美国经济分析：QE成本目前看似不高0129

2013 年 1 月 26 日Issue No: 13/04美国经济分析研究报告QE 成本目前看似不高在联邦公开市场委员会（FOMC）12 月份会议之后的新闻发布会上，美联储主席伯南克表示委员会将持续、谨慎地评估定量放松(QE)的成本与益处。QE 的潜在成本包括：打乱通胀预期，造成资产价格泡沫等金融失衡状况，减弱美联储未来退出当前相当宽松的货币政策的能力，并干扰金融市场的正常运作。然而，我们认为目前以上问题尚且不明显，且短期内这些标准不会成为 QE 继续实施的障碍。尽管 2008 年以来美联储资产负债表显著扩张，但无论是从 TIPS 债券所隐含的均衡通胀率，还是从基于调查的通胀指标来看，通胀预期一直保持稳定。虽然 QE 计划对国债和 MBS 的估值已经产生了显著影响，但美联储可能并未看到更为广泛的风险资产估值过高的迹象。即便在准备金很高的情况下，美联储也应有能力退出宽松的货币立场。尽管在所持证券规模正常化的过程中，美联储可能会出现较大亏损，但在货币政策制定者所要考虑的事项中，资产负债表政策对于财政部相关收入的影响可能不大重要。由于最近几年财政部这部分收入显著上升，任何潜在的亏损只可能将其部分抵消。Jan Hatzius(212) 902-0394 高盛集团Alec Phillips(202) 637-3746 高盛集团Jari Stehn(212) 357-6224 高盛集团Kris Dawsey(212) 902-3393 高盛集团David Mericle(212) 357-2619 高盛集团Shuyan Wu(212) 902-3053 高盛集团Michael Cahill(801) 884-4621 高盛集团面对美联储的大规模购买，国债市场似乎并未承受太大压力。尽管有迹象表明大量购买 MBS 似乎对市场的正常运转构成考验，但我们认为目前还没到美联储官员应该考虑缩减计划规模的时候。投资者不应视本报告为作出投资决策的唯一因素。 有关分析师的申明和其他重要信息，见信息披露附录，或参阅 年 1 月 26 日美国经济分析Costs of QE Look Minor so FarAt the press conference following the December FOMC meeting, Chairman Bernanke noted thatthe Committee would carefully evaluate the costs and benefits of QE on an ongoing basis. In fact,the Chairman mentioned the potential costs of QE 11 times during his prepared remarks and theQ&A, devoting somewhat more time to this topic than he has in the past. For example:“The Committee expects this policy tool large scale asset purchases to continue to be effectiveand the costs and risks to remain manageable, but as the program continues, we will be regularlyupdating those assessments. If future evidence suggests that the programs effectiveness hasdeclined, or if potential unintended side effects or risks become apparent as the balance sheetgrows, we will modify the program as appropriate. More generally, the Committee intends to beflexible in varying the pace of securities purchases in response to information bearing on theoutlook or on the perceived benefits and costs of the program.”Potential costs of QE generally fall into five categories: (1) un-anchoring inflation expectations, (2)creating financial imbalances such as asset price bubbles, (3) reducing the Feds ability to exitfrom the current highly accommodative stance of monetary policy in the future, (4) creating lossesonce exit begins, and as a result reducing remittances to the Treasury Department, and (5)impairing the normal functioning of financial markets. There are also distributional effects of QE,such as disadvantaging retirees who rely most heavily on interest income. However, those whoare disadvantaged (savers) are probably roughly balanced by those who are advantaged(borrowers), and so the overall impact of these distributional effects is less clear. In this weeksUS Economics Analyst, we provide our assessment of these five categories of costs, and outlinea set of indicators that analysts can use to assess whether the costs of QE are shifting over time.While we expect that none of these criteria will be an impediment to continued QE in the neartermin fact we expect QE to continue through 2013 and, at a reduced pace, through 2014 aswellthese issues certainly bear watching over time. But taken as a whole, the costs of QEappear modest at present.Inflation Expectations Remain Well AnchoredThe potential cost most directly linked to the Feds statutory mandate relates to inflationexpectations. Some worry that the large increase in base money (the Feds balance sheet) since2008 may translate into rapid growth in broader measures of the money supply in the future, andeventually result in unanchored inflation expectations or runaway inflation. However, to date thisconcern has not materialized in traditional measures of inflation expectations, and underlyinginflation trends remain quite subdued.Exhibit 1 shows two measures of longer-dated forward inflation expectations, which likely providethe best read on public perceptions of the Feds inflation credibility. Short-term measures ofinflation expectations tend to be heavily influenced by fluctuations in gasoline prices, and henceare of less interest for the purposes of this analysis. The first measure is the forward inflation“breakeven” derived from TIPS marketsthat is, the rate of inflation that would need to occur tomake a hypothetical risk-neutral investor indifferent between holding nominal and inflation-protected securities. The second measure is the median 5- to 10-year ahead rate of inflationexpected by respondents to the University of Michigans survey on consumer sentiment. Clearly,both of these measures are currently within the ranges seen over the past four years, and indeedare within ranges seen during the pre-crisis period as well, providing no evidence that inflationexpectations have become unanchored as a result of QE.The fact that both measures are somewhat above the Feds 2% inflation target can be attributedto 1) the fact that TIPS breakevens refer to CPI inflation while the Fed targets PCE price indexinflation, and CPI inflation tends to track above PCE price index inflation for technical reasons, 2)the likelihood that investors in TIPS are not risk neutral, and hence an inflation risk premium is高盛全球经济、商品和策略研究1132013 年 1 月 26 日美国经济分析also included in the breakeven, and 3) the well-established historical wedge between consumerinflation expectations and measured inflation from official statistics.Exhibit 1: Inflation Expectations Remain AnchoredPercentage pointsPercentage points4.03.5QE1QE1ExpansionQE2Operation TwistQE34.03.53.02.53.02.52.01.55-Yr 5-Yr Forward BreakevenUmich 5-Yr Ahead Expectations2.01.5Jun2008DecJun2009DecJun2010DecJun2011DecJun2012DecSource: Federal Reserve Board, Bloomberg, University of Michigan.We expect that the high level of slack in the economy, and in particular in labor markets, willresult in continued subdued inflationary pressures in coming years, with the core PCE price indexlikely rising around 1.5% in 2013. Specifically, it is hard for us to see wage inflation picking upwith 7.8% unemployment, and it is hard for us to see sustained and significant increases in priceinflation without a pickup in wage inflation. Our past research also indicates that inflationexpectations have become increasingly anchored in recent years around the Feds 2% target.Supply-side shocks present a risk to this view, although we would expect shocks in oil prices, forexample, to affect inflation only over the relatively short term.Risky Asset Bubbles Not Readily ApparentA second potential cost of QE is the possibility that the programs may result in financialimbalances, which we take to mean bubbles in real or financial asset prices. Monetarypolicymakers are probably particularly attentive to this issue, given the criticism the Fed receivedfor maintaining easy policy while the housing bubble (driven at least in part by adjustable ratemortgages) was inflating. That said, an intended effect of the programs is to raise risky assetprices relative to what would otherwise be the case.Stehn, Jari. “A Flatter and More Anchored Phillips Curve” US Economics Analyst. November 10, 2012.高盛全球经济、商品和策略研究2242013 年 1 月 26 日Exhibit 2: Equity and Credit Risk Premium Are Not Historically Low美国经济分析Percentage points8Equity Risk PremiumPercentage points876543210BBB Non-financial Credit Risk Premium76543210199719992001200320052007200920112013Source: GS Global ECS Research.Exhibit 2 shows estimates of risk premia (expected excess returns over the risk free rate) in USequities and investment grade (BBB) corporate credit from GS Global ECS Research. Ifvaluations were particularly stretched relative to fundamentals, we would expect to see very lowlevels of these risk premia, which is clearly not the case. In terms of real assets, our models forhouse prices suggest that residential real estate values are under-shooting fundamentals. Thequestion of whether Fed balance sheet policy is inflating commodity prices, which was moresalient during QE2, seems to be less of a concern now as well. Oil prices have not continued torise on net over the past several years, likely being driven mainly by factors largely unrelated toFed policy.Apart from these widely followed asset prices, some Fed officials have expressed concern thatnon-traditional assets, and in particular farmland, are becoming overvalued, potentially as anunintended consequence of overly-easy monetary policy. USDA researchers do note that“historically low interest rates are likely a significant contributor to farmings current ability tosupport higher land values,” and hence this issue bears monitoring, although farm income hasalso risen strongly in recent years as land prices have risen. Despite some potential warningsigns in agricultural land, it seems unlikely to us that Fed policy has created significant “financialimbalances” in broader asset markets. Of course asset bubbles have a tendency to appear moreobvious with the benefit of hindsight.Fed Can Still Take Away the Punch Bowl, Even with High ReservesA third potential cost of QE relates to whether the large size of the Feds balance sheet wouldmake it more difficult to withdraw monetary accommodation when the time comes. One aspect ofthis question is practical: does the Fed have the tools to raise interest rates even with a very highlevel of reserves in the system? The other aspect is institutional: would the Fed be willing towithdraw accommodation even if it meant negative income and cancelled remittances to theTreasury Department? The potential size of these losses could be expected to grow with the sizeof the Feds balance sheet. We think the answer to both of these questions is yes, although thereis more risk on tools than on intent.Nickerson, Cynthia, Mitchell Morehart, Todd Kuethe, Jayson Beckman, Jennifer Ifft, and Ryan Williams.“Trends in US Farmland Values and Ownership.” Economic Information Bulletin Number 92. USDAEconomic Research Service. February 2012.高盛全球经济、商品和策略研究3352013 年 1 月 26 日美国经济分析The federal funds rate is the interest rate at which banks borrow and lend reserve balances at theFed. With an enormous supply of excess reserves (this is the liability-side impact of the largeincrease in Fed securities holdings through QE), the market clearing price is near zero at present.It is also likely that there will still be an unusually high level of reserves when the Fed intends toraise rates, in our view, around early 2016. And the more QE we have between now and then,the higher the level of reserves in the system come 2016, all else equal. Interest paid on excessreserves (IOER), a tool which the Fed did not have at its disposal until the financial crisis, shouldin theory act as a floor on the fed funds rate, allowing the Fed to raise the effective rate despite ahigh level of reserve balances. Essentially, participants in the fed funds market should not bewilling to lend at a lower rate than IOER, if they can earn this return on their reserves by leavingthem in their own accounts. However, there is something wrong with this picture, as the fed fundseffective rate has averaged around 14 basis points recently, despite IOER currently at 25 basispoints. The floor seems to be leaky due to some participants in the fed funds marketinparticular the GSEsbeing ineligible to earn IOER, and willing to lend at rates below thetheoretical floor. As a result, there is some uncertainty about exactly how effective IOER wouldbe at putting a floor under rates when the time comes to tighten policy, although we anticipatethat these issues would be worked out in due time. If this became a serious issue, we expect thatthe rules would be changed by whatever means necessary, well before monetary policy would besignificantly impaired.The Fed has a number of other tools that it can use to withdraw policy accommodation whenwarranted, in addition to IOER. Term deposits, reverse repos with an expanded set ofcounterparties (including money funds), and outright asset sales might also be used to withdrawreserves from the system. The Feds diversified portfolio of exit strategies reduces the risk of anunwelcome shock should one tool prove to be less effective than expected.The Fed Will Probably Lose Money When it Exits, but Come outAhead for the Programs as a WholeThe impact of exit on Fed earnings and hence remittances to the Treasury Department may besignificant (and more significant the larger the Feds balance sheet). Two issues are at play here:first, when the Fed wants to raise rates, its interest expense associated with paying IOER will rise,while its portfolio income will not essentially its net interest margin will be squeezed. Second, ifthe policy works as intended and the US does not end up like Japan, the Fed is currently “buyingat the top” and will likely be selling its portfolio of securities in a rising rate environment. This willprobably generate some capital losses, although the market value of the SOMA portfolio iscurrently more than $200 billion in excess of book value. While losses as the Fed normalizespolicy could be significant, the impact of balance sheet policy on Treasury remittances is probablylow on the list of considerations for monetary policymakers. Remittances are at record high levelsat the moment, but they still represent a very small share of federal revenues, so the broaderimpact of Fed losses on the budget deficit should any be relatively modest (Exhibit 3).The Federal Reserve is required to remit all income earned over and above dividends to member banks,required capital contributions, and operating expenses to the Treasury Department. Most of this incomehas come from interest earned on the SOMA portfolio in recent years. For example, in 2012 the Fedremitted $89 billion to the Treasury Department.高盛全球经济、商品和策略研究454562013 年 1 月 26 日Exhibit 3: Fed Remittances At Record Levels, Small Share of Federal Budget美国经济分析Billions of dollars100Percent4.09080706050403020100Remittance Amount (left)Percent of Federal Revenue (right)3.53.02.52.01.51.00.50.019901995200020052010Source: Office of Management and Budget, Federal Reserve Board.A recent study by several Fed economists ran projections of the Feds net income and Treasuryremittances over the coming years. For the scenario in which the balance sheet rises by a further$1 trillion during 2013 remittances would be cancelled from 2017-2020 and the “deferred asset”(negative balance on interest payable to Treasury) would reach a peak of $40 billion. While thelosses generated during this period would be modest relative to the total income generated overthe course of the programs, and while a reduced rate of sales relative to the authors assumptionsmight diminish these losses to some extent, we believe that losses on sales and temporarilynegative net income could result in some unfavorable headline risk for the Fed during this period.Markets still functioningFinally, some have raised concerns that the large scale of asset purchases could impair marketfunctioning. While the exact definition of market functioning is open to interpretation, we take it tomean the ability for non-Fed market participants to transact with one another in a normal mannerwith respect to liquidity, asset scarcity, price impact of trades, and similar metrics not directlyrelated to macro valuations. The larger the share of Fed purchases relative to issuance andoutstanding amounts, the more likely purchases are to impair market functioning. Althoughdetailed Fed commentary on the issue of market functioning has been scarce, ChairmanBernanke stated in his 2012 Jackson Hole speech that increasing Fed ownership might makeTreasuries less liquid, eroding the value of the liquidity premium and running counter to the goalof the program of reducing yields.In the Treasury market, measures of market functioning have held up well despite Fed purchasesin the long end being comparable in size to gross issuance of new bonds (Exhibit 4). Borrowingof securities from the SOMA portfolio by dealers is at quite modest levels, according to data fromthe New York Fed, suggesting that SOMA purchases have not heightened the scarcity value ofthese securities.Bid-ask spreads are relatively normal. In contrast to worries that FedCarpenter, Seth, Jane Ihrig, Elizabeth Klee, Alexander Boote, and Daniel Quinn. “The Federal ReservesBalance Sheet: A Primer and Projections.” Finance and Economics Discussion Series paper 2012-56.Federal Reserve Board.The fact that the Feds markets desk excludes securities trading special in repo from the list of eligibleCUSIPs for its purchase operations demonstrates that the Fed views increasing asset scarcity of securitiesalready trading special as something worth avoiding.高盛全球经济、商品和策略研究72013 年 1 月 26 日美国经济分析purchases might cause relative prices of high vs. low SOMA ownership securities to diverge,relative value differences across the curve have narrowed significantly since the Fed startedTreasury purchases, probably due to significant purchases of off-the-run securities. Finally, thereis little evidence that the flow of Fed purchases is having an increasing transient price impact.Exhibit 4: Fed is Purchasing a Significant Share of Gross Treasury IssuancePercent of gross issuancePercent of gross issuance120100806040200-20-40SOMA Purchases:Nominal 2-5Nominal 5-10Nominal 10TotalForecast120100806040200-20-4020092010201120122013Q1Note: Excludes TIPS and Bills. All re-issued securities counted as original time to maturity.Source: Bloomberg, Federal Reserve Bank of New York, GS Global ECS Research.In contrast to the case of Treasuries, purchases of MBS appear to be having more substantialimpacts on market functioning. Specifically, MBS dollar rolls, which are a form of repurchasetransaction in which a party commits to sell an MBS in the current month and to buy back anequivalent security in a future month, are trading special, with an abnormally large drop betweenthe current and forward settlement date prices. Roll specialness is commonly a signal of supply-demand imbalance in MBS markets. In the present case, the specialness is likely related to thelarge demand for mortgage pass-throughs coming from QE asset purchases. Additionally, thefunctioning of the primary mortgage market is in some sense “impaired” in that the spreadbetween primary mortgage rates available to end borrowers and secondary mortgage ratesearned by investors has remained quite wide in recent yearsas declines in secondary MBSrates have not fully passed through to end borrowerssuggesting profit opportunities for newentry in mortgage origination. While this is not an issue caused by Fed purchases themselves,and is likely related to industry capacity constraints, it does likely hamper the ultimateeffectiveness of the MBS purchase program to some extent.FOMC likely sees benefits outweighing costs for foreseeable futureWhile the potential costs of QE are worth careful consideration, we expect that the FOMC will stillcome down squarely on the side of QEs benefits exceeding its costs for the foreseeable future(Exhibit 5). As a result, we anticipate that securities purchases will continue at the current rate of$85bn / month through 2013, and at a reduced rate of $50 bn / month during 2014. However, iffuture analysis suggests that QE has diminishing returns, while costs may be convex, thiscalculus could change.高盛全球经济、商品和策略研究82013 年 1 月 26 日Exhibit 5: FOMC likely sees benefits outweighing costs for foreseeable future美国经济分析Potential CostUn-anchored inflation expectationsFinancial imbalancesReduced ability to exitLower remittances to TreasuryImpaired market functioningLikely View of FOMC LeadershipNon-issue at present, but requires careful monitoring.No obvious bubbles in major risky asset classes. Consider systemicimplications of froth in non-traditional markets.Likely has the tools to exit regardless of the level of starting balance sheetsize.Losses will occur if the program is successful. The extent of those losseslikely not a major concern for the Fed.More apparent in MBS market, but not bad enough to cause Committee topare back the program.Source: Goldman Sachs Global ECS R