Global

Highlights Global Inflation: The worst of the 2020 collapse in global inflation is over; economic growth is starting to rebound, monetary and fiscal policies are highly stimulative, commodity prices are rising and the US dollar is losing some steam. This boosts the investment case for developed market inflation-linked bonds, which appear cheap on our models on a breakeven basis versus nominal government debt. Inflation-Linked Bonds: Starting this week, we are permanently adding inflation-linked bonds as a “discretionary” allocation option in our model bond portfolio framework. We begin with allocations to linkers in the US, Italy and Canada. Tactical Overlay 2.0: We are introducing our revamped Tactical Trade Overlay, using specific securities to implement shorter-term trade ideas in a practical fashion. This week, we begin by initiating inflation-linked bond breakeven trades in the US, Italy and Canada. Feature Chart of the WeekThe Early Days Of An Inflation Expectations Revival With global growth now showing signs of rebounding from the COVID-19 recession as lockdown restrictions ease, inflation expectations in the major developed economies have started to drift upward. Higher inflation breakevens have helped stabilized nominal government bond yields in the majority of countries, even with the latest reads on realized inflation still showing few signs of life (Chart of the Week). In our view, it is still far too soon for bond investors to shift to a below-benchmark stance on overall duration exposure. The threat of a new set of COVID-19 lockdowns is growing, given surging numbers of new infections across much of the southern US and in major emerging economies like Brazil and India. The social and political instability in the US, with elections less than five months away, raises the risk of a renewed flare-up of negative headline risk that can upset overheated equity and credit markets. Amidst all that uncertainty, policymakers worldwide will continue to use aggressive monetary and fiscal stimulus to fight off the risk of an extended recession. That means there is little risk of a big surge in global bond yields from a hawkish repricing of central bank policy expectations over at least the next 6-12 months. At the same time, the extraordinarily loose policy settings, combined with the continued rebound in global commodity prices (most notably, oil), should allow inflation expectations to continue drifting higher. While this will likely also push nominal bond yields higher as well, positioning for wider inflation breakevens remains the “cleaner” way to position for the initial impact of policy reflation. In a report published back on April 28, we introduced a series of valuation models for inflation-linked bonds in the developed economies.1 These models showed that the historic collapse in global oil prices earlier this year, combined with the deflationary impulse from the deep global COVID-19 recession, pushed breakeven inflation rates to levels well below fair value in most countries. Positioning for wider inflation breakevens remains the “cleaner” way to position for the initial impact of policy reflation.  This week, we take the output from our inflation breakeven models to determine specific inflation-linked trade recommendations over both strategic (6-12 months) and tactical (0-6 months) time horizons. For the former, we are adding inflation-linked bonds as an allocation option for all countries in our model bond portfolio. For the latter, we are reviving our Tactical Trade Overlay by introducing some specific trade recommendations using actual inflation-linked bonds in the US, Europe and Canada. Why Global Inflation Expectations Have Bottomed The recent pickup in global market-based inflation expectations has occurred even as actual realized headline inflation rates have fallen dangerously close to 0% in the US, euro area and the UK (Chart 2). Canada is now in outright deflation, with the year-over-year rate of headline CPI inflation falling to -0.4% in May. The decline is not fully attributable to the earlier collapse in oil prices, as core inflation rates have also fallen across the developed world. Chart 2A Threat Of Realized Deflation Despite the plunge in realized inflation, inflation expectations have moved higher for both market-based indicators like inflation breakevens and survey-based measures as well. Chart 3Inflation Expectations Improving Everywhere …. Chart 4… Even Within Europe The German ZEW economic research institute - well known for their surveys of economic forecasters for Germany and the major developed countries - also produces inflation expectations surveys for the same countries. In Charts 3 & 4, we show those ZEW inflation expectations measures alongside the breakeven inflation rates for 10-year government bonds in the US, UK, Japan and the euro area including country-level data for Germany, France and Italy. It is clear that the upturn in breakevens has also occurred as a growing number of economic forecasters have started to anticipate a move higher in both economic growth and inflation over the next year. With recent economic data surprising to the upside in the US, China and in much of Europe, a more optimistic view on global growth is a logical reason helping explain why inflation expectations have been drifting higher. Even more so has been a shift in the deflationary momentum stemming from a rising US dollar and falling commodity prices – trends that are in the process of reversing. Perhaps the strongest deflationary force over the past couple of years has been the persistent strength of the US dollar. World export prices have been contracting on a year-over-year basis since December 2018, which has coincided with a similar period of positive annual growth in the trade-weighted US dollar since June 2018 (Chart 5). While the dollar is still at elevated levels, its momentum has started to roll over (middle panel), suggesting less deflationary pressure from the currency. The same can be said for commodity prices, which reflect both the global demand story and the trend in the US dollar as well, given that important industrial commodities like oil and copper are priced in US dollars. With the prices of those commodities off their lows, the annual growth rates of the CRB Energy and Metals indices have bottomed out, implying less global deflationary pressure from commodities (bottom panel). A reflationary boost to the global economy – and to inflation expectations – from a softer dollar is likely over the next 6-12 months. Looking ahead, the US dollar is likely to continue losing strength for two reasons: less-supportive interest rate differentials and improving global growth (Chart 6). The Fed’s aggressive interest rate cuts over the past year have eliminated much of the attractive carry that helped fuel the dollar’s rise over the past few years. At the same time, the US dollar remains an “anti-growth” currency that tends to weaken during periods of improving global growth, and vice versa. Chart 5Easing Of Disinflationary Pressures From The USD & Commodities Chart 6A Softer USD Will Help Lift Global Inflation Expectations With global growth starting to emerge from the COVID-19 recession, the US dollar is now more exposed to less attractive interest rate differentials. This suggests that a reflationary boost to the global economy – and to inflation expectations – from a softer dollar is likely over the next 6-12 months. Chart 7Rising Oil Prices Will Help Lift Global Inflation Expectations The same can be said for commodity prices like oil, which have considerable upside as global growth improves. Our colleagues at BCA Research Commodity & Energy Strategy are quite bullish on the outlook for oil over the next 12-18 months, given the improved demand/supply balance and aggressive global monetary and fiscal stimulus. Their expect the Brent benchmark to rise to $46/bbl by the end of 2020 and $73/bbl by the end of 2021 – levels that would push inflation expectations in the US and other major developed markets higher given the usual strong correlation between oil and breakevens (Chart 7).2 Summing it all up, the trends that have helped stabilize and lift global inflation expectations look set to continue over the next 6-12 months. Bottom Line: The worst of the 2020 collapse in global inflation is over; economic growth is starting to rebound, monetary and fiscal policies are highly stimulative, commodity prices are rising and the US dollar is losing some steam. This boosts the investment case for developed market inflation-linked bonds, which appear cheap on our models on a breakeven basis versus nominal government debt. Adding Inflation-Linked Bonds To Our Model Bond Portfolio Our model bond portfolio framework is how we translate our main global fixed income strategic themes into actual investment recommendations. We apply specific weightings to government bond and spread product allocations within a fully invested hypothetical portfolio with a custom benchmark index (which is essentially the Bloomberg Barclays Global Aggregate with additional allocations to high-yield and emerging market corporates). We had not included inflation-linked bonds in the model portfolio, as we have always maintained a focus on the larger and more liquid parts of the developed market fixed income universe. We chose to express views on inflation expectations through duration or yield curve positioning, under the assumption that wider breakevens correlate to higher bond yields and/or steeper yield curves. Chart 8Global Inflation Breakevens Are Too Low We now are of the view that inflation-linked bonds should be included in our model portfolio investment universe, but on an “opportunistic” basis. In other words, we are not adding linkers to the custom benchmark index. Instead, we will be using potential allocations to inflation-linked bonds as another way to play for periods of rising inflation expectations beyond recommended duration and curve tilts in the model portfolio – particularly now that we have valuation models for inflation breakevens in almost all countries in the portfolio (the US, UK, Japan, Germany, Italy, France, Canada and Australia). Based on the output of our fundamental fair value framework for 10-year inflation breakevens, inflation protection looks “cheap” in all countries where we have valuation models except the UK (Chart 8). Charts with the details of each country’s 10-year inflation breakeven model can be found in the Appendix on pages 11-14. The inputs to the model are the same for each country: a) the 5-year moving average of headline CPI, representing the medium-term trend that anchors inflation expectations; and b) the annual percentage change in the Brent oil price in local currency terms, which creates deviations from the trend to account for moves in oil and currencies. For all countries excluding the UK, breakevens are below fair value because of the collapse in oil prices earlier this year. Inflation protection looks “cheap” in all countries where we have valuation models except the UK. The UK is the one market that does not appear cheap in our framework, with breakevens very close to both fair value and the medium-term trend in realized inflation. Those relatively high breakevens are also a reflection of the very low real bond yields for UK index-linked Gilts. Chart 9Linkers Offer Better Value In The US & Euro Area Than The UK For the past several years, UK real yields have traded well below measures of equilibrium real interest rates like the New York Fed’s estimates of “r-star”. This differs from real yields for US TIPS or French OATis, which trade roughly in line with the r-star estimates for the US and euro area (Chart 9). We suspect that is because of the chronic demand/supply mismatch for UK inflation-linked bonds, which are always in high demand from UK pension funds who need real assets for asset/liability management and regulatory purposes. So based on the output from the fair value models, inflation-linked bonds look most attractive on a breakeven basis in Italy, Canada, the US, Japan, Germany and France. From this list, we are choosing to add recommended positions in the US, Italy and Canada only. For Germany and France, we are already very underweight both countries in the model portfolio, so it is difficult to make a meaningful switch out of nominal bonds into linkers. For Japan, the Bank of Japan’s Yield Curve Control policy, which caps the level of 10-year bond yields near 0%, makes us reluctant to recommend any breakeven widening positions. The changes to the model bond portfolio can be found in the tables on pages 15-16. Bottom Line: Starting this week, we are permanently adding inflation-linked bonds as a “discretionary” allocation option in our model bond portfolio framework. We begin with allocations to linkers in the US, Italy and Canada. Tactical Trade Overlay 2.0, Starting With Inflation-Linked Bonds This week, we are introducing a remodeled version of our Tactical Trade Overlay, which we put on hiatus a few months ago because of “mission creep”. Many of our recommendations were being held too long to be truly considered tactical, or short-term, in nature, thus defying the original purpose of the Overlay. This week, we are introducing a remodeled version of our Tactical Trade Overlay, which we put on hiatus a few months ago because of “mission creep”. All trades in the new Overlay will have a shorter term investment horizon of six months or less. All recommended trades will be implemented with specific securities, rather than just using generic Bloomberg tickers or bond indices. This will allow for a more transparent process where clients can “follow along” with the performance of our trades. Chart 10Inflation-Linked Bonds Have A Duration To Real Yields, Unlike Nominals To begin, we are putting three inflation-linked bond trades into our new Tactical Trade Overlay, positioning for wider 10-year breakevens in the US, Italy and Canada. All trades will be implemented using a long position in an inflation-linked bond and a short position in the government bond futures contract for each country. We are using futures rather than a short position in a cash government bond for the sake of simplicity, both for implementing the trade and measuring returns. The new trades will be implemented on a duration-matched basis. This means only selling enough of the 10-year bond futures to hedge against any directional move in the yield of the long 10-year inflation-linked bond. A straight comparison of the duration of linkers to futures cannot be made, since inflation-linked bonds have a duration to real yields while futures (and cash government bonds) have a duration to nominal yields. The durations for inflation-linked bonds are always higher than those of nominals (Chart 10), thus the index-linked durations must be adjusted by the beta of changes in real yields to changes in nominal bond yields. To determine the correct duration adjustment, we use betas taken from rolling three-year regressions of monthly changes of 10-year inflation-linked yields on changes in 10-year nominal government yields, using generic Bloomberg tickers. The common convention is to simply apply a yield beta of 0.5 for all inflation-linked bonds (this is the default setting on Bloomberg valuation tools). We think having a variable yield beta is a more accurate way to hedge out the directional risk in each trade from shifts in real bond yields. Chart 11Yield Betas For Inflation-Linked Bonds Vary Across Countries The current yield betas for all eight countries where we have inflation breakeven fair value models are shown in Chart 11 – it is clear from the chart that using a constant yield beta of 0.5 across countries is not accurate, as they vary widely across countries. Multiplying the duration of the actual inflation-linked bond used in our breakeven trades by our rolling yield beta creates a “nominal” duration measure that can then be compared to the duration on the short leg of the breakeven trade. For futures, we use the empirical duration estimates from Bloomberg using the “FRSK” function. The ratio of the beta-adjusted linker duration to the empirical duration of the bond futures creates the hedge ratio that we will use when measuring the returns of this now “risk-matched” breakeven trade. The actual bonds, futures contracts and hedge ratios for all of our new breakeven trades can all be found in the table on page 18, with initial entry prices for all securities. We will begin to monitor the trade returns in next week’s report. Bottom Line: We are reviving our Tactical Trade Overlay with inflation-linked bond breakeven trades in the US, Italy and Canada. Appendix: 10-Year Inflation Break Even Model Chart 12Our US 10-Year Inflation Breakeven Model Chart 13Our UK 10-Year Inflation Breakeven Model Chart 14Our France 10-Year Inflation Breakeven Model Chart 15Our Italy 10-Year Inflation Breakeven Model Chart 16Our Japan 10-Year Inflation Breakeven Model Chart 17Our Germany 10-Year Inflation Breakeven Model Chart 18Our Canada 10-Year Inflation Breakeven Model Chart 19Our Australia 10-Year Inflation Breakeven Model   Robert Robis, CFA Chief Fixed Income Strategist rrobis@bcaresearch.com Footnotes 1 Please see BCA Research Global Fixed Income Strategy Weekly Report, "Global Inflation Expectations Are Too Low", dated April 28, 2020, available at gfis.bcaresearch.com. 2 Please see BCA Research Commodity & Energy Strategy Weekly Report, "Low Vol, High Uncertainty Keeps Oil-Price Rally On Tenterhooks", dated June 18, 2020, available at ces.bcaresearch.com. Recommendations The GFIS Recommended Portfolio Vs. The Custom Benchmark Index Duration Regional Allocation Spread Product Tactical Trades Yields & Returns Global Bond Yields Historical Returns

Special Report

Highlights We test the out-of-sample performance of Mean-Variance Optimization (MVO) portfolios. We find that MVO portfolios based on historical estimates have historically underperformed static-weight portfolios. However, MVO portfolios using momentum-based return estimates and a “shrunk” correlation matrix dramatically improved performance, allowing them to outperform our benchmarks. Additionally, MVO portfolios where the asset weights could not deviate more than 10% from the benchmark still added value, making them a more attractive option for benchmarked investors. Feature “Let every man divide his money into three parts, and invest a third in land, a third in business, and a third let him keep in reserve” – Rabbi Isaac bar Aha, Babylonian Talmud, 200 C.E.   Diversification is the main pillar of asset allocation. Its advantages are clear: By spreading out funds to different assets, a portfolio can remain resilient to the failings of individual investments, provided they do not occur at the same time. But, while it is universally accepted that “you should not put all your eggs in one basket”,1 putting this concept into practice remains challenging: After all, how exactly do you diversify? The first attempt to answer this question dates back to the Babylonian Talmud, which suggested that an equally weighted portfolio between a growth asset, a real asset, and a safe asset was the best way to allocate one’s funds. After that, it took almost 2000 years for the first formal theory on portfolio construction to emerge, with the derivation of the mean-variance solution by Harry Markowitz in the 1950s. Today, while other methods such as risk parity, factor-based diversification, and the Kelly criterion have also emerged, mean-variance optimization remains the standard theoretical framework for portfolio construction. However, mean-variance optimization (MVO) has not proven to be the final solution to the asset allocation puzzle. Far from it. With time, investors have realized that, while MVO might provide the optimal allocation in theory, it has several drawbacks when used to construct real world portfolios. Specifically, MVO portfolios have become notorious due to their inability to deal with estimation error, as well as for their large concentrated positions on a single asset. In this report we examine the strengths and weaknesses of MVO portfolios from an empirical standpoint, and we suggest three practical solutions to solve some of the problems described above. The report is structured as follows: We first describe our data and methodology in our Methodology section. We then provide a summary of our main findings in the Summary Of Results section. For readers who wish to read a more detailed description of our analysis, please refer the Results In Full section. Methodology Data In order to build optimized portfolios, we use monthly returns for seven assets from the perspective of a US-based investor: US equities, international equities (developed markets ex-US), US Treasurys, US investment-grade corporate bonds, commodity futures, US REITs, and US cash. We use a sample starting in 1973 and ending in April 2020.2 Optimization We build three types of mean-variance optimized portfolios: A conservative portfolio, with a target volatility of 6%, a moderate portfolio, with a target volatility of 9%, and an aggressive portfolio, with a target volatility of 12%.3 In the optimization procedure, each portfolio seeks to maximize returns subject to the risk constraint. In addition to this risk constraint, we also do not allow for leverage or short positions. The optimization and rebalancing are done on a monthly basis. Benchmarking As a benchmark we build three portfolios (an aggressive benchmark, a moderate benchmark, and a conservative benchmark) with constant weights. The portfolios are rebalanced on a monthly basis. Be advised that this is not a completely fair comparison, as these benchmarks are constructed ex-post, which means that in contrast to the MVO portfolios, these benchmark portfolios have the benefit of hindsight to achieve their target volatility. However, they will prove useful to evaluate whether MVO portfolios are doing a good job at maximizing returns. Table 1 describes the benchmark portfolios. Table 1Benchmark Portfolio Weights Turnover limits In order to get a better picture of how MVO performs under realistic conditions, we assess the performance of MVO portfolios with turnover limits. These turnover limits are put in place to take into account that changing portfolio weights by large amounts from month to month is not feasible for most asset managers. We limit the month-to-month change in the weight of each asset to 5%. As an example, if the current weight in US Treasurys is 20%, and the MVO finds that the optimal weight for the following month is 40%, the new weight will only be 25%. In a few months this results in buying and selling not being equal. In those cases, turnover for certain assets might be limited further to ensure that all of the weights add up to 100%. Transaction costs Table 2One-Way Transaction Costs To calculate transaction costs, we multiply the absolute value of the month-on-month change in the weight of each asset by the transaction costs shown in Table 2. Note that this approach does not take into account drift. However, this will not be important when comparing portfolios, given that drift will also not be considered for the benchmarks (since weights stay static from month to month, transaction costs for the benchmarks are zero in our analysis).   Summary Of Results After conducting our analysis, our main findings were the following (please see full details below in the following sections): Historical performance MVO portfolios using historical estimates as inputs have historically underperformed static-weight benchmarks in both raw return and risk-adjusted terms (Table 3). The realized volatility of MVO portfolios was relatively close to their target risk over the sample. However, this was not true when looking at shorter periods. Table 3Summary Of Results Solution #1: Momentum-based return estimates Expected return estimates can be improved by taking into account that prices usually trend (price momentum). These momentum-based return estimates improve both the return and the Sharpe ratio. Additionally, the volatility of MVO portfolios using momentum-based return estimates remains below the target risk more consistently. Solution #2: Shrinking correlation matrix  Historical correlation estimates are often too noisy and lead to estimation error.  “Shrinking” cross-asset correlations to a pre-specified value often results in improvements in both raw returns and risk-adjusted returns.   Very high shrink factors had the best performance. Solution #3: Constraining weights MVO portfolios result in large concentrations in one or two assets. Such an allocation is not practical for most asset managers, particularly if they are benchmarked. However, MVO can improve performance even when asset weights are constrained. Specifically, the information ratio can be improved when using the two solutions described above. We remind clients that MVO should be used prudently. Specifically, even though many of the measurements we have suggested here concentrate in making MVO more robust to noisy estimates, MVO is always vulnerable to the risk of estimation error. Other inputs should be considered when making final asset allocation decisions. Results In Full Historical performance The theory underpinning mean-variance optimizations assumes perfect knowledge of expected returns, volatility, and correlation. In practice, however, this is never the case. Instead, these inputs need to be estimated – a process which unavoidably carries error with it. This error in the estimated inputs can lead to significant deviations from the optimal weights, resulting in lower performance. But to what extent does performance suffer when using imperfect inputs? To answer this question, we construct MVO portfolios as they are often built in practice: using historical estimates. The portfolios are built as follows: Each month we use the historical mean return, historical volatility, and historical correlation matrix up to that point in time, as inputs for every asset with the exception of cash.4 To ensure that the historical inputs are robust enough, we start building the portfolios when we have at least 10 years of historical data (since our data begins in 1973, we start building the portfolios in 1983). Table 4 shows several key metrics for these MVO portfolios with the different risk targets as well as for their benchmarks. Overall, MVO portfolios underperformed their respective benchmarks in every single category regardless of the risk level. This result is in line with previous research, which has shown that MVO usually underperforms equally weighted or static-weight portfolios.5 Table 4MVO Versus Benchmarks The one positive trait of the MVO portfolios was that the volatility of returns over the entire sample stayed relatively close to the risk target. This happened because, over the very long term, volatility remains fairly stable within assets, in stark contrast to returns or correlations, which can undergo dramatic changes even over very long horizons. (Chart 1). However, it is important to point out that, while the volatility of the MVO portfolios stayed relatively close to the risk target throughout the almost 40-year sample, the same was not true if we look at subperiods. Volatility of MVO portfolios was often significantly higher than the risk constraint in the medium term (Chart 2). This is a problem, given that asset managers are usually evaluated in these shorter time frames. Chart 1Volatility Is Relatively Stable Over The Long-Term Chart 2MVO Volatility Can Stray Well Above Target   Solution #1: Momentum-based return estimates Expected returns are the most important input for mean-variance optimization. In general, there are many ways to improve on the dismal track record of historical estimates. However, in this report we will focus on a simple way to improve them. Specifically, we take into account the fact that prices usually trend. As we discussed in our July 2019 report, momentum in asset prices has been one of the most persistent forces in the history of financial markets.6 The propensity of returns to trend, was first discussed by the academic literature in the 1990s but has possibly been known amongst practitioners since the primitive financial markets of the 17th century. We use this simple stylized fact about returns for a couple of reasons: First, there is extensive literature arguing that there are structural forces which causes prices to trend. This gives us some assurance that this tendency is not a random trait of the data, but rather a result of an internal mechanism in the market. Second, the existence of price momentum has been known for a very long time, which means that any sophisticated investor could have realistically used this fact during our sample to improve his or her estimate of expected returns. To take price momentum into account, we estimate expected returns as follows for each asset except for cash: Each month we compute the average return of the asset up to that point, following periods when price was above the 12-month moving average (uptrend average). We also compute the average return of the asset up to that point, following periods when price was below the 12-month moving average (downtrend average). If the asset’s price is above its 12-month moving average, we use the uptrend average as our expected return estimate. If the asset’s price is below its 12-month moving average, we use the downtrend average as our expected return estimate. Much like with our historical estimates, we wait until we have 10 years of data to begin constructing the portfolios. Table 5 shows the result of the MVO portfolios using these momentum-based expected return inputs, as well as the MVO portfolios using historical estimates and the benchmarks. Overall, using momentum-based expected returns significantly improved the performance of all of the MVO portfolios. Additionally, the volatility of the enhanced MVO portfolios was much better behaved, staying much closer to the risk target than the volatility of the MVO that used historical return estimates (Chart 3). Table 5Momentum-Based Return Estimates Enhance Performance Chart 3Momentum-Based Return Estimates Lead To Better Behaved Volatility Solution #2: Shrinking correlation matrix What about improving correlation estimates? Correlations are notoriously noisy and difficult to forecast since they often change depending on the regime. Moreover, while the number of estimates necessary for expected returns and volatility increases linearly with the number of assets in consideration, the number of estimates necessary for the correlation matrix increases at a much higher rate.7 One alternative solution proposed by practitioners to deal with this noise and limit the amount of estimation error is to anchor correlations to some pre-specified value (also known as “shrinking” the correlation matrix). This reduces the noise of the correlation estimate, making for a much more robust input. Shrinking the correlation matrix can be done as follows: Choose shrinkage target: The anchor value or the shrinkage target, is the number to which cross-asset correlations are anchored. Often the average cross-asset correlation between all assets is chosen as the shrinkage target.8 Choose a shrinkage factor: The shrinkage factor is the weight we put on the anchor versus our estimate. The weighting is done with the formula below: Shrunk correlation = (Shrinkage factor) * (Shrinkage target) + (1-Shrinkage factor)* (Correlation estimate) But what exactly should the shrinkage factor be? To answer this question, we test the performance of MVO portfolios that use historical estimates for expected returns and variance, but where the correlation matrix is shrunk. Additionally, we examine how varying the shrinkage factor affected historical performance for different types of investors. Our results indicate that correlation matrices with high levels of shrinkage (70%+ shrink factor) have invariably improved Sharpe ratios and reduced volatility, while also improving returns in almost all cases (Table 6). Table 6High Shrinkage Factors Lead To Better Risk-Adjusted Returns This result was also robust to using other types of expected return estimates. Table 7 shows that shrinking the correlation continued to either maintain or improve performance when using the momentum-based expected return inputs from the section above. Importantly, MVO portfolios using momentum-based returns and shrunk correlation with high shrinkage factors were able to beat the benchmarks at various metrics regardless of the type of portfolio. Table 7Momentum-Based Return Estimates Enhance Performance Solution #3: Constraining weights MVO often results in portfolios that have highly concentrated positions in one or two assets (Chart 4). Additionally, asset weights can experience dramatic changes, even when turnover limits are imposed. For several reasons this is not a desirable trait for most asset managers. Benchmarked managers in particular are often constrained in how much they can deviate from their benchmarks. Moreover, even if they do not have explicit deviation limits, benchmarked managers are often limited implicitly since they are evaluated on their tracking error. Chart 4MVO Weights Are Highly Concentrated One solution to the extreme weights problem is to limit the amount by which the weights can deviate from the benchmark. To test this hypothesis, we perform the optimization as described previously, using momentum-based return estimates and a shrunk correlation matrix, but we limit the amount by which weights can deviate from the benchmark weights by 10%. Table 8A shows the result of MVO portfolios with weight limits, as well the benchmark. Overall, constraining the amount of deviation from the benchmark weights resulted in significant performance improvements across every metric we tested on. Additionally, while the limited MVO portfolios did not always outperform the unconstrained MVO portfolios in terms of return, they were able to have much lower tracking risk and better information ratios, making them a better option for benchmarked investors (Table 8B). Finally, the constrained portfolios had much better-behaved asset weights than the unconstrained ones9 (Chart 5). Table 8APortfolios With Shrunk Correlation (80% Shrinkage Factor) & Momentum-Based Return Estimates Table 8BPortfolios With Shrunk Correlation (80% Shrinkage Factor) & Momentum-Based Return Estimates Chart 5Limited MVO Leads To More Realistic Assets Weights For Benchmarked Investors   Juan Correa Ossa, CFA  Associate Editor juanc@bcaresearch.com Footnotes 1  The origin of this phrase is often attributed to Don Quijote, the main character of the famous 17th century Spanish novel of the same name by Miguel de Cervantes. 2 Our sources are MSCI Inc (Please see copyright declaration), Bloomberg /Barclays Indices, National Association Of Real Estate Investment Trusts and Goldman Sachs via Datastream. 3 Our approach is loosely based on the work from Bessler, Opfer and Wolff. For more details, please see Wolfgang Bessler, Heiko Opfer, and Dominik Woff, “Multi-Asset Portfolio Optimization and Out-of-Sample Performance: An Evaluation of Black-Litterman, Mean-Variance, and Naïve Diversification Approaches”, European Journal of Finance, Vol. 23, no.1, 2017. 4 For the expected return of cash we just use the current cash yield, since this will be the return of cash with certainty. Volatility and correlation are calculated using historical estimates. 5 Please see De Miguel, Garlappi, Uppal, “Optimal Versus Naïve Diversification: How Inefficient is the 1/N Portfolio Strategy?”, The Review of Financial Studies, Vol. 22, no. 5, May 2009. 6 Please see Global Asset Allocation Report “Swimming With The Tide: Momentum Strategies In Financial Markets,” dated July 23 2019, available at gaa.bcaresearch.com 7 Specifically, if you need to build a portfolio of n assets, you need n estimates of expected returns, n estimates of variances but (n*(n-1))/2 estimates of asset correlations. 8 We use this method to compute the shrinkage target in our analysis since we have a balanced variety of assets (two risk assets, two fixed income assets, two real assets and cash). However, be advised that this shrinkage target might not always be appropriate. In general, judgement should be used to choose an appropriate shrinkage target. 9 In general, there might be other benefits to constraining asset weights. A recent paper by Pedersen et al showed that MVO tends to suffer due to taking large exposure in problematic portfolios that arise due to noise. Constraining MVO weights is a solution to control for this noise and keep the optimization from overweighting this problem portfolios. For more details please see Pedersen, Lasse Heje and Babu, Abhilash and Levine, Ari, “Enhanced Portfolio Optimization”, NYU Stern School of Business, 2020.

Despite the strong rally in stocks since mid-March and a looming second wave of the pandemic, we continue to recommend that investors overweight equities on a 12-month horizon. Needless to say, this view has raised some eyebrows. With that in mind, this week we present a Q&A from the perspective of a skeptical reader who does not fully share our enthusiasm. Q: You said last week that a second wave of the pandemic is now your base case, yet you’re still sticking with your positive 12-month equity view. Why? A: A second wave of the pandemic, along with uncertainty about how the coming fiscal cliff in the US will be resolved, could unnerve investors temporarily. Nevertheless, we expect global equities to rise by about 10% from current levels over the next 12 months, handily outperforming bonds. While low interest rates and copious amounts of cash on the sidelines will provide a supportive backdrop for stocks, the main impetus for higher equity prices will be a recovery in economic activity and corporate profits. Q: It is hard to see the economy recovering very much if there is a second wave. A: It is important to get the arrow of causation right. Part of the reason we expect a second wave is because we think policymakers will continue to relax lockdown measures even if, as has already occurred in a number of US states, the infection rate rises. Granted, a second wave will moderate the pace at which containment measures can be dismantled. It will also prompt people to engage in more social distancing. Thus, a second wave would make the economic recovery slower than it otherwise would have been. However, it is doubtful that growth will grind to a halt. The appetite for continued lockdowns has clearly waned. For better or for worse, most western nations will follow the “Swedish model” of trying to limit the spread of the virus without imposing draconian restrictions on society. Chart 1CBO Projects The Unemployment Rate Will Fall Very Slowly Q: Even if the Swedish model works, and I doubt it will, we are still in a very deep economic hole. The unemployment rate in many countries is the highest since the Great Depression. The Congressional Budget Office does not foresee the US unemployment rate falling below 5% until 2028. A return to positive growth seems like a very low bar for success. We may need many years of above-trend growth just to get back to the pre-pandemic level of GDP! A: The Congressional Budget Office is too pessimistic in assuming that the recovery will be as sluggish as the one following the Great Recession (Chart 1). That recovery was weighed down by the need to repair household balance sheets after the bursting of a debt-fueled housing bubble. The current downturn was caused by external forces – an exogenous shock in econospeak. Historically, recoveries following exogenous shocks have tended to be more rapid than recoveries following recessions that were instigated by endogenous problems. Q: That may be so, but Wall Street is already penciling in a very rapid recovery. Last I checked, analysts expect S&P 500 earnings next year to be close to where they were last year. A: One has to be careful when comparing earnings estimates with economic growth projections. Chart 2 shows a breakdown of S&P 500 EPS estimates by sector. Appendix A also shows the evolution of these estimates over time. While analysts expect overall earnings per share (EPS) to return to last year’s levels in 2021, this is mainly because of the resilient profit outlook in the technology and health care sectors (the two biggest sectors in the S&P 500 by market cap). Outside those two sectors, EPS in 2021 is expected to be down 8.6% from 2019 levels, or 11.2% in real terms. Chart 2Breakdown Of S&P 500 EPS Estimates By Sector If one looks at the cyclically-sensitive industrials sector, earnings are projected to fall by 16% between 2019 and 2021. Energy sector earnings are projected to decline by 65%. Earnings in the consumer discretionary sector are expected to decline by 8%, despite the fact that Amazon accounts for nearly half of the sector by market cap.1 This suggests that analysts are expecting more of a U-shaped economic recovery than a V-shaped one. Chart 3The Present Value Of Earnings: A Scenario Analysis Q: Fair enough, but I am ultimately more interested in what the market is pricing in than what analysts are expecting. It seems to me that stock prices have rebounded much more rapidly than one would have anticipated based on the evolution in earnings estimates. A: That is true, but it is important to keep in mind that the fair value of the stock market does not solely depend on the expected path of earnings. It also depends on the discount rate we use to deflate those earnings. For the sake of argument, let us suppose that S&P 500 earnings only manage to reach $144 per share next year (10% below current consensus) and take five years to return to their pre-pandemic trend. All things equal, such a decline in earnings would reduce the present value of stocks by 4.2% relative to what it was at the start of the year (Chart 3). However, all things are not equal. The US 30-year Treasury yield, adjusted for inflation, has declined by 59 basis points this year. If we use this real yield as a proxy for the discount rate, the fair value of the S&P has actually increased by 8.7% since January 1st, despite the decline in earnings. Q: I think you’re doing a bit of a bait and switch here. You’re assuming that earnings estimates return to trend by the middle of the decade, but that long-term bond yields remain broadly unchanged over this period. If the economy and corporate earnings recover, won’t bond yields just go back to where they were last year, if not higher? A: Not necessarily. Conceptually, there is not a one-to-one mapping between interest rates and the full-employment level of aggregate demand.2 For example, consider a case where an adverse economic shock hits the economy, making households and businesses more reluctant to spend. If that were all there was to the story, the stock market would go down. But there is more to the story than that. Suppose the central bank cuts interest rates in response to this shock, which boosts demand by enough to return the economy to full employment. Now we have a new equilibrium where the level of demand – and by extension, the level of corporate profits – is the same as before but interest rates are lower. The fair value of the stock market has gone up! Q: Hold on. Central banks came into this recession with little fire power left. I agree that their actions have helped the stock market, but they have not been enough to rehabilitate the economy. A: Good point. That is where the role of fiscal policy comes in. One of the unsung benefits of lower interest rates is that they have incentivised governments to borrow more at a time when the economy needs all the fiscal support it can get. As Chart 4 shows, the fiscal response during this year’s downturn has been significantly larger than during the Great Recession. Thus, it is more correct to say that the combination of lower interest rates and fiscal easing have conceivably increased the fair value of the stock market. Chart 4Fiscal Stimulus Is Greater Today Than It Was During The Great Recession Q:  And yet despite all this fiscal and monetary support, GDP remains depressed. A: The point of the stimulus was not to raise output or employment. It was to keep households and businesses solvent during a time when their regular flow of income had dried up. Q: If households and businesses did not spend much of that money, where did it go? A: Much of it remains in the banking system. The US savings rate shot up to 33% in April. As Chart 5 illustrates, this was almost perfectly mirrored by the increase in bank deposits. Anyone who claims that savings have nothing to do with deposits should study this chart. Chart 5Lots Of Savings Slushing Around   Chart 6Stocks That Are Popular With Retail Investors Are Outperforming Q: And now, I suppose, these deposits are flowing into the stock market? A: Correct. That is one reason why stocks popular with retail investors have outperformed the S&P 500 by 30% since mid-March (Chart 6). Q: Have these retail flows really been important enough to matter? A: They have probably been more important than widely portrayed. Many of the online brokerages touting zero-commission trades make their money by selling order flow to hedge funds. Thus, the trading of individuals is magnified by the trading of institutional investors. More liquid markets tend to generate higher prices. There is also another subtle multiplier effect worth considering. You mentioned that money was “flowing into the stock market.” Technically speaking, “flow” is not the best word to use. For the most part, if I decide to buy some shares, someone else has to sell me their shares. On a net basis, there is no inflow of cash into the stock market. Rather, what happens is that my buy order lifts the price of the shares by enough to entice someone to sell their shares. Thus, if retail investors bid up the price of stocks to the point that institutions are forced to sell, those institutions are now left with excess cash that they have to deploy elsewhere in the stock market. As the value of investors’ stock portfolios rises, the percentage of their net worth held in cash falls. This game of hot potato only ends when the percentage of cash held by investors shrinks to a level that is consistent with their preferences. Importantly, this means that changes in the amount of cash on the sidelines can have a “multiplier” effect on stock prices. For example, if cash holdings go up by a dollar, and people want to hold ten times as much stock as cash, then stock market capitalization has to go up by ten dollars. Q: How far along are we in this game of hot potato? A: Despite the rally in stocks since mid-March, cash held in money market funds and savings deposits is still 10% higher as a share of market capitalization than at the start of the year. This suggests that the firepower to fuel further increases in the stock market has not been fully spent. Chart 7Equity Risk Premium Is Still Quite High Q: Wouldn’t you think that after a pandemic people would be more risk-averse and hence inclined to hold more cash? A: That would be a logical assumption, but it is not clear whether it is empirically true. There is some evidence from the psychological literature that people who survive life-threatening events tend to become less risk averse rather than more risk averse after the event has passed.3 A pandemic seems to qualify as a life-threatening event. In any case, when considering the equity risk premium, we should not only think about the riskiness of stocks; we should also think about the riskiness of bonds. Bond yields are near record lows. To the extent that yields cannot fall much from current levels, this makes bonds a less attractive hedge against downside economic news than they once were. So perhaps the equity risk premium, which is still quite high, should actually be lower than it currently is (Chart 7). Q: It seems that much of your optimism is based on the assumption that policy will stay stimulative. On the monetary side, that seems like a safe assumption. However, as you yourself mentioned at the outset, there is a risk that stocks will be upended by a premature tightening in fiscal policy. A: This is indeed a risk. In the US, the Paycheck Protection Program (PPP) will run out of funds over the coming month. The additional $600 per week in benefits that jobless workers are receiving will expire on July 31st, causing average unemployment payments to fall by about 60%. Direct payments to households have also ceased. Together, these three fiscal measures amount to about 5.5% of GDP. Furthermore, most states begin their fiscal year on July 1st. Despite receiving $275 billion in federal aid, they are still facing a roughly $250 billion (1.2% of GDP) financing shortfall in the coming fiscal year, which could force widespread layoffs. The good news is that both Republicans and Democrats want to avert this fiscal cliff. While negotiations over the next stimulus package could unnerve investors for a while, they will ultimately culminate in a deal. The Democrats want more spending, as does the White House. And if public opinion polls are to be believed, congressional Republicans will also cave in to voter demands for continued fiscal largess (Table 1). Table 1There Is Much Public Support For Fiscal Stimulus Q: It seems to me that the fiscal cliff is not the only political risk to worry about. Tensions with China are running high and there is domestic unrest in many cities around the world. Even if fiscal policy remains accommodative, President Trump will probably lose in November. This makes a repeal of his tax cuts more likely than not. A: It is true that betting markets now expect Joe Biden to become president (Chart 8). They also expect Democrats to regain control of the Senate. My personal view is that Trump has a better chance of being reelected than implied by betting markets. While the protests have hurt Trump’s favorability ratings in recent weeks, ongoing unrest could help him, given his claim of being the “law and order” president. It is worth recalling that after falling for more than 20 years, the nationwide homicide rate spiked by 23% between 2014 and 2016 following protests in cities such as St. Louis and Baltimore (Chart 9). This arguably helped Trump get elected, just like the Watts Riot in Los Angeles helped Ronald Reagan get elected as Governor of California in 1966. Chart 8Betting Markets Now Expect Joe Biden To Become President If Senator Biden were to prevail, then yes, Trump’s corporate tax cuts would be in jeopardy. A full repeal of the Trump tax cuts would reduce EPS of S&P 500 companies by about 12%. Chart 9Continued Unrest May Help Trump, As It Has In The Past However, it is possible that Democrats would choose to only partially reverse the corporate tax cuts, while also lifting taxes on higher-income households. One should also note that trade tensions with China would probably diminish under a Biden presidency, which would be a mitigating factor for equity investors. Chart 10Cyclical Sectors Should Outperform Defensives As Global Growth Recovers... And A Weaker Dollar Should Also Help Non-US Stocks Q: So to sum up, you are still bullish on stocks over a 12-month horizon, although you see some near-term risks stemming from the likelihood of a second wave of the pandemic and uncertainty about how and when the fiscal cliff problem in the US will be resolved. What are your favorite sectors, regions, and styles? A: Cyclical sectors should outperform defensives over the next 12 months as global growth recovers. Cyclicals are overrepresented outside the US, which should favor overseas markets. A weaker dollar should also help non-US stocks (Chart 10). The dollar generally trades as a countercyclical currency, implying that it will sell off as global growth recovers. Moreover, unlike last year, the greenback no longer enjoys the benefit of higher interest rates than those abroad. In terms of style, value should outperform growth. Growth stocks have done very well in a falling interest rate environment (Chart 11). However, interest rates cannot fall much further from current levels. Small caps should outperform large caps, both because small caps are more growth-sensitive and because they tend to be more popular among day traders. Google searches for “day trading” have spiked in the past few months (Chart 12). Chart 11Interest Rates Cannot Fall Much Lower From Current Levels, Which Will Allow Value To Outperform Growth Chart 12Day Trading Is Back In Vogue These Days Beyond the pure macro plays, the pandemic could lead to a number of unexpected changes that have yet to be fully discounted by markets. For example, we will likely see a surge in the demand for automobiles as people shun public transit. The pandemic could also accelerate the reshoring of manufacturing activity, particularly in the health care sector. Contract manufacturing companies with significant domestic operations will benefit. Additionally, more people will move to the suburbs to work from home and escape the virus and rising crime. This could boost the demand for new houses and lift suburban real estate prices. Since most suburbs are built on top of land previously zoned for agriculture, farmland prices could also rise. Appendix A Evolution Of S&P 500 EPS Estimates By Sector Peter Berezin Chief Global Strategist peterb@bcaresearch.com Footnotes 1  Amazon EPS is projected to rise by 54% between 2019 and 2021, from 11% of overall consumer discretionary earnings to 19%. 2 One can see this within the context of the IS-LM model that is taught to economics undergraduates. If the LM curve shifts outward while the IS curve shifts inward, one could end up with the situation where aggregate demand is the same as before, but the equilibrium interest rate is lower. 3  For example, Gennaro Bernile, Vineet Bhagwat, and P. Raghavendra Rau investigated the link between the intensity of early-life experiences on CEO’s attitudes towards risk. Their results suggest that CEOs who witnessed extreme levels of fatal natural disasters appear more cautious in approaching risk. In contrast, those that experience disasters without very negative consequences become desensitized to risk. For details, please see Gennaro Bernile, Vineet Bhagwat, and P. Raghavendra Rau, “What Doesn't Kill You Will Only Make You More Risk-Loving: Early-Life Disasters and CEO Behavior,“ The Journal of Finance, (72:1) February 2017.   Global Investment Strategy View Matrix Current MacroQuant Model Scores

Highlights Falling volatility in oil-trading markets will remain suspect while the massive economic uncertainty plaguing global markets persists. Geopolitical risk also will remain high, as the US and China return to loggerheads and India and China move closer to war. Positive consumer and employment data in the US could presage a sharp recovery in demand generally; however, it is immediately countered with fears of a second COVID-19 wave, which now is the baseline scenario of our global investment strategists. Despite lower EM oil-demand growth this year – spurred by weaker GDP growth – deeper production cuts by OPEC 2.0 will keep oil markets on track to rebalance beginning in 3Q20. Massive fiscal and monetary stimulus will bridge global economic activity to a return to normal next year, provided the second wave of the COVID-19 pandemic does not result in renewed lockdown measures. Our updated supply-demand balances keep our expectation for Brent prices at $40/bbl this year and put next year’s average price at $65/bbl, $3/bbl below last month’s forecast. We continue to expect WTI to trade $2-$4/bbl lower than Brent. Feature As the OPEC 2.0 Joint Ministerial Monitoring Committee convenes today, members will be attempting to sort out the appropriate supply response to a highly uncertain oil-demand evolution over the balance of this year and next. Indeed, global economic policy uncertainty is scaling heights unimagined even in the depths of the Global Financial Crisis (GFC) of 2007-09 or the European sovereign-debt crisis of 2010-12, which followed in the GFC’s wake (Chart of the Week). This uncertainty is driving the policy responses of central banks and governments around the world, as they attempt to bridge COVID-19-induced demand destruction and the return to normality they seek in re-opening their economies. The data informing policy are suspect, as are the responses of firms and households to the stimulus they provide. This reflects the near-complete uncertainty in re current economic conditions. This translates directly to estimates of fundamental supply and demand variables, particularly in oil, which has been hardest-hit among the major commodities (Chart 2). Chart of the WeekEconomic Uncertainty Plagues Oil Markets Chart 2Oil Hardest Hit Commodity In 2020 COVID-19 Pandemic Demand To Weaken More Than Expected In 2020 OPEC 2.0’s agreement earlier this month to extend its 9.7mm b/d production cuts into July likely were informed by weaker physical demand. Our updated oil-demand model – driven by World Bank estimates of DM and EM GDP growth – indicates global oil consumption will fall by close to 9mm b/d this year, or ~ 1mm b/d more than we estimated last month.1 For next year, we expect a stronger rebound – 8.5mm b/d vs. last month’s estimate of 8mm b/d – off a lower base this year. This change is driven by the Bank’s more pessimistic assessment of EM GDP growth for 2020 than the IMF growth estimates we used in last month’s forecast (Chart 3). DM demand will take a harder hit than EM, given the extent of the lockdowns in major systematically important economies. This will set up a stronger rebound in oil demand next year, which, among many things spawned by the COVID-19 pandemic, is rarely seen. Chart 3EM Oil Demand Growth Estimate Lowered OPEC 2.0’s agreement earlier this month to extend its 9.7mm b/d production cuts into July likely were informed by weaker physical demand – appearing as unintended inventory accumulation – reflecting slower GDP growth. Global Oil Supply Expansion Required In our updated balances, we expect OPEC 2.0 supply to contract 3.2mm b/d y/y in 2Q20, and to increase in 2H20 and 2021 to keep prices from overshooting in the event the global demand response to fiscal and monetary stimulus is underestimated. We expect US shales to contract 600k b/d this year to 9.3mm b/d of production, and to gradually rebound in 2021 (Chart 4).2 The contraction in US shales will lead non-OPEC 2.0 supply losses in our estimation (Table 1). Chart 4Cuts By OPEC 2.0, US Shales Will Remove 9.4mm b/d Table 1BCA Global Oil Supply - Demand Balances (MMb/d, Base Case Balances) The combination of reduced supply and higher demand growth beginning next month will produce a physical deficit in 2H20 and in 2021 (Chart 5). This will be apparent in falling storage levels (Chart 6) and in a further flattening and eventual backwardating of the Brent and WTI forward curves (Chart 7). Chart 5Physical Markets Will Tighten Chart 6... Causing Storage to Drain ... Chart 7... And Forward Curves To Flatten, Then Backwardate Chart 8Massive Stimulus Flooding Global Economy Upside Favored, But Uncertainty Dominates We reckon even a second wave of the pandemic – now our Global Investment Strategy’s base case – will not derail a recovery in commodity demand. We continue to maintain a bias toward the upside price risk prevailing over the downside – driven by our expectation the massive fiscal and monetary stimulus unleashed globally will serve as an effective bridge from the COVID-19 pandemic to normal economic activity (Chart 8). This is being picked up in BCA Research's Global Nowcast, which closely tracks current economic conditions in leading manufacturing economies (Chart 9). We reckon even a second wave of the pandemic – now our Global Investment Strategy’s base case – will not derail a recovery in commodity demand.3 But the balance could tip the other way, with downside risk dominating the upside. The unprecedented uncertainty now dominating markets makes falling price volatility in oil markets – as measured by the implied volatility of Brent crude oil options’ implied volatility – highly suspect (Chart 10). We continue to emphasize two-way price risk in commodities remains pronounced despite the decline in the implied volatility of traded crude-oil options.4 Chart 9Global Economic Activity Turning Higher Chart 10Falling Vol Does Not Mean Lower Uncertainty Investment Implications The dynamics laid out above continue to point to a tightening physical oil market this year and next and higher prices. However, that does not come without substantial two-way risk. Indeed, the evolution of supply-demand information alone can trigger sharp adjustments in prices, as data revisions – to be expected, given the uncertainty prevailing at present – upend earlier preliminary estimates. We are leaving our 2020 forecast for Brent at $40/bbl and expect 2021 prices to average $65/bbl, $3/bbl below last month’s forecast. We continue to expect WTI to trade $2-$4/bbl lower than Brent (Chart 11). We also expect forward curves to flatten and return to backwardation in Brent and WTI, as the underlying physical markets tighten and inventories draw. Chart 11Brent To Average /bbl In 2021     Robert P. Ryan Chief Commodity & Energy Strategist rryan@bcaresearch.com Hugo Bélanger Associate Editor Commodity & Energy Strategy HugoB@bcaresearch.com Fernando Crupi Research Associate Commodity & Energy Strategy FernandoC@bcaresearch.com     Commodities Round-Up Energy: Overweight Brent prices are recovering from the dual supply and demand shocks delivered by the COVID-19 pandemic and the short-lived OPEC 2.0 internal market-share war. Brent price are now down 42% ytd vs. -72% two months ago. The contango in the Brent futures curve continues to narrow as voluntary and involuntary production cuts take effect and lockdown measures are relaxed in major economies. Continued production losses and demand recovery will force inventories lower, flattening the oil forward curves and ultimately backwardating them. Base Metals: Neutral As of Tuesday’s close, the LMEX index was up 17% since bottoming in March, 2ppt lower than the level reached last week. Positive data out of China – fueled by stimulative fiscal and monetary policies – indicates demand for industrial metals will grow: Year-on-year industrial production, infrastructure spending, and steel production grew by 4.4%, 10.9%, and 4.2%, respectively, in May (Chart 12). Moreover, y/y floor space started and sold moved up to positive territory. As government support continues to reach the economy, these sectors will encourage base metal consumption, providing further upside to the LME index. Still, fresh outbreaks of COVID-19 cases in Beijing – and associated lockdown measures – illustrate the fragility of the recovery over the short-term. Precious Metals: Neutral Gold prices remain range-bound at ~ $1,700/oz, mimicking movements in US real rates. Going forward, both the Fed and market participants expect US interest rates will remain pinned near zero through the end of 2022 (Chart 13). Our US Investment strategists expect the Fed will err to the side of providing too much accommodation as it navigates the uncertain consequences of the current economic shock. A gradual rebound in inflation next year could push real rates deeper in negative territories, which will be supportive for gold. Ags/Softs:  Underweight July soybean prices are up more than 3% since the beginning of the month. Strong export prospects going forward contributed to the strength in prices this past week. On June 4th the USDA reported new sales of soybeans of 1.21 MM MT, a huge week-on-week jump, which brought outstanding sales for the next marketing season to 4.1 MM MT. China was responsible for close to half of these sales and private exporters have since reported a little over an additional 1 million MT of exports to China. Chart 12Chinese Infrastructure Investment Rising Chart 13US Rates Expected To Remain Near Zero Until End 2022       Footnotes 1     Please see p. 3 of the World Bank’s June 2020 Global Economic Prospects. 2     We proxy US shales using the sum of crude production from the top 5 tight oil basins (i.e. Anadarko, Bakken, Eagle Ford, Niobrara, and Permian). Recent news reports suggest as much as 500k b/d of previously shut-in production will be back on line by the end of the month as a consequence of higher prices. This is slightly above our estimates shown in Chart 4. Please see US shale companies to boost oil output by 500,000 bpd by month-end published June 17, 2020, by reuters.com. 3    Please see A Second Wave Is Now The Base Case (But Stocks Will Eventually Shrug It Off) published by BCA Research’s Global Investment Strategy June 12, 2020. It is available at gis.bcaresearch.com. 4    For a discussion of how options markets price risk – i.e., known economic and political factors with outcomes that can be assigned probabilities – please see Ryan, Bob and Tancred Lidderdale (2009), Energy Price Volatility and Forecast Uncertainty, published by the US EIA October 2009. Risk can be thought of a “known unknowns” that can be measured across time and assigned a probability (conditional or otherwise), while uncertainty literally consists of unknown unknowns that cannot be measured.     Investment Views and Themes Recommendations Strategic Recommendations Tactical Trades Trade Recommendation Performance In 2020 Q1 Commodity Prices and Plays Reference Table Trades Closed in 2020 Summary of Closed Trades

Highlights We conservatively estimate lost output from shutdowns and social distancing will equal $10 trillion, and we expect the jobs market to be permanently scarred. Inflation, even at 2 percent, is a pipe dream, which leads to three investment conclusions on a 1-year horizon: Overweight US T-bonds and Spanish Bonos versus German Bunds and French OATs. Any high-quality bond yield that can decline will decline. Overweight CHF/USD. The tightening yield spread will structurally favour the CHF, while the haven status of the CHF should prevent it from underperforming in periods of market stress. Overweight defensive equities (technology and healthcare) versus cyclical equities (banks and energy). This implies underweight European equities versus other markets. Fractal trade: Short Germany versus the UK. The recent outperformance of German equities is technically extended. Feature Chart of the WeekCredit Impulses Are Large, But The Hole In Output Is Much Larger Big numbers befuddle us. Hardly a day passes without someone listing the unprecedented global stimulus unleashed to counter the coronavirus forced shutdowns – the trillions in government spending promises, tax relief, loan guarantees, money supply growth, and central bank asset-purchases. The most optimistic estimates quantify the total stimulus at $15 trillion. This includes $7 trillion of loan guarantees plus increases in central bank balance sheets which do not directly boost demand. So the direct stimulus is closer to $7 trillion.1 Yet the size of the stimulus is meaningless until we quantify the massive hole in economic output that needs to be filled. Assuming no further large-scale shutdowns, we conservatively estimate that the hole will amount to 12 percent of world output, or $10 trillion. A $10 Trillion Hole In Output Last week, the UK’s Office for National Statistics (ONS) helped us to estimate the hole in output, because unusually the ONS calculates UK GDP on a monthly basis. Between February and April, when the UK economy went from fully open to full shutdown, UK GDP collapsed by 25 percent. This despite the UK having an outsized number of jobs suitable for ‘working from home.’ For a more typical economy, we estimate that a full shutdown collapses output by 30 percent (Chart I-2). Chart I-2A Full Shutdown Collapses Output By 30 Percent The next question is: how long does the full shutdown last? Assuming it lasts for three months, output would suffer a hole amounting to 7.5 percent of annual GDP.2  But in practice, the economy will not fully re-open after three months. Social distancing will persist until people feel confident that the pandemic is under control. An effective vaccine against Covid-19 is unlikely to be available for a year. So, even without government policy to enforce social distancing, many people will choose to avoid crowds and congregations for fear of catching the virus. The size of the stimulus is meaningless until we quantify the massive hole in economic output. This means that the sectors that rely on crowds and congregations – leisure and hospitality and retail trade – will be operating at half-capacity, at best. Given that these sectors generate 9 percent of GDP, operating at half-capacity will create an additional hole amounting to 4.5 percent of output. More worryingly, these two sectors employ 21 percent of all workers, so operating at sub-par will leave the jobs market permanently scarred.3 Combining the 7.5 percent existing hole with the 4.5 percent future hole, the full hole in economic output will amount to around 12 percent of annual GDP. As global GDP is worth around $85 trillion, this equates to $10 trillion.  Crucially though, our estimate assumes that a second wave of the pandemic will not force a new cycle of shutdowns. If it does, the hole will become even bigger. Don’t Be Fooled By Money Supply Growth The recent growth in broad money supply seems a big number. Since the start of the year, the outstanding stock of bank loans has increased by around $0.7 trillion in the euro area, and by $1 trillion in both the US and China (Chart I-3 and Chart I-4). This has boosted the 6-month credit impulses in all three economies. Indeed, the US 6-month credit impulse recently hit its highest value of all time, and the combined 6-month impulse across all three blocs equals around $2 trillion (Chart of the Week). Chart I-3Don't Be Fooled By Money Supply Growth In The Euro Area And The US... Chart I-4...And In ##br##China This 6-month credit impulse quantifies the additional borrowing in the most recent six-month period compared to the previous period. Ordinarily, a $2 trillion impulse would create a huge boost to demand. After all, the private sector does not usually borrow just to hold the cash in a bank. Yet in the coronavirus crisis this is precisely what has happened. While the shutdowns lasted, firms drew on existing bank credit lines to build up emergency cash buffers. Therefore, much of the money growth will not generate new demand. While the shutdowns lasted, firms drew on existing bank credit lines to build up emergency cash buffers.  To the extent that this cash is sitting idly in a firm’s bank account, the monetary velocity will decline. Meaning there will be a much-reduced transmission from credit impulses to spending growth. Furthermore, when the economy re-opens, many firms will relinquish the precautionary credit lines. There is no point holding cash in the bank when there are few investment opportunities. Hence, credit impulses will fall back – as seems to be the case right now in the US. QE: The Great Misunderstanding To repeat, big numbers befuddle us. They must always be put into context. No truer is this than when it comes to central bank asset-purchases. The great misunderstanding is that the act of central banks buying assets, per se, drives up those asset prices. Central banks act as lenders of last resort to solvent but illiquid banks and sovereigns. If there is ample liquidity in these markets – as is the case now – then the primary function of central bank asset-purchases is to set the term-structure of interest rates. In turn, the term-structure of global interest rates establishes the prices of $500 trillion of global assets. The prices of these assets are inextricably inter-connected and inter-dependent4 (Chart I-5). Chart I-5The Prices Of $500 Trillion Of Assets Are Inextricably Inter-Connected The great misunderstanding is that the act of central banks buying assets, per se, drives up those asset prices. Yet central banks set no price target for their asset-purchases. They leave that to the market. Moreover, in the context of the $500 trillion of inter-dependent asset prices, the $10-15 trillion or so of central bank asset-purchases to date constitutes chicken feed (Chart I-6). Hence, the mechanism by which asset-purchases work is through the signal they give to the $500 trillion market on the likely course of interest rate policy. This sets the term-structure of interest rates, which in turn sets the required return on all the $500 trillion of assets (Chart I-7). Chart I-6$10-15 Trillion Of QE Is Chicken Feed... Chart I-7...Compared To $500 Trillion Of Assets Priced By The Term-Structure Of Interest Rates As the ECB’s former Chief Economist, Peter Praet, explains: “There is a signalling channel inherent in asset purchases, which reinforces the credibility of forward guidance on policy rates. This credibility of promises to follow a certain course for policy rates in the future is enhanced by the asset purchases, as these asset purchases are a concrete demonstration of our desire (to keep policy rates at the lower bound.)” The credible commitment to keep policy rates near the lower bound for an extended period depresses bond yields towards the lower bound too. But once bond yields have reached their lower bound the effectiveness of central bank asset-purchases becomes exhausted. Three Investment Conclusions The main purpose of this report was to put the $7 trillion of direct stimulus dollars unleashed into the economy into a proper context. With lost output estimated at $10 trillion and the jobs market permanently scarred, inflation – even at 2 percent – is a pipe dream. Moreover, a second wave of the pandemic and a new cycle of shutdowns would inject a further disinflationary impulse. This leads to three investment conclusions on a 1-year horizon: Any high-quality bond yield that can decline – because it is not already near the -1 percent lower bound to yields – will decline. An excellent relative value trade is to overweight US T-bonds and Spanish Bonos versus German Bunds and French OATs (Chart I-8). Long CHF/USD is a win-win. The tightening yield spread will structurally favour the CHF, while the haven status of the CHF should prevent it from underperforming in periods of market stress. Overweight defensive equities versus cyclical equities, with technology correctly defined as defensive, not cyclical. The performance of cyclicals (banks and energy) versus defensives (technology and healthcare) is now joined at the hip to the bond yield (Chart I-9). This implies underweight European equities versus other markets. Chart I-8Bond Yields That Can Decline Will Decline Chart I-9The Performance Of Cyclicals Versus Defensives Is Joined At The Hip To The Bond Yield Fractal Trading System* The recent outperformance of German equities is technically extended. Accordingly, this week’s recommended trade is to go short Germany versus the UK, expressed through the MSCI dollar indexes. Set the profit target and symmetrical stop-loss at 5 percent. In other trades, long euro area personal products versus healthcare achieved its 7 percent profit target at which it was closed. The rolling 1-year win ratio now stands at 65 percent. When the fractal dimension approaches the lower limit after an investment has been in an established trend it is a potential trigger for a liquidity-triggered trend reversal. Therefore, open a countertrend position. The profit target is a one-third reversal of the preceding 13-week move. Apply a symmetrical stop-loss. Close the position at the profit target or stop-loss. Otherwise close the position after 13 weeks. * For more details please see the European Investment Strategy Special Report “Fractals, Liquidity & A Trading Model,” dated  December 11, 2014, available at eis.bcaresearch.com. Footnotes 1 Source: Reuters estimate. 2 A 30 percent loss in output for a quarter of a year (3 months) amounts to a 30*0.25 = 7.5 percent loss in annual output. 3 Using the weights of leisure and hospitality and retail trade in the US economy as a proxy for the global weights. 4 The $500 trillion of assets comprises: real estate $300 trillion, public and private equity $100 trillion, corporate bonds and EM debt $50 trillion, and high-quality government bonds $50 trillion.   Dhaval Joshi Chief European Investment Strategist dhaval@bcaresearch.com Fractal Trading System   Cyclical Recommendations Structural Recommendations Closed Fractal Trades Trades Closed Trades Asset Performance Currency & Bond Equity Sector Country Equity Indicators Bond Yields Indicators To Watch - Bond Yields Indicators To Watch - Bond Yields Indicators To Watch - Bond Yields Indicators To Watch - Bond Yields Interest Rate Indicators To Watch - Interest Rate Expectations Indicators To Watch - Interest Rate Expectations Indicators To Watch - Interest Rate Expectations Indicators To Watch - Interest Rate Expectations

Highlights The relaxation of lockdown measures, along with mass protests over the past two weeks, have made a second wave of the pandemic more likely than not in many countries. Unlike during the first wave, most governments will not shutter their economies in response to a renewed spike in infection rates. For better or for worse, the “Sweden strategy” will become commonplace. As today’s stock market selloff illustrates, a second wave could significantly unnerve investors, especially since it is coming on the heels of a substantial rally in stocks. However, global equity prices will still rise over a 12-month horizon. Easy monetary policy, improving labor market conditions, and significant amounts of cash on the sidelines should allow the equity risk premium to decline, especially outside the US where valuations remain quite cheap. The US dollar has entered a cyclical bear market. This is especially positive for commodities, economically-sensitive equity sectors, and non-US stocks. Opening The Hatch Chart 1Governments Are Lifting Lockdown Restrictions Three months after the virus burst out of China, countries around the world are starting to relax lockdown measures. Our COVID-19 Government Response Stringency Index, created by my colleague Jonathan LaBerge and showcased in last week’s Global Investment Strategy report, has been on an easing course since May. A similar measure developed by Goldman Sachs broadly shows the same loosening pattern. Reflecting these developments, the Dallas Fed’s index of “mobility and engagement” has been slowly returning to normal (Chart 1). The reopening of economies is taking place despite limited success in containing the virus. While some countries have seen a considerable drop off in the number of new cases and deaths, others continue to experience an increase in both metrics (Chart 2). Globally, the number of new cases has begun to trend higher after remaining flat for most of April. The number of deaths — which lags new cases by about three weeks but is less vulnerable to statistical distortions caused by changes in testing prevalence — has also ticked higher after falling for nearly two months. Mass protests starting in Minneapolis and spreading to much of the western world have the potential to further increase the infection rate. As Jonathan noted last week, large gatherings have been an important vector of transmission for the virus. While the protests have occurred outdoors, many protestors did not wear masks while singing and shouting nor practise social distancing. Chart 2Globally, The Number Of New Cases and Deaths Has Started To Trend Higher Again A Risky Gambit How markets react to a second wave of the pandemic will depend a lot on how policymakers and the broader public respond. For better or for worse, the patience for continued lockdowns has waned. The US and a number of other countries appear to be moving towards the “Swedish model” of trying to keep a lid on the virus without imposing draconian lockdown restrictions. It is a risky gambit, especially in light of the jump in infections that Sweden has reported in the past two weeks. While some countries such as China and New Zealand, which have effectively eradicated the virus, can allow most activities – with the exception of international travel – to resume, others should arguably wait longer until they too have defeated the disease. As Professor Peter Doherty, renowned immunologist and co-recipient of the 1996 Nobel Prize for Medicine, discussed in a webcast with my colleague Garry Evans on Monday, significant progress has been made towards developing a vaccine for COVID-19. Opening up economies now could cause a lot of needless death before a vaccine becomes available. Near-Term Risks To Stocks… Chart 3Earnings Estimates Have Taken It On The Chin Even if governments continue opening up their economies despite rising infection rates, some people will increase the amount of social distancing they practise regardless of official recommendations. Airline, cruise ship, and restaurant stocks had rallied mightily off their March lows before giving up some of their gains over the past few days. If a second wave occurs, they will fall further. The rally in stocks linked to the reopening of the economy occurred alongside a retail investor speculative frenzy. In one of the more bizarre episodes in financial history, stocks of bankrupt or soon-to-be-bankrupt companies surged on Monday as novice day traders snapped up shares of companies that most institutional equity investors had left for dead. Meanwhile, earnings estimates have taken it on the chin (Chart 3). Many companies chose not to provide guidance for the second quarter, citing unprecedented uncertainty over the near-term business outlook. Since Q2 will be the worst quarter for economic growth, it will probably also be a very bad quarter for earnings. The prospect of a slew of poor earnings reports in July could further dent investor sentiment, exacerbating the stock market correction we have seen over the past few days. All this suggests that global equities could experience some further weakness over the next few months. …But Still Sticking With Our 12-Month Overweight To Equities Chart 4Economic Activity Has Started Rebounding Despite these short-term risks, we are not ready to abandon our cyclical overweight view on stocks. While many people have remarked that the equity market has diverged from the economy, in fact, the rebound in the stock market has tracked the peak in initial unemployment claims and the trough in current activity indicators quite closely (Chart 4). A second wave would certainly slow the economic rebound. However, it would probably not reverse it completely given that the mortality rate from the virus now appears to be somewhat lower than initially feared and an increasing number of medical treatments are becoming available. If output and employment keep rising, stocks are likely to trend higher. A Deep Hole This does not mean that everything will return to normal soon. Even though global growth appears to have bottomed in April, the level of employment remains at depression-like levels (Chart 5). About 12% of US workers are employed in the hospitality, restaurant, and travel sectors. A return to normalcy in those sectors will take several years at best. Nevertheless, the recovery will not be nearly as drawn out as the one following the Global Financial Crisis. The Congressional Budget Office expects that it will take another eight years for the US unemployment rate to fall back to 5% (Chart 6). That seems unduly pessimistic. Chart 5Employment Remains At Depression-Like Levels Chart 6CBO Projects The Unemployment Rate Will Fall Very Slowly Cyclical Versus Structural Unemployment Chart 7Residential Construction Accounted For Less Than 20% Of The Job Losses During The Great Recession Commentators like to talk about structural unemployment, but the truth is that large increases in joblessness usually reflect deficient labor demand rather than insufficient supply. For example, the decline in residential construction employment and related sectors accounted for less than one-fifth of the job losses during the Great Recession (Chart 7). You don’t have to fill a half-empty pool through the same pipe from which the water escaped. As long as there is enough demand throughout the economy, workers who lose their jobs will likely find new jobs elsewhere, whether it be at an Amazon distribution center or any number of manufacturing companies that will benefit from the repatriation of production back onshore. The shift in jobs from one sector to the next is not instantaneous, but it need not drag on for years either. Policy Will Stay Stimulative This is where the role of monetary and fiscal policy takes center stage. Despite the improving economic outlook, government bond yields have barely moved off their lows as investors have become increasingly convinced that central banks will keep rates at rock-bottom levels (Chart 8). This week’s FOMC meeting made it clear that the Fed has no intention of raising rates through 2022. “We’re not thinking about raising rates. We’re not even thinking about thinking about raising rates,” Fed Chairman Jerome Powell declared during his press conference. Granted, the zero lower bound has prevented yields from falling as much as they normally would. Fortunately, fiscal policy has stepped in to fill the void. Chart 9 shows that governments have eased fiscal policy much more this year than they did in 2008-09. If governments tighten fiscal policy prematurely like they did after the Great Recession, the recovery will indeed be sluggish. Such a risk cannot be ignored. BCA’s geopolitical team, led by Matt Gertken, has argued that Republican Senators will initially resist the proposed $3 trillion in new stimulus, until they are forced to act by a major new round of financial or social turmoil. Nevertheless, Matt thinks that the Republican Senate will ultimately buckle under the political pressure, knowing full well that a large dose of fiscal largess could prevent a Democratic sweep in November. Chart 8Yields Remain Close To Recent Lows Chart 9Will It Be Enough? Chart 10China Has Ramped Up Stimulus Outside the US, fiscal support shows little sign of being scaled back. Germany has pushed forward with additional stimulus, going so far as to propose a risk-sharing arrangement via the creation of an EU Recovery Fund. On Wednesday, the Japanese House of Representatives approved a draft supplementary budget of 32 trillion yen ($296 billion) providing additional funding for small businesses and medical workers. Jing Sima, BCA Research's chief China strategist, expects Chinese credit formation as a share of GDP to reach the highest level since 2009 and the budget deficit to widen to the largest on record (Chart 10). The upshot is that we may find ourselves in an environment over the next few years where global GDP and corporate profits are moving back to trend, while interest rates (and the implied discount rate used for valuing stocks) stay at very low levels. If profits return back to normal but interest rates do not, the surreal implication is that the pandemic could end up increasing the fair value of the stock market. Ample Cash On The Sidelines Stocks also have another factor working in their favor: huge amounts of cash on the sidelines (Chart 11). The combination of massive fiscal income transfers and low spending has led to a surge in private-sector savings. The US personal savings rate reached 33% in April, the highest on record. Reflecting this increase in savings, private sector bank deposits have ballooned (Chart 12). Chart 11Sizable Amount Of Dry Powder Chart 12Savings Have Spiked Amid Stimulus     Investors often talk about cash “flowing” in and out of the stock market. This is a somewhat misleading characterization. Setting aside the impact of corporate buybacks and public share offerings, the decision by one person to buy shares requires a corresponding decision by someone else to sell shares. The buyer of the shares loses some cash, while the seller gains some cash. On net, there is no inflow of cash into the stock market. Rather, what happens is that the price of shares adjusts to ensure that there is a seller for every buyer. If people feel that they have too much cash relative to the value of their equity holdings, they will bid up the price of stocks until enough sellers come forward. This will cause the amount of cash that people hold as a percentage of their total wealth to shrink, even if the dollar value of that cash remains the same. The process will only stop when the amount of cash that people hold is in line with their preferences. The amount of cash held in US money market funds and personal cash deposits has surged by $2.6 trillion since February. Despite the rally in equities, cash holdings as a percent of stock market capitalization remain near multi-year highs. This suggests that the firepower to fuel further increases in the stock market has not been exhausted. Start Of The Dollar Bear Market After peaking in March, the broad trade-weighted US dollar has weakened by 5.3%. The dollar is a countercyclical currency, meaning that it tends to move in the opposite direction of the global business cycle (Chart 13). While the dollar could strengthen temporarily in response to a second wave of the pandemic, global growth should continue to recover in the second half of the year provided that severe lockdown measures are not reintroduced. Stronger global growth will push the greenback lower. Chart 13The US Dollar Is A Countercyclical Currency Unlike last year, the dollar no longer has support from higher US interest rates. Indeed, US real rates are below those of many partner countries due to the fact that US inflation expectations are generally higher than elsewhere (Chart 14). Chart 14The Dollar Has Been Losing Interest Rate Support A Weaker Dollar Will Support Non-US Stocks The combination of a weaker dollar and stronger global growth should disproportionately help the more cyclical sectors of the stock market, particularly commodity producers. Since cyclical stocks tends to be overrepresented outside the US, non-US equities should outperform their US peers over the next 12 months. A weaker dollar will also reduce the local- currency value of dollar-denominated debt. This will be especially helpful for emerging markets. Despite the recent rally, the cyclically-adjusted PE ratio for EM stocks remains near historic lows (Chart 15). EM equities should fare well over the next 12 months.   Chart 15EM Stocks Are Very Cheap Peter Berezin Chief Global Strategist peterb@bcaresearch.com Global Investment Strategy View Matrix Current MacroQuant Model Scores