1 F ebruary 2019

Via Em ail: [email protected]

Dr. Shane Worner International Organization of Securities Commissions Calle O quen do 12 28006 M adrid Sp ai n

Re: Comm ents on IOSCO Report: Leverage

D ear Si r ,

AQR Capital M anagement, LLC (“AQR”) welcomes the opportunity to respond to the International Organization of Securities Commissions’ (“IOSCO”) consultation report on assessi n g th e use of leverage by in vestm en t fu n ds (th e “ Report”). W e are very grat efu l to h ave had the opportunity to contribute to IOSCO’s work on this im portant topic over the past 15 months, speci fically i n bei n g able to at ten d I OSCO m eet in gs an d m eet I OSCO coun try represen t atives. T h is en gagem en t has given us an appreciation for the im port an t ch allen ges presen t ed to I OSCO and its constituent national regulators in im plem enting the Financial Stability Board’s (“FSB”) Recom m en dation s an d w e com m en d I O SCO for i ts th ough t ful an alysi s i n th e Report addr essi n g Recom m en dat i on 10, t h e iden ti fication an d developm en t of con si sten t m easures of leverage for fin an cial st abili ty purposes (th e “R ecom m en dation ”).

Background to AQR

E st ablish ed in 1999, A Q R is a global asset m an agem en t firm based in G reen w ich , Con n ecticut , Un ited States1 and as at 31 December 2018 has approximately $200 billion in assets under m anagem ent across both traditional long-only equity strategies and alternative investm ent strategies, of which the firm is one of the world’s largest providers. We offer investors access to

1 With additional offices in , , and Hong Kong.

AQR Capital Management, LLC | Two Greenwich Plaza | Greenwich, CT 06830 | U.S. | p: +1.203.742.3600 | f: +1.203.742.3100 | w: aqr.com 2

th ese i n vestm en t s th rough separ ate accoun ts, U CI T S fun ds, L u xem bourg SI F s, offsh ore an d onshore pooled vehicles, and through sub-advised mutual funds and U.S. open-end mutual funds.

A QR h as been a leader i n offeri n g alt ern ative in vestm en t strat egi es as regist ered m utual an d U CI T S fun ds an d em ph asi zes ri sk con trol an d best practi ces i n fun d govern an ce. W e possess a substan ti al understanding of derivatives, securi ties fin an ce an d alt ern ative in vestm en t st rat egi es, w h ich w e beli eve positi ons us well to com m ent knowledgeably on the Recom m endation and m ore speci fically th e R eport.

Object i ves of t h e R ecom m en dat ion an d Repor t

W e are m i n dful of th e specific objectives set out by the FSB in the Recommendation that m easures of leverage ou gh t to be con si sten t, com par able across fun ds an d to th e exten t possi ble across sectors, m inim ize the shortcom ings of existin g m easu res, additi ve an d facilit at e m ore m ean in gful m on itorin g of leverage for fin an cial st ability purposes. W e n ote th at in th e R eport IOSCO h as also suggest ed th at key m et rics sh ould be of gen eral application across st rategi es an d funds, avoid model risk and facilitate the identification of funds which may pose a risk to financial stability.

W e furth er recogn ize th at on e of th e cen tral ch allen ges posed by th e R ecom m en dation is th e ten sion bet w een a preferen ce for th e sim plicity of a si n gle m easure of leverage such as total gross notional exposure, which can be misinformative, and m odest ly m ore com plicated, yet inform ative m easurem en ts th at en able regulators to meaningfully assess sources of syst em i c ri sk.

Suggest ed A pproach

W e beli eve th at in large part, the Report can ach ieve th e objectives of t h e R ecom m en dation th rough th e proposed t w o step fram ew ork. I n particu lar, th e R eport clearly an d accurat ely articulat es th e st ren gth s an d w eakn esses of th e th ree key pot en ti al “Step On e” m etrics for m easurin g leverage, as w ell as the need to analyze each m etric by asset class rather than as a sin gle aggregat e n um ber. W e support th e overall approach proposed in th e R eport , an d i n th e rest of th is lett er provide ou r suggest ed refin em en ts to th e “St ep O n e” process w ith in th e Report’s broader construct.

The approach w e advocate for m eets several, if n ot all, of th e objectives set out i n th e Recom m en dation an d Report. I t is also relatively sim ple, con sist en t , m odel in depen den t an d importantly, allows regulators to enhance their underst an di n g of poten tial risks that leverage m ay 3

pose to financial stability by allow i n g i n form at i ve sum m ati on across fun ds th at distinguishes betw een exposures to m ean in gfully di fferen t risks in securiti es an d deri vatives m arket s.

A v oi d a si n gl e a ggr ega t ed n u m ber

The most important aspect of the “Step One” process is that it excludes any gross notional leverage m etric based on an aggregated single num ber. It is our strongly held view that aggregation is the fatal step that destroys the value of any n oti on al m easur e as i t adds up n u m bers that do not reflect even rough ly com parable risks, allowing important risks to be disguised within broader benign risks. F or th i s r eason , aggr egat ed gross n oti on al i s a useless sum of usefu l part s.

T h ere are vast differen ces in th e relati ve riski n ess of un derlyi n g asset types i n derivatives contracts an d securiti es w h ich m ake a sin gle aggregat ed leverage n um ber m ean in gless in term s of assessi n g th e riskin ess or th e speculative ch aracter of a fun d. A si n gle aggregat ed n um ber m i xes exposures to high risk asset classes with exposu res to low ri sk asset classes an d m akes i t m ore difficult to differentiate between hedged exposures and unhedged exposures, or between portfolios th at use leverage to am plify risk and on es that use it to reduce risk.

Maintaining the distinction between high and low risk exposures by reporting on each underlying asset class en ables m ore m ean i n gful di sclosu res that do not allow exposures to h igh risk assets to be m asked by the often-larger exposures to low ri sk assets.

T h e problem s w ith reportin g an aggregat ed n u m ber exi st n ot on ly w h ere such a n um ber i s reported in isolation, but even w h ere it is reported alon g w i th oth er supplem en t ary m easures. Despite the additional con text th at m ay be provided by supplem en t ary m easures, a si n gle aggregated num ber will focus attention on the misleading measurement and lessen the import of any other measures that might be reported.

Su m ea ch a sset cl a ss’ l on g ex p osu r es a n d sh or t ex p osu r es, sep a r a t el y

W e beli eve th at a furth er disti n ction sh ould be m ade betw een lon g exposures an d sh ort exposures, w ith leverage reported for each , broken down by asset class. Just as aggregating exposure across asset classes of very di ffer en t ri sk levels provides an u n in form ative view of leverage, so to do measures that aggregate long and positions. There are typically vastly different contributions to both portfolio and systemic risk from positions of the sam e aggregate notional w h ere on e is lon g-only and the other is long and short. 4

U se a com b i n a t i on of m ea su r es a t “ St ep On e”

T h e R eport proposes an d di scusses th r ee separat e “St ep O n e” gross n oti on al m easur es; un adjusted gross notional (GN E), adjusted gross notional (Adjusted GN E) and net notional exposure (N N E). W e agree with th e vi ew expressed i n th e Report th at each on e of th ese “ Step O n e” m easures m ay not provide regulators with a means to exclude from consideration funds that are unlikely to pose ri sks to th e fi n an ci al system an d so do n ot w ar ran t furth er an alysi s. W e support the notion that each on e of th ese m easures m ay prove m ore m ean in gful w h en u sed in com bin ation with each other and that it would n ot be appropriate to run a “St ep One” an alysi s usi n g only on e of th ese m easures as a stan dalon e m etric. As stat ed above, it is important that each of th ese m easures, even when used in combination, should differentiate between exposures to different asset classes rather than appearing as a single aggregate num ber.

W e ackn ow ledge th e st ren gth s an d w eakn esses of th ese th ree di fferen t m etrics as i den ti fied in th e Report, an d u ltim at ely beli eve th at th e m ost effective leverage m easurem en t w ould reflect a com bination of Adjusted GN E with N N E, differentiated across exposures to different types of asset classes. T he application of the netting principles from N N E to Adjusted GN E further refin es th e “Step On e” an alysis in th at it en ables econ om ic exposures w ith in a fun d portfolio to be offset an d th ereby assi sts in th e iden ti fication of effective leverage gen erated by securiti es an d derivatives positions. H ow ever, w e recogn ize th at th e application of N N E n ecessari ly raises question s about th e scope of n et tin g allow ed, an d beli eve th at th e ran ge of per m i ssi ble netting for th ese purposes sh ou ld be fairly narrow.

T h e com bin ed A djust ed G N E an d N N E m easu rem en t sh ould th en be appli ed to a fun d’s asset s i n correspon den ce w i th an asset class br eakdow n such as th e on e provi ded i n th e table on page 11 of th e R eport, which separates out m ajor asset classes such as equi ti es, cr edi t , i n t er est, and com m odities, each broken down into long and short positions. H ow ever, for th e sam e reason s articulated in our com m ents above about a single aggregated num ber, we suggest that any reporti n g t able n ot i n clude th e “T otals” row on page 11 that aggregates across asset classes as t h i s data point is misinformative.

A simple, consistent and comparable approach

T his approach, though m ore com plex than a single aggregated gross notional num ber, is relatively straightforward to apply and allows a m uch m ore soph i sti cat ed un der stan di n g of a fun d’s use of leverage. W e describe t h e approach i n greater det ai l below . 5

First, we adopt the concept of “GNL” as bei n g t h e sum of n oti on al exposures to di fferen t asset s, som e qui te ri sky oth ers less so, som e posi ti on s long an d som e sh ort .

GNL = GNLEQ-long + GNLFI-long + GN LFX-long + GN LCOM M -long + GNLEQ-sh or t + GNLFI-sh or t + GNLFX-sh or t + GNLCOM M -sh or t + GNLCR-long + GNLCR-sh or t W h ere:

GNLEQ-long and GN LEQ-sh or t are the notional exposure of long and sh ort positions made up of all equity underlier exposures, cash and derivatives;

GNLFI-long and GNLFI-sh or t are the notional exposure of long and sh ort positions m ade up of fixed in com e u n derliers. Further, FX would den ote curren cy exposure, COM M den otes com m odity exposure, an d, i f credit exposures are to be treat ed separat ely from fi xed i n com e, then CR for credit.

The method is simple; fund managers would report the component parts, but not total GNL, to regu lator s. Reportin g t h e pieces en ables regu lators to su m an d com p ar e m or e si m i lar exposur es across r elevan t sets of leveraged fu n ds an d bet ter assess ri sks across t h e fi n an ci al syst em . T h i s type of reportin g also perm i t s r egulators to an alyze ri sk across set s of fun ds even w h ere n on e of th e fu n ds m easured alon e by w ay of an aggregate gross m easu re w ould h ave been iden tifi ed for furth er revi ew .

One of the benefits of this approach is that it would also allow regulators to calculate risk m easures when analyzing a group of different funds, regar dless of th ei r advi sor. To undertake this additional type of analysis, regulators could take the inform ation provided and apply a haircutting m ethodology or even a simple risk model, whether at the individual fund level or across a set of fun ds expected to sh ar e corr elat ed ri sk exposur es.

T he calculations of GN L of the com ponent parts should incorporate the duration adjustm ents associ at ed w i th A djust ed G N E an d n etti n g ben efi t s of N N E , as di scussed above.

W e explore h ow this approach m ight work by way of sever al exam ples below :

Example 1: Repor t i n g for a r i sk par i t y fu n d com posed of st ock, bon d, an d com m odi t y fu t u r es, w i t h som e i n f l a t i on - linked bonds

E xposures are:

• Equity index futures 35% 6

• Govern m en t bon d futures 120% • Inflation-linked bonds 40% • Com m odities 20%

T he portfolio is long-only and completely unhedged. Under the reporting regime proposed, portfolio reporting would m ake the following calculations and report three non-zero data elem en ts:

GNLEQ-long = 35% GNLFI-long = 120% + 40% = 160% GNLCOM M -long = 20% GNLEQ-sh or t = GNLFI-sh or t = GNLCOM M -sh or t = 0%

In this exam ple portfolio, the total gross notional exposure is 215%. But is 215% a lot or a little? It would be a higher risk fund if all 215% of the exposure was to equities or com m odities, but not if th at exposure w ere en tirely attribut able to T reasu ry N ot es. T h is sli gh tly m ore det ailed accoun ti n g sh ow s qui ckly th at th ere i s m oder at e exposur e to h i gh ri sk assets (stocks and commodities), but that most exposure is to lower risk bonds and bond futures. Further, it shows that all exposures are directional and that the leverage is not used to create exposures with risks that are potentially offset (hedges). Unlike with risk-based reporting measures, the accounting inform ation is all available for aggregation across m any portfolios (say, all portfolios m an aged by all m an agers) w h i le th e ri skin ess of th e port folio can be ju dged by regu lators without first being processed by th e asset m an ager.

N ote that we make no distinction between the forms of the underlying exposure, be it cash securiti es or derivatives. A ll exposu res are reported.

Example 2: Por t foli o r epor t i n g for a m ar ket -neutral equity portfolio that is “ b et a ” n eu t r a l

T he volatility of a $1 long by $1 short equity portfolio is low; too low to be of m uch investm ent interest since it isn’t risky enough to generate th e return s n eeded to be con sidered a productive use of capital for m ost investors. In contrast, a portfolio that is $3 long by $3 short on $1 of N AV could generate enough portfolio risk and possibly provide enough uncorrelated return to be of in terest to in vestors. H ow w ould w e m easure it s leverage? A gain , usi n g th e sam e defi n ition s, w e report both long and short positions:

GNLEQ-long = 300% GNLEQ-sh or t = 30 0 % 7

GNLFI-long and GN L FI-sh or t = 0% (sam e for ot h er asset classes)

Regulators and experienced investors would understand from this report that this is a portfolio with meaningful exposure to risky equity underliers but constructed with the intent to offset m uch of th ose ri sks lon g vs. sh or t. I t does n ot say th ose risks h ave been successfully offset; on ly that that m ay be th e portfolio m an ager’s i n t en t , an d it does allow the iden tification of th e full valu e of th ose exposures. Everyone can now observe that total notional exposure to risky assets is fairly high, but that the portfolio m ay be very different from a portfolio of sim ilar notional exposure th at i s lon g-only equities.

This example also shows that the relatively high exposure is neither an attempt to construct a highly risky portfolio nor is it a portfolio th at em ploys leverage because th e u n derlyin g asset s are low risk (th ey aren ’t , th ey’re stocks.) Righ t aw ay, w e can see th e poten ti al issue; th e portfoli o reli es on r eason ably effecti ve ri sk offsets across ri sky assets an d i t i s th at pr esum ed property that m ust be carefu lly analyzed by i n vestors or exam i n ers.

Example 3: L on g-on ly p or t foli o of for ei gn equ i t i es w i t h a 5% ca sh bu f f er

GNLEQ-long = 95% GNLEQ-sh or t = 0%

Example 4: L on g-on ly p or t foli o of for ei gn equities with a 5% cash buffer, 50% cu r r en cy h ed ged

GNLEQ-long = 95% GNLEQ-sh or t = 0% GNLFX-sh or t = 50 %

Examples 3 and 4 are straightforward but note that in contrast to the commitment method, the curren cy h edge i s sti ll r eport ed. N either portfolio likely presen t s m uch risk from leverage, but the two portfolios are not in fact the same and should not be reported as if they were. T h e curren cy h edged port foli o i s li kely less ri sky, but i t does h ave con t ractu al agr eem en ts th at require potential mark-to-m arket paym en t s an d it does in fact h ave exposures th at exceed 100%. T h is exam ple h i gh ligh ts som e of th e distin ction s betw een m arket risk an d leverage. H ere, leverage is used t o reduce port folio volati lity, but w h ile in t erestin g from an oversi gh t perspective, portfolio volatility is not the only r i sk. I n th e h edged port folio, th e presen ce of contractual com m itm ents m ay also be relevant and with this method, it is exposed.

Can a regulator ever “add up” the parts? 8

As noted above, we rem ain highly critical of aggregation. H owever, our suggested approach does n ot preven t th e aggregation of th is m et ric w h ere a regulator m ay fin d som e use for th e n um ber as par t of i t s oversi gh t. R egu lator s can sti ll add th e pi eces up into a single number, but we strongly believe that this step should be taken only after a meaningful initial individual asset class assessm en t of leverage. T h e iden ti fication of leverage by asset class ret ain s th e basic, critical information that will be m ost usefu l for regu lator s i n assessi n g poten ti al ri sk.

Supplem entary D ata Points

We broadly support the use by regulators of supplementary and additional data points in conjunction with other “Step One” measures, as they may assist regulators in further refining the identification of potential systemic risk arising from leverage in invest m en t fun ds. Som e of th ese supplem en tary data poin t s are likely already collect ed by regulators w ith in curren t regu latory reportin g regim es an d/or w ou ld i n volve m odest ch an ges to existin g regulatory reporting categories. W e recom m end as an initial m atter that regulators seek to utilize these easi ly accessible dat a poin ts in th e “St ep O n e” process.

W e do n ot propose an y speci fic data poin ts at t h is tim e because w e believe th at th e ch oice of which additional data points are m ost useful m ay differ from regulator to regulator based on the types of fu n ds un der th eir oversigh t an d th eir existin g regulatory reportin g regim es. W h i le supplem en t ary data poi n ts can be useful, w e do n ot w ish for th ei r in clusion to add un n ecessary com plexity to th e “St ep On e” an alysi s of syst em i c ri sk. W e th erefor e w ould en courage r egu lat ors to adopt supplementary data points that elucidate rather than complicate their analysis at “Step One” with no undue emphasis being placed on any one additional data point.

Conclusion

W e beli eve th at leverage can be an effecti ve tool for i n vestm en t fun ds t o use in pursuin g th e in terest s of th eir clien ts. A t th e sam e tim e, w e recogn ize th e n eed for a con sisten t m easu rem en t of leverage th at th e i n t ern ation al regu latory com m un ity can use to assess poten ti al system i c ri sks to th e fi n an ci al syst em . T o th at en d, w e support t h e w ork th at I O SCO h as don e th us far i n r espon se to the Recommendation. In particular, we support IOSCO’s acknowledgment in the Report of the flaws inherent in the use of a single gross notional metric, which would fail to achieve the objective of det erm i n in g a usefu l an d m ean in gful m easure of leverage for th e purposes set out in th e R ecom m en dation an d R eport.

T h e approach w e propose in th is lett er reflects our un derst an di ng of the best way that IOSCO can ach ieve th e objectives of th e Recom m en dation . A t th e very least w e h ope th at it serves to illust rat e the point that modest changes to the aggregated gross notional metrics that have been used in the 9

past will allow regulators to m ore m ean in gfu lly m easure an d assess th e use of leverage in i n vestm en t fun ds for system i c ri sk purposes.

W e are ext rem ely grat eful for th e opportun ity t o con tribute to th e R eport an d sh ould you h ave an y questi on s or w i sh to speak to th e m att ers set ou t in th is let ter, w e w ould be m ore th an h appy t o assist. Please feel free to contact us at m ichael.m [email protected] , [email protected] , or [email protected] .

Yours sincerely,

/s/ Michael Mendelson /s/ Andrew Bastow /s/ Richard Grant

Michael Mendelson Andrew Bastow Richard Grant AQR Capital Management, LLC