先進国

Highlights Over the short term – 1-2 years – the pick-up in re-infection rates in Asia and LatAm states with large-scale deployments of Sinopharm and Sinovac COVID-19 vaccines will re-focus attention on demand-side risks to the global recovery (Chart of the Week). The UAE-Saudi impasse re extending the return of additional volumes of OPEC 2.0 spare capacity to the oil market over 2H21 will be short-lived.  The UAE's official baseline production will be increased to 3.8mm b/d from 3.2mm b/d presently, and its output in 2H21 will be adjusted accordingly.  Over the medium term – 3-5 years out – the risk to the expansion of metal supplies needed for renewables and electric vehicles (EVs) will rise, as left-of-center governments increase taxes and royalties, and carbon prices move higher. Rising metals costs will redound to the benefit of oil and gas producers, and accelerate R+D in carbon- and GHG-reduction technologies. Longer-term – 5-10 years out – the active discouragement of investment in hydrocarbons will contribute to energy shortages. In anticipation of continued upside volatility in commodity prices and share values of oil, gas and metals producers, we remain long the S&P GSCI and COMT ETF, and long equities of producers and traders via the PICK ETF. Feature Our conversations with clients almost invariably leads us to considering the risks to our long-standing bullish views for energy and metals. This week, we reprise some of the highlights of these conversations. In the short term, our bullish call on oil is underpinned by the assumption of continued expansion in vaccinations, which we believe will lead to global economic re-opening and increased mobility, as the world emerges from the devastation of COVID-19. This expectation is once again under scrutiny. On the supply side, the very public negotiations undertaken by the UAE and the leaders of OPEC 2.0 – the Kingdom of Saudi Arabia (KSA) and Russia – over re-basing the UAE's production reminds investors there is substantial spare capacity from the coalition available for the market over the short term. The slow news cycle going into the US Independence Day holiday certainly was a fortuitous time to make such a point. Chart of the WeekWorrisome Uptick Of COVID-19 Cases KSA-UAE Supply-Side Worries The abrupt end to this week's OPEC 2.0 meeting was unsettling to markets. Shortly after the meeting ended – without being concluded – officials from the Biden administration in the US spoke with officials from KSA and the UAE, presumably to encourage resolution of outstanding issues and to get more oil into the market to keep crude oil prices below $80/bbl (Chart 2). We're confident the KSA-UAE impasse re extending the return of additional volumes of spare capacity to the oil market over 2H21 will be short-lived. The UAE's official baseline production number (i.e., its October 2018 output level) will be increased to 3.8mm b/d from 3.2mm b/d presently, and its output in 2H21 will be adjusted accordingly. Coupled with a likely return of Iranian export volumes in 4Q21, this will bring prices down into the mid- to high-$60/bbl range we are forecasting. Chart 2US Pushing For Resolution of KSA-UAE Spat Longer term, markets are worried this incident is a harbinger of a breakdown in OPEC 2.0's so-far-successful production-management strategy, which has lifted oil prices 200% since their March 2020 nadir. At present, the producer coalition has ~ 6-7mm b/d of spare capacity, which resulted from its strategy to keep the level of supply below demand. A breakdown in this discipline – in extremis, another price war of the sort seen in March 2020 or from 2014-2016 – could plunge oil markets into a price collapse that re-visits sub-$40/bbl levels. In our view, economics – specifically the cold economic reality of the price elasticity of supply – continues to work for the OPEC 2.0 coalition: Higher revenues are realized by members of the group as long as relatively small production cuts produce larger revenue gains – e.g., a 5% (or less) cut in production that produces a 20% (or more) increase in price trumps a 20% increase in production that reduces prices by 50%. Besides, none of the members of the coalition possess the wherewithal to endure another shock-and-awe display from KSA similar to the one following the breakdown of the March 2020 OPEC 2.0 meeting. We also continue to expect US shale-oil producers to be disciplined by capital markets, and to retain a focus on providing competitive returns to their shareholders, which will limit supply growth to that which maintains profitability. Until we see actual evidence of a breakdown in the coalition's willingness to maintain its production-management strategy, we will continue to assume it remains operative. Worrisome COVID-19 Re-Infection Trends Reports of increased re-infection rates in Latin American and Asia-Pacific states providing Chinese Sinopharm and Sinovac COVID-19 vaccines will re-focus attention on demand-side risks to the global recovery. Conclusive data on the efficacy of these vaccines is not available at present, based on reporting from Health Policy Watch (HPW).1 The vast majority of these vaccines were purchased in Latin America and the Asia-Pacific region, where ~ 80% of the 759mm doses of the two Chinese vaccines were sold, according to HPW's reporting. This will draw the attention of markets to this risk (Chart 3). Of particular concern are the increases in re-infection rates in the Seychelles and Chile, where the majority of populations in both countries were inoculated with one of the Chinese vaccines. Re-infections in Indonesia also are drawing attention, where more than 350 healthcare workers were re-infected after receiving the Sinovac vaccination.2 The risk of renewed global lockdowns remains small, but if these experiences are repeated globally with adverse health consequences, this assessment could be challenged. Chart 3COVID-19 Returning In High-Vaccination States Transition Risks To A Low-Carbon Economy Over the medium- to long-terms, our metals views are premised on the expectation the build-out of the global EV fleet and renewable electricity generation – including its supporting grids – will require massive increases in the supply of copper, aluminum, nickel, and tin, not to mention iron ore and steel. This surge in demand will be occurring as governments rush headlong into unplanned and unsynchronized wind-downs of investment in the hydrocarbon fuels that power modern economies.3 The big risk here is new metal supplies will not be delivered fast enough to build all of the renewable generation, EVs and their supporting grids and infrastructures to cover the loss of hydrocarbons phased out by policy, legal and boardroom challenges. Such a turn of events would re-invigorate oil and gas production. Renewable energy and electric vehicles are the sine qua non of the drive to achieve net-zero carbon emissions by 2050. However, the rising price of base metals will add to already high costs of rebuilding power grids to make them suitable for green energy. Given miners’ reluctance to invest in new mines, we do not expect metals prices to drop anytime soon. According to Wood Mackenzie, in 2019 the cost of shifting just the US power grid to renewable energy over the next 10 years will amount to $4.5 trillion.4 Given these cost and supply barriers, fossil fuels will need to be used for longer than the IEA outlined in its recent and controversial report on transitioning to a net-zero economy.5 To ensure that fossil fuels can be used while countries work to achieve their net zero goals, carbon capture utilization and storage (CCUS) technology will need to be developed and made cheaper. The main barrier to entry for CCUS technology is its high cost (Chart 4). However, like renewable energy, the more it is deployed and invested in, the cheaper it will become, following the trend seen in the development of renewable energy and EVs, which were aided by large-scale subsidies from governments to encourage the development of the technology. These cost reductions are already visible: In its 2019 report, the Global CCS Institute noted the cost of implementing CCS technology initially used in 2014 had fallen by 35% three years later. Chart 4CCUS Can Be Expensive Metals Mines' Long Lead Times In 2020 the total amount of discovered copper reserves in the world stood at ~ 870mm MT (Chart 5), according to the US Geological Service (USGS). As of 2017, the total identified and undiscovered amount of reserves was ~ 5.6 billion MT.6 The World Bank recently estimated additional demand for copper would amount to ~ 20mm MT p.a. by 2050 (Chart 6).7 Glencore’s recently retired CEO Ivan Glasenberg last month said that by 2050, miners will need to produce around 60mm MT p.a. of copper to keep up with demand for countries’ net zero initiatives.8 Even with this higher estimate, if miners focus on exploration and can tap into undiscovered reserves, supply will cover demand for the renewable energy buildout. Chart 5Copper Reserves Are Abundant Chart 6Call On Base Metals Supply Will Be Massive Out To 2050 While recent legislative developments in Chile and Peru, which together constitute ~ 34% of total discovered copper reserves, could lead to significantly higher costs as left-of-center governments re-write these states' constitutions, geological factors would not be the main constraint to copper supply for the renewables energy buildout: Even if copper mining companies were to move out of these two countries, there still is about 570 million MT in discovered copper reserves, and nearly ten times that amount in undiscovered reserves. As we have written in the past, capital expenditure restraint is the principal reason the supply side of copper markets – and base metals generally – is challenged (Chart 7). Unlike in the previous commodity boom, this time mining companies are focusing on providing returns to shareholders, instead of funding the development of new mines (Chart 8). Chart 7Copper Prices Remains Parsimonious Chart 8Shareholder Interests Predominate Metals Agendas Of course, it is likely metals miners, like oil producers, are waiting to see actual demand for copper and other base metals pick up before ramping capex. Sharp increases in forecasted demand is not compelling for miners, at this point. This means metals prices could stay elevated for an extended period, given the 10-15-year lead times for copper mines (Chart 9). For example, the Kamoa-Kakula mine in the Democratic Republic of Congo (DRC) now being brought on line took roughly 24 years of exploration and development work, before it started producing copper. Technological breakthroughs that increase brownfield projects’ productivity, or significant increases in the amount of recycled copper as a percent of total copper supply would address some of the price pressures arising from the long lead times associated with the development of new copper supply. Another scenario with a non-trivial probability that threatens the viability of metals investing is a breakthrough – or breakthroughs – in CCUS technology, which allows oil and gas producers to remove enough carbon from their fuels to allow firms using these fuels to achieve their net-zero carbon goals. Chart 9Long Lead Times For Mine Development Investment Implications Short-term supply-demand issues affecting the oil market at present are transitory, and do not signal a shift in the fundamentals supporting our bullish call on oil. Our thesis based on continued production discipline remains intact. That said, we will continue to subject it to rigorous scrutiny on a continual basis. Our average Brent forecast for 2021 remains $66.50/bbl, with 2H21 prices averaging $70/bbl. For 2022 and 2023 we continue to expect prices to average $74 and $81/bbl, respectively (Chart 10). WTI will trade $2-$3/bbl lower. Our metals view has become slightly more nuanced, thanks to our client conversations. One of the unintended consequences of the unplanned and uncoordinated rush to a net-zero carbon future will be an improvement in the competitive position of oil and gas as transportation fuels and electric-generation fuels going forward. This will be driven by rising costs of developing and delivering the metals supplies needed to effect the net-zero transition. We expect markets will provide incentives to CCUS technologies and efforts to decarbonize oil and gas fuels, which will contribute to the global effort to arrest rising temperatures. This suggests the rush to sell these assets – which is underway at present – could be premature.9 In the extreme, this could be a true counterbalance to the metals story, if it plays out. Chart 10Our Oil Price View Remains Intact     Robert P. Ryan Chief Commodity & Energy Strategist rryan@bcaresearch.com Ashwin Shyam Research Associate Commodity & Energy Strategy ashwin.shyam@bcaresearch.com   Commodities Round-Up Energy: Bullish The monthly OPEC 2.0 meeting ended without any action to increase monthly supplies, following the UAE's bid to increase its baseline reference production – determined based on October 2018 production levels – to 3.8mm b/d, up from 3.2mm b/d. S&P Global Platts reported the UAE's Energy Minister, Suhail al-Mazrouei, advanced a proposal to raise its monthly production level under the coalition's overall output deal, while KSA's energy minister, Prince Abdulaziz bin Salman, insisted the UAE follow OPEC 2.0 procedures in seeking an output increase. We do not expect this issue to become a protracted standoff between these states. The disagreement between the ministers is procedural to substantive. Remarks by bin Salman last month – to wit, KSA has a role in containing inflation globally – and his earlier assertions that production policy of OPEC 2.0 would be driven by actual oil demand, as opposed to forecasted oil demand, suggest the Kingdom is not aiming for higher oil prices per se. Base Metals: Bullish Spot benchmark iron ore (62 Fe) prices traded above $222/MT this week in China on the back of stronger steel demand, according to mining.com (Chart 11). Market participants are anticipating further steel-production restrictions and appear to be trying to get out in front of them. Precious Metals: Bullish The USD rally eased this week, allowing gold prices to stabilize following the June Federal Open Market Committee (FOMC) meeting. In the two weeks since the FOMC, our gold composite indicator shows that gold started entering oversold territory (Chart 12). We believe gold prices will start correcting upwards, expecting investor bargain-hunting to pick up after the price drop. The mixed US jobs report, which showed the unemployment rate ticked up more than expected, implies that interest rates are not going to be raised soon. Our colleagues at BCA Research's US Bond Strategy (USBS) expect rates to increase only by end-2022.10 This, along with slightly higher odds of a potential COVID-19 resurgence, will support gold prices in the near-term. Ags/Softs: Neutral The USDA's Crop Progress report for the week ended 4 July 2021 showed 64% of the US corn crop was in good to excellent condition, down from the 71% reported for the comparable 2020 date. The Department reported 59% of the bean crop was in good to excellent shape vs 71% the year earlier. Chart 11 Chart 12     Footnotes 1     Please see Are Chinese COVID Vaccines Underperforming? A Dearth of Real-Life Studies Leaves Unanswered Questions, published by Health Policy Watch, June 18, 2021. 2     According to HPW, the World Health Organization's Emergency Use Listing for these two vaccines "were unique in that unlike the Pfizer, AstraZeneca, Moderna, and Jonhson & Johonson vaccines that it had also approved, neither had undergone review and approval by a strict national or regional regulatory authority such as the US Food and Drug Administration or the European Medicines Agency. Nor have Phase 3 results of the Sinopharm and Sinovac trials been published in a peer-reviewed medical journal.  More to the point, post-approval, any large-scale tracking of the efficacy of the Sinovac and Sinopharm vaccine rollouts by WHO or national authorities seems to be missing." 3    Please see A Perfect Energy Storm On The Way, which we published on June 3, 2021 for additional discussion.  It is available at ces.bcaresearch.com. 4    Please refer to The Price of a Fully Renewable US Grid: $4.5 Trillion, published by greentechmedia 28 June 2019. 5    Please refer to the IEA's Net Zero By 2050, published in May 2021. 6    Please refer to USGS Mineral Commodity Summaries, 2021. 7     Please refer to Minerals for Climate Action: The Mineral Intensity of the Clean Energy Transition, published by the World Bank. 8    Please refer to Copper supply needs to double by 2050, Glencore CEO says, published by reuters.com on June 22, 2021. 9    Please see the FT's excellent coverage of this trend in A $140bn asset sale: the investors cashing in on Big Oil’s push to net zero published on July 6, 2021. 10   Please refer to Watch Employment, Not Inflation, published by the USBS on June 15, 2021.   Investment Views and Themes Strategic Recommendations Tactical Trades Commodity Prices and Plays Reference Table Trades Closed in 2021 Summary of Closed Trades

Underweight (Upgrade Alert) We are currently underweight US banks, but the macro environment is changing and today we put this sub-sector on an upgrade alert looking to push it to a neutral allocation. The news on the buyback and dividend fronts is encouraging as banks will be allowed to resume their shareholder friendly activities that were halted last year due to the Fed’s Stress Test. Already, financials stocks are at the front of the pack with a roughly 3% total yield that is likely to increase further. Tack on the current search for yield environment, and the allure of financials equities becomes even more tempting.  Bottom Line: We are putting banks on our upgrade alert watchlist. Please see an upcoming Strategy Report where we delve deeper into the buyback and dividend topics.

Highlights Inflation is set to decelerate, job creation has a speed limit, and super-spreaders of new-variant Covid-19 infections will create speed bumps in the economy.  This means that in the second half of the year: Bonds will rally. The US dollar will rally. Growth stocks will outperform value stocks. US stocks will outperform non-US stocks. Fractal trade shortlist: Brazilian real, Saudi Tadawul All Share, and Marine Transportation.  Feature Chart of the WeekThe 60 Percent Correction In Lumber Shows What Happens When Supply Bottlenecks Ease. Are Used Cars Next? As Supply Bottlenecks Ease, Inflation Will Cool Since mid-March, US inflation has surged to 5 percent. Yet bond yields have drifted lower, by almost 50 bps in the case of the 30-year T-bond yield, equating to a handsome return of 12 percent. The seeming contradiction between rising inflation and declining bond yields has puzzled some people, but it shouldn’t. In 2009, the same pattern occurred in reverse. Inflation collapsed, culminating in a modern era low of -2 percent in July 2009. Yet while inflation was collapsing, bond yields rose sharply (Chart I-2 and Chart I-3). Chart I-2In 2009, Bond Yields Rose When Year-On-Year Inflation Fell Chart I-3In 2021, Bond Yields Fell When Year-On-Year Inflation Rose We can explain this seeming contradiction with an analogy from driving. The inflation rate is like your average speed over the past mile. But the bond market cares much more about your average speed over the next mile, or even over the next 5-10 miles. If you are driving at a constant speed, then your speed over the past mile is a good guide to your future speed. But if you have been driving unusually fast or unusually slowly, there is a more important predictor of your future speed. That important predictor is your acceleration – meaning, what is happening to your speed over successive hundred yards stretches. In the same way, during episodes of unusually low or unusually high inflation, the bond market focusses on the monthly rate of inflation, and specifically the moment that it stops decreasing, as in early-2009, or stops increasing, as in mid-2021. In 2008, after a long sequence of declining monthly rates of inflation that went deep into negative territory, the December 2008 print marked the first substantial increase. Hence, the bond yield also bottomed in December 2008 (Chart I-4), even though annual inflation did not bottom until July 2009. Chart I-4In 2009, Bond Yields Bottomed When Month-On-Month Inflation Bottomed Similarly, in 2020-21, after a six month sequence of increasing monthly rates of inflation, the May 2021 print marked the end of the rising trend. To the extent that this was anticipated, most of the decline in the bond yield has happened since mid-May (Chart I-5). Chart I-5In 2021, Bond Yields Topped When Month-On-Month Inflation Topped Since mid-May, the 60 percent crash in the lumber price shows what happens when supply bottlenecks ease. Other prices that are being supported by temporary supply constraints – such as used car prices – are likely to suffer the same fate (Chart of the Week). Hence, so long as the coming monthly prints confirm an ongoing deceleration in inflation, the current rally in bonds will stay intact. Jobs: The Hard Work Starts Now Staying on the theme of speed, there is a well-defined speed limit to every post-recession jobs recovery. In A Fed Rate Hike By Early 2023 Is Pie In the Sky, we pointed out the remarkable consistency in the pace of post-recession US jobs recoveries. The last five recessions had different causes, severities, durations and peak unemployment rates. Yet in the recoveries that followed each recession, the unemployment rate declined at a remarkably consistent pace of 0.4-0.5 percent per year (Table I-1). Table I-1After Every Recession, The Pace Of Recovery In The Jobs Market Is Near-Identical Reassuringly at the last FOMC press conference, Jay Powell supported this thesis: Most of the act of sort of going back to one's old job – that's kind of already happened. So, this is a question of people finding a new job. And that's just a process that takes longer. There may be something of a speed limit on it. You've got to find a job where your skills match, you know, what the employer wants. It's got to be in the right area. There's just a lot that goes into the function of finding a job. Powell’s comments lead to two further points: The act of going back to one’s old job for those on ‘temporary layoff’ is relatively straightforward. For job creation, this is the low hanging fruit, most of which has already been picked. Now comes the much harder part – finding jobs for those ‘not on temporary layoff’ whose numbers have barely declined from the peak (Chart I-6). Chart I-6For Job Creation, The Low Hanging Fruit Has Already Been Picked One way of encapsulating this is to observe that the unemployment rate – including those on temporary layoff – has already made 80 percent of the journey from its recession peak to the February 2020 trough, which makes it seem that the jobs recovery is largely done. However, the unemployment rate for those not on temporary layoff has made only 25 percent of the journey (Chart I-7). Moreover, this process is not a straight line, it is a curve. The first quarter of the journey is the easiest, then it gets harder. Chart I-7The Hard Part Is Finding Jobs For Those Unemployed 'Not On Temporary Layoff' As we, and Jay Powell, have pointed out, the process to reduce this unemployment rate has a remarkably consistent speed limit of 0.4-0.5 percent per year. Starting at the current rate of 2.5 percent and a target of 1.5 percent, this means full employment will not be reached before the second half of 2023. And even this assumes clear blue skies for the world economy through the next two years, which is a tall order. We conclude that the market pricing of a Fed funds rate lift-off in December 2022 is much too optimistic, making the December 2022 Eurodollar contract a good buy. The End Of Pandemic Restrictions Will Unleash Super-Spreaders On July 19, the UK will remove all its domestic pandemic restrictions – meaning no more facemasks, social distancing, and limits on the size of gatherings. This doesn’t mean that the pandemic is over in the UK. Far from it. The delta variant of the virus is rampant. Rather, with a large portion of the population vaccinated, the government is replacing state-imposed laws and regulations with a libertarian onus on personal responsibility. Given that Covid-19 is not going away, the UK strategy raises a fundamental question. Other than implementing a vaccination program, what role should a government take in containing the virus? In Who’s Right On The Pandemic – Sweden Or Denmark? we revealed two important findings: First, it is a misunderstanding that state-imposed restrictions cause the collapse in social consumption. This is a classic confusion between correlation and causation. The true cause of the recession is that a virulent disease focuses millions of people on self-preservation, shunning crowds and public places. But to the extent that the pandemic also leads to state-imposed restrictions, many people blame the slowdown on these correlated restrictions rather than on the underlying cause – the voluntary change in behaviour. Second, without state-imposed restrictions, the majority will voluntarily change their behaviour to avoid catching and spreading the virus, but a minority will not. When a virus is spreading, this is critical because a tiny minority of so-called ‘super-spreaders’ is responsible for most infections. Put simply, economic growth depends on the behaviour of the majority and in a pandemic the majority will voluntarily reduce their social consumption. This explains why libertarian Sweden and lockdown Denmark suffered similar contractions in their economies (Chart I-8). Chart I-8Libertarian Sweden Has Not Significantly Outperformed Lockdown Denmark... In contrast, containing the virus depends on restricting the minority of super-spreaders. Which explains why libertarian Sweden suffered a much worse outbreak of the disease than lockdown Denmark (Chart I-9). Chart I-9...But Libertarian Sweden Has Suffered Many More Covid-19 Casualties The worry now is that the end of state-imposed restrictions will unleash super-spreaders and super-spreading events. This will allow the virus to replicate, mutate, and create new variants which are potentially more transmissible and resistant to existing vaccines. Pulling together our three themes for the second half of the year, inflation is set to decelerate, job creation has a natural speed-limit, and super-spreaders of new-variant Covid-19 infections will create speed bumps in the economy. This means that:  Bonds will rally. The US dollar will rally. Growth stocks will outperform value stocks. US stocks will outperform non-US stocks Candidates For Countertrend Reversal This week, we present three candidates for countertrend reversal. First, the Brazilian real’s recent surge has hit expected resistance at 65-day fractal fragility. A good way to play a continued reversal is to short BRL/COP (Chart I-10). Chart I-10The Brazilian Real Is Correcting Second, within emerging markets, the strong rally in the Saudi equity market is vulnerable to a setback, especially versus other markets. A good way to play this is to short the Saudi Tadawul All Share index versus the FTSE Bursa Malaysia KLCI, given that the 260-day fractal structure is at the point of fragility that marked the major top in 2014 (Chart I-11). Chart I-11The Saudi Stock Market Is Vulnerable To A Setback Finally, coming full circle to short-term supply bottlenecks, one major beneficiary has been the Marine Transportation sector which, since February, has outperformed the world market by 70 percent. As the supply bottlenecks ease, this is vulnerable to correction, especially as the 260-day fractal structure is at the point of fragility that marked the major top in 2007 (Chart I-12). Chart I-12Underweight Marine Transportation Hence, this week’s recommended trade is to underweight Marine Transportation versus the market, setting the profit target and symmetrical stop-loss at 16.5 percent.   Dhaval Joshi Chief Strategist dhaval@bcaresearch.com Fractal Trading System Fractal Trades 6-Month Recommendations Structural Recommendations Closed Fractal Trades Closed Trades Asset Performance Equity Market Performance   Indicators To Watch - Bond Yields Chart II-1Indicators To Watch - Bond Yields - Euro Area Chart II-2Indicators To Watch - Bond Yields - Europe Ex Euro Area   Chart II-3Indicators To Watch - Bond Yields - Asia Chart II-4Indicators To Watch - Bond Yields - Other Developed   Indicators To Watch - Interest Rate Expectations Chart II-5Indicators To Watch - Interest Rate Expectations Chart II-6Indicators To Watch - Interest Rate Expectations   Chart II-7Indicators To Watch - Interest Rate Expectations Chart II-8Indicators To Watch - Interest Rate Expectations  

特別レポート

Highlights Complementing the US Political Strategy Quantitative Presidential Election Model, we introduce our revised Quantitative Senate Election Model. Our senate election model measures the probability of the incumbent party (Democratic Party) to retain the Senate in the 2022 midterm election. The model predicts that Democrats are slightly favored to retain control of the Senate, though it is too early to call, which in combination with the high likelihood that the GOP will retake the House, points to a US political gridlock from 2023 to 2025. The “Blue Sweep” policy setting will end as early as the end of the year as Democrats pass Biden’s signature legislation. Post-midterm gridlock implies that taxes unlikely to rise further from 2023 while spending will not be subjected to cuts. While markets will not be alarmed if growth keeps up, near term surprises from potential tax hikes, rate hikes, and China’s slowdown warrants a more defensive positioning. Feature 2020 was not only the year of a highly contested US Presidential election, but also a close-knit battle for control of the US Senate, which had 351 seats up for reelection. The Republican party initially retained control of the Senate at the start of 2021 and the 117th congress, but this was short-lived. The Democrats secured victories in both run-off triggered Senate races in the state of Georgia, putting them at an even 50-50 hold with Republicans in the Senate. The inauguration of Vice President Kamala Harris who too became the Senate President, was the tie breaker the Democrats needed to take control of the Senate, and ultimately secure a “blue sweep” of holding the House of Representatives, the Senate and the White House. We recently introduced BCA US Political Strategy readers to our quantitative presidential election model. If you have not yet read it, you can access it here. In this week’s report we introduce our US Political Strategy Senate election model. We acknowledged that it was still early days in the presidential election cycle when we published our presidential election model but there were however some interesting takeaways from an early model forecast. For control of the Senate, however, the cycle is much shorter, with voting of one third of the Senate taking place every two years. The mid-term elections of 2022 are not that far-out, and with 34 seats up for reelection, we believe that introducing our readers to our Senate election model now will start to provide valuable insight going forward. Like our presidential election model, our Senate election model is a state-by-state model that uses both economic and political variables to predict the number of seats the incumbent party will win in the 2022 Senate election. Our Senate model covers a large sample size, consisting of 19 Senate elections (1984 to 2020), across 50 states, amounting to 950 observations. The Six Variables Our Senate model is based off a Probit regression that produces a 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 the Senate election in each state; or 0 = Incumbent party did not win the Senate election in each state. This method allows us to measure the probability that a state with certain characteristics will fall into one of two categories above. We can then predict the probability of the incumbent party winning all the Senate seat/s in each of the 50 states (although this is only relevant to one-third of the states that have a Senate seat up for election in 2022). 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. The incumbent party’s margin of victory in previous Senate elections in each state Senate race. This is measured as the incumbent party’s share of the popular vote minus the non-incumbent party’s share. If the incumbent party failed to secure a solid win in each state in the previous Senate election, the probability of securing a solid win in the current election becomes smaller. Moreover, the larger the margin of victory in a previous Senate election race, the more likely that incumbent party will win re-election in said state. Net average approval level of the incumbent president in a Senate election year. This is the difference between the incumbent president’s approval and disapproval level in a Senate election year, from the start of the year up until the end of October of that year – taken as an average. Generic congressional ballot (net support rate). The generic congressional ballot asks people which party they are likely to vote for in Congress. We take the average net support rate in a Senate election year (that being whichever party leads the other in congressional ballot polling). Democrats are usually favored in congressional generic ballot voting, so the net rate is more predictive than the gross rate Dummy variable for congressional ballot. A dummy variable is assigned to variable number four. For example, dummy takes the value of 1 when Democrats have a positive net support rate in generic congressional ballot voting, and 0 when Republicans have a net positive support rate. We assign only one dummy variable to avoid a dummy variable trap.2 A “time for change” variable, a categorical variable indicating whether the incumbent party has controlled the Senate for three or more terms (six or more years). If the Senate has been controlled for three or more terms, the model will “punish” the incumbent party, as we would expect to see a change in control of the Senate the longer one incumbent party controls it. Democrats Retain Control Of The Senate As it stands, our election model predicts that Democrats will retain control of the Senate in 2022 (Chart 1). The Democrats are predicted to win 49 seats, a gain of one seat over the 2020 Senate election outcome,3 and when coupled with the two seats of Independent Senators, give them a majority of 51 seats. Chart 1Quant Model Gives Democrats 54% Chance Of Retaining The Senate The additional seat for Democrats stems from our model allocating both North Carolina and Pennsylvania (which are currently occupied by Republicans) to the Democrats (+ two seats) and allocating one of Georgia’s seats occupied by Raphael Warnock4 back to Republican control. The Democrats overall probability of retaining control of the Senate is 54%, three percentage points higher than early market predictions (Chart 2). The market implied odds highlight another close battle between Democrats and Republicans to control the Senate in 2022. Chart 2Market Narrowly In Favor Of Democratic Senate Control North Carolina is the only toss-up state,5 with a 51% chance of a Democratic victory. Pennsylvania will switch to Democrats and Georgia to Republicans. Note that North Carolina and Pennsylvania are both currently under Republican control. Both incumbents have decided not to run again. While both Georgia Senate run-off races were won by Democrats earlier this year, the sum of first-round voting in November 2020 was higher for Republican candidates than for Democrats. There was also extra-ordinary voter turnout in favor of Democrats for both run-offs, which ultimately played a big role in Democrats securing victory. Voter turnout was largely spurred on by voting against Republicans, and ultimately Donald Trump. This may not be the case come 2022, if turnout for Democrats is unmatched to 2020/2021. Our model’s prediction will evolve over time as new data become available, which could produce more toss-up states, or swing the prediction in favor of the opposing party. For now, the model provides us with a preliminary prediction as we draw nearer to the 2022 midterm elections. Senate Races Of Interest Comparing our model’s prediction to online betting markets, we group nine races into a category of “interest”. All nine races have varying degrees of probability for a Democratic win, ranging from approximately 30% to 60%. Five races are overestimated, and four races are underestimated by consensus (Chart 3). The remaining 25 races are decidedly in favor of either Democrat or Republican control, according to our model, so are therefore excluded from this analysis. Betting markets are overestimating Nevada, Arizona, Pennsylvania, Georgia, and Wisconsin, while underestimating New Hampshire, North Carolina, Florida and Ohio. Chart 3Senate Odds Compared With The Bookies All nine of these races are precariously balanced, even at this stage of the mid-term election cycle. Small or local factors could ultimately decide the outcome. This is an important limitation on our macro model, highlighting our ultimate emphasis on qualitative analysis. For example, it is not at all clear that Democrats will win Georgia. Our model gives Democrats a 43% chance of victory. Betting markets are a lot more optimistic, penning a 55% chance of a Democratic win. But even by our model’s standard, Georgia remains a toss-up. Georgia may not be as close of a race as it was in 2020/2021, if voters are not as motivated as they were to vote Democrat. Will turnout be as large in 2022? That remains to be seen. One or two races with unique makeup can contribute to maintaining or shifting the balance of power in the Senate come 2022. Back Testing Our Model Our Senate model performs at an acceptable level during in-sample and out-sample back testing. For in-sample testing, we test our model over our entire sample period (1984 – 2020) and find that 74% of Senate elections (control of the Senate) are correctly predicted, with the model predicting the outcome of the last five Senate elections correctly (Chart 4). Chart 4In-Sample Back Testing Results During out-sample back testing, we look at a sample period of 2000 – 2020, comprising of 11 Senate elections, where our model correctly predicts 73% of actual outcomes. The previous five Senate elections are predicted correctly too (Chart 5). Chart 5Out-Sample Back Testing Results In comparison to our presidential election model, prediction accuracy of our Senate model is lower across its sample period. Predicting control of the Senate can sometimes be more uncertain than that of the White House. Both statistical and event based (Senate elections) reasons give way to a lower accuracy rate in this case. For example, there could be several idiosyncratic state-level variables not captured by our model, which could have played a leading role in determining any one state’s Senate election outcome over our sample period, and ultimately, control of the Senate. Where To From Here? In comparison to the presidential election cycle, we are a lot closer to election day. That means that Senate races will begin to heat up as we move closer toward November 8, 2022 – the date of the midterm elections. For now, our model ratifies the current control of the Senate, that is, Democratic. Our Model also suggests that come 2022, the Democrats will retain control of the Senate. But this is all but an early forecast. If any long-standing conclusion can be drawn right now, it is that the battle for control of the Senate in 2022 will be highly contested. From a qualitative point of view, our model may be overestimating the Democrats’ odds in 2022 as things stand today. Midterm elections have historically seen the sitting president’s party lose seats in the Senate and House of Representatives. We already expect Republicans to retake the House after a poor showing by Democrats in 2020. This narrative may play into the Republicans taking the Senate too – and is plausible given how closely the battle for the Senate is wound. But congressional approval has ticked higher lately under a Democratic run congress (Chart 6). Most likely, the American public have largely approved of COVID-19 government relief, and the Democrats will pass at least one more major piece of legislation covering infrastructure. Republicans are deeply divided, so there is some chance that they underperform in 2022. Nevertheless, the historical pattern clearly favors the opposition. The takeaway is to expect the GOP to retake the house but to monitor the Senate closely with both quantitative and qualitative tools. Chart 6US Public Approving Of Congress ?!? Lastly, and importantly, we should note that in both the case of the presidential and Senate models, a probability between 50% and 55% for the incumbent party retaining control of the White House or Senate is indicative of an outcome “too close to call.” Both models are touting Democratic wins, but high conviction views about either the 2024 presidential election or 2022 Senate election are not warranted at this time. Investment Takeaway Unless 2022 is one of the rare cases of an incumbent party legislative victory after a national shock, like 1934 and 2002, Republicans will take the House at least. This is likely notwithstanding our model’s slight tilt in favor of Democrats in the Senate. This means that the “Blue Sweep” policy setting will cease as early as 2023, but de facto it would cease as early as the end of this year when Biden’s signature legislation is passed, since Congress will get little done in 2022. Our model suggests Republicans are slightly disfavored in the Senate. The truth is that as long as they gain one chamber of the legislature then US fiscal stimulus will virtually freeze. Taxes will no longer be able to rise from 2023 but spending will not be subject to cuts. Gridlock is reinforced by our presidential quant election model’s slightly higher odds of Democrats retaining the White House, which we think is underestimated at present. Hence Biden will retain veto power even if Democrats squander the Senate and House in 2022. Gridlock is thus looming from 2023 until at least 2025. The financial markets will not be alarmed by this forecast as long as growth keeps up. In the very near term, however, the clouds on the horizon of tax hikes, Fed rate hikes, and China’s tight-fisted economic policy pose rising headwinds to US equities in 2022 — and hence markets should respond negatively sooner than later. We are tactically growing more defensive.   Guy Russell Research Analyst GuyR@bcaresearch.com   Statistical Appendix Some clients may be curious as to how our US Political Strategy Senate election model differs from our Geopolitical Strategy model used in the 2020 elections, and where it has made improvements in its predictive accuracy. We discuss these improvements herein. Changes To The Geopolitical Strategy Senate Election Model A notable property in our dependent variable data requires a brief discussion. Our dependent variable classified as “elected” takes the form of a binary outcome. This data, however, is what’s called “unbalanced,” since incumbent Senators are re-elected approximately 80% of the time. This means that most outcomes in our dependent variable are coded as “1,” with fewer “0’s” because of the strong incumbency effect in Senate races. There are many data sets that exhibit this type of property, such as events like wars, vetoes, cases of political activism, or epidemiological infections, where non-events occur rarely. To alleviate this statistical property in the data, we estimate our model using a weighted maximum likelihood estimate as opposed to the ordinary maximum likelihood estimate usually used in a Probit regression.6 This method assigns more weighting to the unbalanced data, or what is known theoretically as “rare event” data, to aid the Probit regression in assigning higher probabilities to “0” outcomes. Through this process, we effectively deal with our unbalanced dependent variable data. The last update to the BCA Geopolitical Strategy Senate election model was published on January 6, 2021. Our model suggested that Republican’s would retain control of the Senate. Our model was limited in dealing with a unique twin Georgia run-off race that ultimately swung Senate control into the hands of the Democrats. The Geopolitical Strategy, which we will refer to as the 2020 model, only missed the Republican victory in Maine, but correctly predicted losses in Arizona and Colorado. The model missed both Georgia races, signaling they would remain red states – this was proven otherwise. Also, our model has become a better predictor in terms of in and out-sample forecasting (compared to our 2020 model). The 2022 version correctly predicts 74% (vs 72%) of in-sample and 73% (vs 70%) of out-sample outcomes. Methodology And Variables Our Senate model retains the methodology and suite of economic and political variables used in the model we first introduced in 2020. For long-time clients and those who are new to the US Political Strategy and Geopolitical Strategy service, the first version of our model can be found here. 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 term preceding a Senate 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, they all remain the same as the 2020 model. Model Performance Classification The 2022 model correctly classifies predicted outcomes at a rate of exactly 81%. That is, when the model makes a prediction of a certain state’s Senate election outcome from 1984-2020, it is correct 81% of the time. This level of classification is higher than our 2020 model, which classified outcomes at a rate of 79% (Table 1). Table 1New Model Classifies Outcomes At A Higher 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 Senate election in each state; or 0 = Incumbent party did not win the Senate election 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. A robust model has an 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. Chart 7Receiver Operating Characteristic Curve Of 2022 Model Table 2… Is A Better Fit … The ROC curve for our 2022 model has an AUC of 0.9609 (Chart 7), a higher AUC than our 2020 model (Table 2). This means that the true positive rate for classifying outcomes is high and the false positive rate is low, improving on our model’s robustness. F1 Scores A final grading of the 2022 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 2022 model produces a higher F1 score compared to our 2020 model (Table 3). Table 3… And Is More Accurate Than The 2020 Model Considering the improvement in forecast accuracy and overall better model specification over our 2020 model, we accept our 2022 model as our new base case Senate 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. Appendix Tables Table A1USPS Trade Table Table A2Political Risk Matrix Chart A1Presidential Election Model Table A3APolitical Capital: White House And Congress Table A3BPolitical Capital: Household And Business Sentiment Table A3CPolitical Capital: The Economy And MarketsTable A4Political Capital Index Footnotes 1     Two of which were open Senate seats for the state of Georgia. 2     A dummy variable trap is a scenario in which the independent variables are multicollinear — a scenario in which two or more variables are highly correlated; or, in simple terms, in which one variable can be predicted from the others. To avoid such a trap, we must exclude one of the categorical variables. Since there are two categorical variables that can be represented here (Republican or Democrat), we use k-1 (where k = the number of categorical variables). 3    In reference to the Senate election outcome after the Georgia run-off races which concluded in early January 2021. 4    This seat formed part of the 2020 special Senate election race which was decided by a run-off election between Raphael Warnock and Kelly Loeffler. The seat was always up for reelection in 2022 no matter which party won it in the 2020 special election. 5    Toss-ups are defined as having a probability between 45% and 55% according to our model. 6    Weighted maximum likelihood estimation is a reasonable approach in dealing with dependent variables that show significant imbalance in their data set. See: King, G. and Zeng, L., 2001. Logistic regression in rare events data. Political analysis, 9(2), pp.137-163.