outcome by 27 Electoral College votes. Chart A2New Model Reduces Predicted Error Over Old Model … The 2024 model also performs well against a different version of the 2020 model, a “bare bones” version that relied exclusively on economic data. This version excluded Trump’s approval data, relying only on an economic explanatory variable to explain the variation in the model’s evolving prediction over time. Our last update to this bare bones model predicted a Trump loss, hence the low prediction error (Chart A3). We published this result alongside our official 2020 model (and other alternatives) for the sake of transparency and to enable clients to choose which of our models better suited their assumptions over ours. We still believe the incumbent president’s job approval data plays a significant role in the presidential election, which is why we included this variable in the GPS and USPS models. But the bare bones model was especially powerful given the economic backdrop in the US last year. Now that the US economy is showing increasing signs of making a full recovery, our 2024 model has learnt from past data and modeling, and still manages to predict 2020’s election outcome despite its inclusion of non-economic (i.e. political) variables. Chart A3… And Performs Well Against “Bare Bones” Economic Model If we create a new bare bones 2024 model and compare it to a comparable 2020 model we arrive at essentially the same outcome (Chart A4). These are two pure economic models, but the new version has a different (smoother) transformation applied to the coincident economic index. That is, changes in economic activity are less volatile. The older version under-predicted the 2020 election outcome by two crucial Electoral College votes, while the new one over-predicted the outcome by 16 votes. Chart A4New “Bare Bones” Economic Model However, for our official 2024 model we will not take this bare bones economic approach but rather will incorporate hard political data (presidential approval, state margin of victory, and a time for change variable). Minimizing predictive error while retaining an explanatory variable that we believe is causal provides us with the most robust model. Classification The 2024 model correctly classifies predicted outcomes at a rate of exactly 90%. That is, when the model makes a prediction of a certain state’s electoral outcome from 1984-2020, it is correct 90% of the time. This level of classification is the highest we have achieved across the several versions we have published since 2016 (Table 1). A close second is the bare bones 2020 model, at 89.11%. Table 1New Model Classifies Outcomes At The Highest Rate … Sensitivity And Specificity – Receiver Operating Characteristic Curve A Receiver Operating Characteristic (ROC) curve is a performance measurement for classification problems of binary modelled outcomes, among others. An ROC curve tells us how much the model is capable of distinguishing between classes. In our case, we have two classes: the dependent variable (classified as “elected”) is stated as 1 = incumbent party wins the Electoral College votes in each state; or 0 = incumbent party does not win the Electoral College votes in each state. The higher the area under the curve (AUC), the better our model is at predicting 0 classes as 0 and 1 classes as 1. An excellent model has AUC near to one. A poor model has an AUC near to zero, which means it has the worst measure of classifying classes correctly, labelling zeros as ones and vice versa. In fact, at a level of zero AUC, the model is reciprocating incorrect classes by predicting zeros as ones and ones as zeros. Statistically, more AUC means that the model is identifying more true positives while minimizing the number/percent of false positives. The ROC curve for our 2024 model has an AUC of 0.9668 (Chart A5), the highest AUC of all models we have developed and tested (Table 2). This means that the true positive rate for classifying outcomes is high and the false positive rate is low, further bolstering the model’s robustness. Chart A5Receiver Operating Characteristic Curve Of New Model Table 2… Has The Best Fit Compared To Older Models … F1 Scores A final grading of the 2024 model is by means of the F1 score. The F1 score is a measurement that considers both precision (specificity in the above ROC curve) and recall (sensitivity in the above ROC curve) to compute the score. The F1 score can be interpreted as a weighted average of the precision and recall values, where an F1 score reaches its best value at 1 and worst value at 0. The 2024 model produces the highest F1 score across our suite of historic models (Table 3). Table 3… And Is The Most Accurate Across All Models After discussing the above statistical metrics and elements of the 2024 model, we are happy to accept it as our new base case presidential election model, premised on its improvement in accuracy at predicting election outcomes in the past, as well as its ability to correctly classify outcomes as they were realized, relative to past published models of this nature. Appendix Tables Table A1USPS Trade Table Table A2Political Risk Matrix Table A3Political Capital Index Table A4APolitical Capital: White House And Congress Table A4BPolitical Capital: Household And Business Sentiment Table A4CPolitical Capital: The Economy And Markets Footnotes 1     We assume that the District of Columbia will vote for the Democratic candidate due to past voting outcomes overwhelmingly favoring Democrats. 2     Large numbers of people polled in the 2016 and 2020 elections declined to say they were voting for President Trump, who was stigmatized in the mainstream media and society at large, or refused to participate in opinion polling. While some analysts rejected this idea after the 2016 election, the large polling misses in 2020 revived it. As many as one-fifth of Trump voters in 2020 might have kept their support secret. See Gregory Korte, “‘Shy Trump Voters’ Re-Emerge As Explanation For Pollsters’ Miss,” Bloomberg, November 19, 2020, bloomberg.com. See also Ed Kilgore, “What Did We Learn About Political Polling In 2020?” New York Magazine, March 26, 2021, nymag.com. 3    From here on out, the updated 2020 Geopolitical Strategy model will be referred to as the “2020 model”. 4    We we had originally introduced four measures covering this topic back in 2019, two require a longer period of job approval data to be put into estimation, these being the  “October momentum” and the “2-year change” job approval variables. We will revisit additional job approval measures and determine if they should be included in later estimations.