<div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://files.passeidireto.com/7e6bc13c-bf65-4cc9-ae46-d7a88a02abc2/bg1.png"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls5 ws4">BIS <span class="ls0">W</span>orking <span class="ws0">P</span>apers </div><div class="t m0 x1 h3 y2 ff1 fs1 fc0 sc0 ls1 ws6">No 68<span class="ls5 ws7">5 </span></div><div class="t m0 x1 h4 y3 ff1 fs2 fc0 sc0 ls5 ws8">Why so low for <span class="_0 blank"></span><span class="ls2">s<span class="ls5">o <span class="ls3">l</span>ong? </span></span></div><div class="t m0 x1 h4 y4 ff1 fs2 fc0 sc0 ls6">A<span class="wsc"> long</span><span class="ls5 ws1">-t<span class="ls7">erm<span class="wse"> view<span class="wsf"> o</span><span class="wsd">f real </span></span></span></span></div><div class="t m0 x1 h4 y5 ff1 fs2 fc0 sc0 ls7">interest<span class="ws10"> ra</span><span class="wsd">tes </span></div><div class="t m0 x1 h5 y6 ff1 fs3 fc0 sc0 ls8">by<span class="ws12"> </span><span class="ls9 ws2">Claudio<span class="ws13"> </span>Borio</span><span class="ws14">, </span><span class="ls9 ws15">P<span class="_1 blank"> </span>iti Disyatat</span><span class="ls5 ws7">, M</span><span class="ls9 ws2">ikae<span class="_1 blank"> </span><span class="ws16">l </span>Juselius<span class="ws17"> </span></span><span class="lsa ws18">an<span class="_1 blank"> </span>d </span></div><div class="t m0 x1 h6 y7 ff1 fs3 fc0 sc0 ls9 ws2">Phurichai<span class="fs4 ls5 ws19 v1"> </span><span class="lsb ws3 v0">Rungcharoe<span class="_0 blank"></span>nkitkul<span class="ls5 ws1a"> </span></span></div><div class="t m0 x1 h7 y8 ff1 fs5 fc0 sc0 lsc ws7">Monetary and<span class="ws1b"> E</span>conomic Department<span class="ls5 ws1c"> </span></div><div class="t m0 x1 h5 y9 ff1 fs3 fc0 sc0 lsd">December<span class="ws1d"> </span><span class="ws12">2017 </span></div><div class="t m0 x1 h8 ya ff1 fs3 fc0 sc0 ls9">JEL<span class="ws1e"> classification:</span><span class="ls5 ws9"> <span class="fs6 lse">E32,<span class="ws20"> E40, E44, E50,<span class="ws21"> E52<span class="ls5">.</span></span></span></span><span class="ws22"> </span></span></div><div class="t m0 x1 h8 yb ff1 fs3 fc0 sc0 lsf">Keywords:<span class="ws24"> </span><span class="fs6 lse ws25">Real interes<span class="ws26">t rate<span class="ls5 ws27">, <span class="ws28">natural <span class="ws29">interest <span class="ws2a">rate, </span></span></span></span></span></span></div><div class="t m0 x1 h8 yc ff1 fs6 fc0 sc0 ls5 ws7">saving, investment, inflation, m<span class="ls4 ws2b">onetary p<span class="lse">olicy</span></span>.<span class="fs3 ws2c"> </span></div></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div> <div id="pf2" class="pf w0 h0" data-page-no="2"><div class="pc pc2 w0 h0"><div class="t m0 x2 h5 yd ff1 fs3 fc0 sc0 ls10">B<span class="ls5 ws33 v0">IS Working Papers <span class="_0 blank"></span>are written by <span class="_0 blank"></span>members of the Mo<span class="_0 blank"></span>netary and Economic </span></div><div class="t m0 x2 h5 ye ff1 fs3 fc0 sc0 ls12 ws34">Departme<span class="_0 blank"></span>nt of the<span class="_0 blank"></span> Bank f<span class="_0 blank"></span>or Intern<span class="_0 blank"></span>ational<span class="_0 blank"></span> Settleme<span class="_0 blank"></span>nts, a<span class="_0 blank"></span>nd from ti<span class="_0 blank"></span>me to time<span class="_0 blank"></span> by othe<span class="_0 blank"></span>r </div><div class="t m0 x2 h5 yf ff1 fs3 fc0 sc0 ls13 ws35">economists, and are published by the Ban<span class="_1 blank"> </span>k. The papers are on subjects of topical </div><div class="t m0 x2 h5 y10 ff1 fs3 fc0 sc0 ls14 ws36">interest and are techn<span class="_1 blank"> </span>ical in character. The view<span class="_1 blank"> </span>s expressed in them are those<span class="_1 blank"> </span> of their </div><div class="t m0 x2 h5 y11 ff1 fs3 fc0 sc0 ls12 ws37">authors a<span class="_0 blank"></span>nd not nece<span class="_0 blank"></span>ssarily<span class="_0 blank"></span> the views of the<span class="_0 blank"></span> BIS.<span class="ls5 ws7"> </span></div><div class="t m0 x2 h5 y12 ff1 fs3 fc0 sc0 ls14 ws2d">Th<span class="ws38 v0">is publicat<span class="_1 blank"> </span>ion is available on th<span class="_1 blank"> </span>e BIS website <span class="ls5 ws2e">(<span class="ls15 ws2f">www.bis.org</span><span class="ws7">). </span></span></span></div><div class="t m0 x2 h9 y13 ff2 fs3 fc0 sc0 ls5 ws7">© <span class="_2 blank"> </span><span class="ls10">B<span class="ls13 ws39 v0">ank for International Sett<span class="_1 blank"> </span>lements <span class="ls5">2017</span>. All rights reserved. Brief ex<span class="_1 blank"> </span>cerpts may be </span></span></div><div class="t m0 x3 ha y14 ff2 fs3 fc0 sc0 ls13 ws3a">reproduced or translat<span class="_1 blank"> </span>ed provided the source is stated.<span class="ls5 ws7"> </span></div><div class="t m0 x2 h5 y15 ff1 fs3 fc0 sc0 ls10 ws30">I<span class="v0">SSN<span class="ls11 ws7"> <span class="ls16 ws31">102<span class="ls5 ws2e">0-</span>0959</span> <span class="ls12 ws32">(print)</span><span class="ls5"> </span></span></span></div><div class="t m0 x2 h5 y16 ff1 fs3 fc0 sc0 ls14 ws2d">IS<span class="ls5 ws7">SN 1682-<span class="ls16 ws31">7678</span> (</span>onlin<span class="_1 blank"> </span>e<span class="ls5 ws7">) </span></div><a class="l"><div class="d m1" style="border-style:none;position:absolute;left:284.305000px;bottom:267.959000px;width:58.284000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div> <div id="pf3" class="pf w2 hb" data-page-no="3"><div class="pc pc3 w2 hb"><img fetchpriority="low" loading="lazy" class="bi x4 y17 w3 hc" alt="" src="https://files.passeidireto.com/7e6bc13c-bf65-4cc9-ae46-d7a88a02abc2/bg3.png"><div class="t m0 x5 hd y18 ff3 fs7 fc1 sc0 ls27 ws42">Why so low for so long? </div><div class="t m0 x6 he y19 ff3 fs7 fc1 sc0 ls27 ws43">A long-term view of real interest rates<span class="ff4 fs8 sc1 ls5 ws3b v2">\uf02a</span><span class="ls5 ws7 v0"> </span></div><div class="t m0 x7 hf y1a ff3 fs3 fc0 sc0 ls17 ws44">Claudio Borio, Piti<span class="_0 blank"></span> Disyatat, Mikael<span class="_0 blank"></span> Juselius and Phuric<span class="_0 blank"></span>hai Rungcharoenkitk<span class="_0 blank"></span>ul<span class="fs4 ls5 ws3c v1">\u2020</span><span class="ls5 ws7"> </span></div><div class="t m0 x8 h5 y1b ff3 fs3 fc0 sc0 ls28 ws45">2 December 2017 </div><div class="t m0 x9 h7 y1c ff3 fs5 fc1 sc0 ls18 ws3d">Abstract<span class="ff5 ls5 ws7"> </span></div><div class="t m0 x4 h5 y1d ff3 fs3 fc0 sc0 ls29 ws46">Prevailing explanations of the decline in real int<span class="_1 blank"> </span>erest rates since the early 1980s are premised </div><div class="t m0 x4 h5 y1e ff3 fs3 fc0 sc0 ls2a ws47">on the notion that real interest rat<span class="_1 blank"> </span>es are driven by variations in desired saving and investment. </div><div class="t m0 x4 h5 y1f ff3 fs3 fc0 sc0 ls5 ws48">But based on data stretching back to 187<span class="_0 blank"></span>0 for 19 countries, our systematic analysis<span class="_0 blank"></span> casts doubt </div><div class="t m0 x4 h5 y20 ff3 fs3 fc0 sc0 ls29 ws49">on this view. The link <span class="ls19 ws4a">between real interest rates and sa<span class="ls2b ws4b">ving-investment determi<span class="_0 blank"></span>nants appears </span></span></div><div class="t m0 x4 h5 y21 ff3 fs3 fc0 sc0 ls2c ws4c">tenuous. While it<span class="_1 blank"> </span> is possible to find some relationships consistent with th<span class="_1 blank"> </span>e theory in some </div><div class="t m0 x4 h5 y22 ff3 fs3 fc0 sc0 ls28 ws4d">periods, particularly over the la<span class="ls2d ws4e">st 30 years, they do not surviv<span class="ls2b ws4f">e over the extended sample. Thi<span class="_0 blank"></span>s </span></span></div><div class="t m0 x4 h5 y23 ff3 fs3 fc0 sc0 ls1a ws50">holds both at the <span class="_0 blank"></span>national<span class="ls2e ws51"> and glo<span class="_1 blank"> </span>bal<span class="_1 blank"> </span> leve<span class="_1 blank"> </span>l. B<span class="_1 blank"> </span>y co<span class="_1 blank"> </span>ntr<span class="_1 blank"> </span>ast<span class="_1 blank"> </span>, we<span class="_1 blank"> </span> fi<span class="ls2f ws52">nd evidence that <span class="ls30 ws53">persistent<span class="_1 blank"> </span> shifts </span></span></span></div><div class="t m0 x4 h5 y24 ff3 fs3 fc0 sc0 ls31 ws54">in real interest rates coincide with changes in <span class="_0 blank"></span><span class="ls5 ws55">monetary regimes. Moreov<span class="ls32 ws56">er, external influence<span class="_0 blank"></span>s </span></span></div><div class="t m0 x4 h5 y25 ff3 fs3 fc0 sc0 ls33 ws57">on countries\u2019 real intere<span class="ls34 ws58">st rates appear to reflect idiosyncratic variations in interest rates of </span></div><div class="t m0 x4 h5 y26 ff3 fs3 fc0 sc0 ls1b ws59">countries that domi<span class="_0 blank"></span>nate global monetary <span class="_0 blank"></span>and financial co<span class="_0 blank"></span>nditions rather than co<span class="_0 blank"></span>mmon </div><div class="t m0 x4 h5 y27 ff3 fs3 fc0 sc0 ls35 ws5a">movements in global saving and inves<span class="_0 blank"></span>tment. All th<span class="ls36 ws5b">is points to an underra<span class="ls29 ws5c">ted role of monetary </span></span></div><div class="t m0 x4 h5 y28 ff3 fs3 fc0 sc0 ls1c ws5d">policy in determining real<span class="_0 blank"></span> interest rates over long<span class="_0 blank"></span> horizons. </div><div class="t m0 x4 h8 y29 ff3 fs6 fc0 sc0 ls37 ws5e">JEL classification: E32, E40, E44, E50, E52. </div><div class="t m0 x4 h8 y2a ff3 fs6 fc0 sc0 ls38 ws5f">Keywords: Real interest rate, nat<span class="ls39 ws60">ural interest rate, saving, in<span class="ls5 ws61">vestment, inflation, monetary </span></span></div><div class="t m0 x4 h8 y2b ff3 fs6 fc0 sc0 ls3a ws7">policy. </div><div class="t m0 x4 h10 y2c ff4 fs9 fc0 sc0 ls1d">\uf02a<span class="ff3 fsa ls1e ws62 v3">We would like to thank I<span class="ff6 ls5">ñ<span class="ff3 ls3b ws63">aki Aldasoro, Marco Buti and colleagues, S<span class="_1 blank"> </span>tijn <span class="ls3c ws64">Claessens<span class="_0 blank"></span>, Andy Filardo, M<span class="_0 blank"></span>arc Flandreau,</span></span></span></span></div><div class="t m0 xa h11 y2d ff3 fsa fc0 sc0 ls3d ws65">Joseph Gagnon, Gasto<span class="_1 blank"> </span>n Gelos, Charles Goo<span class="_1 blank"> </span>dhart, James Ha<span class="_1 blank"> </span><span class="ls3e ws66">milton, Esa Jokiv<span class="_1 blank"> </span>uolle, David Laidler, En<span class="_1 blank"> </span>rique Martínez-</span></div><div class="t m0 xa h11 y2e ff3 fsa fc0 sc0 ls3f ws67">García, Luis Brandao Marques, Elma<span class="ls40 ws68">r Merte<span class="_1 blank"> </span>ns, Emanuel Mönch,<span class="_1 blank"> </span> Francisco Nadal de Simone, Ed<span class="_1 blank"> </span>ward Nels<span class="v0">on,</span></span></div><div class="t m0 xa h11 y2f ff3 fsa fc0 sc0 ls1f ws3e">Lu<span class="ws69 v0">kasz Rachel, Umang Rawat, Daniel Rees, Larry Sc<span class="_0 blank"></span>hembri, <span class="ls20">Hyun Song Shin, Nathan Sussman<span class="_0 blank"></span>, Alan Taylor, Kostas</span></span></div><div class="t m0 xa h11 y30 ff3 fsa fc0 sc0 ls41 ws6a">Tsatsaroni<span class="_1 blank"> </span>s, Gregory T<span class="_1 blank"> </span>hwaites and<span class="_1 blank"> </span> BIS sem<span class="_1 blank"> </span>inar par<span class="_1 blank"> </span>ticipants fo<span class="_1 blank"> </span>r helpf<span class="_1 blank"> </span>ul comment<span class="_1 blank"> </span>s and di<span class="_1 blank"> </span>scussions.<span class="_1 blank"> </span> Amy Wood,</div><div class="t m0 xa h12 y31 ff3 fsa fc0 sc0 ls42 ws6b">Diego Urbina and G<span class="_1 blank"> </span>iulio Cornelli <span class="_1 blank"> </span>provided excellent stati<span class="_1 blank"> </span>s<span class="ls43 ws6c">tical ass<span class="_1 blank"> </span>istance. All remai<span class="_1 blank"> </span>ning errors are<span class="_1 blank"> </span> ours. The vie<span class="_1 blank"> </span>w<span class="v0">s</span></span></div><div class="t m0 xa h11 y32 ff3 fsa fc0 sc0 ls21 ws3f">expr<span class="ws6d v0">essed are those of the authors and<span class="_0 blank"></span> do not necessarily represent thos<span class="_0 blank"></span>e of the Bank for Internationa<span class="v0">l</span></span></div><div class="t m0 xa h13 y33 ff3 fsa fc0 sc0 ls1e ws6e">Settlements, <span class="v0">Bank of Finland or<span class="_1 blank"> </span> the Bank of Thailand</span>.</div><div class="t m0 x4 h14 y34 ff3 fs9 fc0 sc0 ls22">\u2020<span class="fsa ls3f ws6f v3">Rungcharoenkitk<span class="_1 blank"> </span>ul (correspondi<span class="_1 blank"> </span>ng author), Senior E<span class="_1 blank"> </span>conomist, Bank for Internati<span class="_1 blank"> </span>onal Settlements,</span></div><div class="t m0 xa h11 y35 ff3 fsa fc0 sc0 ls3f ws40">phu<span class="ws70 v0">richai.rungcharoenkit<span class="_1 blank"> </span>kul@bis.org; <span class="_1 blank"> </span>Juselius: Senior Resea<span class="ls23 ws71">rch Economist, Ba<span class="_1 blank"> </span>nk of Finl<span class="ls24 ws72">and, mikael.juselius@bof.fi; </span></span></span></div><div class="t m0 xa h11 y36 ff3 fsa fc0 sc0 ls42 ws73">Borio: Head of Monetary a<span class="_1 blank"> </span>nd Economic De<span class="_1 blank"> </span>partment, Bank<span class="ls25 ws74"> for I<span class="_1 blank"> </span>nternational Settlements, claudio.b<span class="_1 blank"> </span>orio@bis.or<span class="v0">g;</span></span></div><div class="t m0 xa h11 y37 ff3 fsa fc0 sc0 ls43 ws41">Dis<span class="ws75 v0">yatat: Execut<span class="_1 blank"> </span>ive Director,<span class="_1 blank"> </span> Puey Ungpha<span class="_1 blank"> </span>korn Insti<span class="_1 blank"> </span>tute fo<span class="ls26 ws76">r Economic Research, <span class="_1 blank"> </span>Bank of Thailand,<span class="_1 blank"> </span> pitid@bot.or.th.</span></span></div><a class="l" data-dest-detail='[3,"XYZ",0,119,-32768]'><div class="d m1" style="border-style:none;position:absolute;left:484.920000px;bottom:682.253000px;width:2.430000px;height:8.424000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[3,"XYZ",0,211,-32768]'><div class="d m1" style="border-style:none;position:absolute;left:449.700000px;bottom:728.594000px;width:4.740000px;height:11.613000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l"><div class="d m1" style="border-style:none;position:absolute;left:465.548000px;bottom:71.980000px;width:58.418000px;height:16.640700px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l"><div class="d m1" style="border-style:none;position:absolute;left:125.350000px;bottom:99.261300px;width:164.795000px;height:10.640700px;background-color:rgba(255,255,255,0.000001);"></div></a></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div> <div id="pf4" class="pf w2 hb" data-page-no="4"><div class="pc pc4 w2 hb"><div class="t m0 xb h15 y38 ff6 fsb fc0 sc0 ls4b ws7">iv </div><div class="t m0 xb h16 y39 ff6 fsc fc2 sc0 ls5 ws7"> </div><div class="t m0 xb h16 y3a ff6 fsc fc2 sc0 ls5 ws7"> </div><div class="t m0 xc h7 y3b ff3 fs5 fc1 sc0 ls18 ws7b">Table of contents<span class="ff5 ls5 ws7"> </span></div><div class="t m0 xb h5 y3c ff3 fs3 fc0 sc0 ls31 ws7c">Introduction ..................................................................................................................<span class="_0 blank"></span><span class="ls2d ws7">............................................... <span class="_3 blank"></span><span class="ls5 ws2e">1<span class="ff7 fsd ws7"> </span></span></span></div><div class="t m0 xb h5 y3d ff3 fs3 fc0 sc0 ls4c ws77">1.<span class="ff7 fsd ls44 ws7"> </span><span class="ls2f ws7d">Real interest rate determination<span class="ls15 ws7e">: an over<span class="_0 blank"></span>view of approaches.................................................... <span class="_3 blank"></span><span class="ls45">2<span class="ff7 fsd ls5 ws7"> </span></span></span></span></div><div class="t m0 xb h5 y3e ff3 fs3 fc0 sc0 ls4c ws77">2.<span class="ff7 fsd ls44 ws7"> </span><span class="ls2f ws7d">Real interest rate determination<span class="ls15 ws7e">: the role<span class="_0 blank"></span> of real factors <span class="_4 blank"></span>............................................................. <span class="_3 blank"></span><span class="ls45">6<span class="ff7 fsd ls5 ws7"> </span></span></span></span></div><div class="t m0 x4 h5 y3f ff3 fs3 fc0 sc0 ls4d ws78">2.1<span class="ff7 fsd ls46 ws7"> </span><span class="ws7f">Essential elements of th<span class="ls2d ws80">e empirical strategy <span class="_5 blank"></span>........................................................................ <span class="_3 blank"></span><span class="ls5 ws2e">6<span class="ff7 fsd ws7"> </span></span></span></span></div><div class="t m0 x4 h5 y40 ff3 fs3 fc0 sc0 ls4d ws78">2.2<span class="ff7 fsd ls46 ws7"> </span><span class="ls31 ws81">Data and definition of variables <span class="_4 blank"></span>..............................................................................................<span class="_0 blank"></span><span class="ls2d ws7">.. <span class="_3 blank"></span><span class="ls5 ws2e">8<span class="ff7 fsd ws7"> </span></span></span></span></div><div class="t m0 x4 h5 y41 ff3 fs3 fc0 sc0 ls4d ws78">2.3<span class="ff7 fsd ls46 ws7"> </span><span class="ls2d ws82">A first look at the data <span class="_6 blank"></span>......................................................................................................<span class="_0 blank"></span><span class="ls15 ws83">......... 10<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y42 ff3 fs3 fc0 sc0 ls4d ws78">2.4<span class="ff7 fsd ls46 ws7"> </span><span class="ls4e ws84">Tests and main<span class="ls2d ws82"> results <span class="_5 blank"></span>........................................................................................................<span class="_0 blank"></span><span class="ls15 ws83">........ 12<span class="ff7 fsd ls5 ws7"> </span></span></span></span></div><div class="t m0 xb h5 y43 ff3 fs3 fc0 sc0 ls4c ws77">3.<span class="ff7 fsd ls44 ws7"> </span><span class="ls4f ws85">Real interest rate determination: <span class="_0 blank"></span><span class="ls15 ws7e">the role of monetary factors <span class="_4 blank"></span>.............................................. <span class="_4 blank"></span>21<span class="_0 blank"></span><span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y44 ff3 fs3 fc0 sc0 ls4d ws78">3.1<span class="ff7 fsd ls46 ws7"> </span><span class="ls31 ws81">Analytical considerations <span class="_6 blank"></span>.....................................................................................................<span class="_6 blank"></span><span class="ls28 ws86">..... 21<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y45 ff3 fs3 fc0 sc0 ls4d ws78">3.2<span class="ff7 fsd ls46 ws7"> </span><span class="ls31 ws87">Previous work <span class="_6 blank"></span>.................................................................................................................<span class="_0 blank"></span><span class="ws88">............... 25<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y46 ff3 fs3 fc0 sc0 ls4d ws78">3.3<span class="ff7 fsd ls46 ws7"> </span><span class="ls31 ws81">New evidence on monetary policy regimes <span class="_4 blank"></span>...................................................................... <span class="_4 blank"></span>26<span class="_0 blank"></span><span class="ff7 fsd ls5 ws7"> </span></span></div><div class="t m0 x4 h5 y47 ff3 fs3 fc0 sc0 ls4d ws78">3.4<span class="ff7 fsd ls46 ws7"> </span><span class="ws7f">A monetary narrative of the evolution of<span class="_0 blank"></span> real interest rates <span class="_7 blank"></span>...................................... <span class="_4 blank"></span>34<span class="_0 blank"></span><span class="ff7 fsd ls5 ws7"> </span></span></div><div class="t m0 xb h5 y48 ff3 fs3 fc0 sc0 ls2d ws89">Conclusion ....................................................................................................................<span class="ws8a">............................................. 36<span class="_0 blank"></span><span class="ff7 fsd ls5 ws7"> </span></span></div><div class="t m0 xb h5 y49 ff3 fs3 fc0 sc0 ls2d ws8b">References ....................................................................................................................<span class="_0 blank"></span><span class="ws8a">............................................. 38<span class="ff7 fsd ls5 ws7"> </span></span></div><div class="t m0 xb h5 y4a ff3 fs3 fc0 sc0 ls31 ws81">Annex A: Data and plots <span class="_8 blank"></span>.......................................................................................................<span class="_0 blank"></span><span class="ws88">............................... 44<span class="ff7 fsd ls5 ws7"> </span></span></div><div class="t m0 x4 h5 y4b ff3 fs3 fc0 sc0 ls47 ws79">A.1<span class="ff7 fsd ls48 ws7"> </span><span class="ls31 ws87">Data sources and coverage <span class="_5 blank"></span>.....................................................................................................<span class="ls4d ws8c">. 44<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y4c ff3 fs3 fc0 sc0 ls47 ws79">A.2<span class="ff7 fsd ls48 ws7"> </span><span class="ls2d ws82">Data plots of real interest rates <span class="_8 blank"></span>.............................................................................................<span class="ls4d ws8c">. 46<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y4d ff3 fs3 fc0 sc0 ls47 ws79">A.3<span class="ff7 fsd ls48 ws7"> </span><span class="ls19 ws8d">Saving-invest<span class="_0 blank"></span>ment factors<span class="ff8 ls5 ws7"> </span><span class="ls4f ws8e">for the United States <span class="ls2d ws80">and United Kingdom <span class="_7 blank"></span>................. <span class="_4 blank"></span>52<span class="ff7 fsd ls5 ws7"> </span></span></span></span></div><div class="t m0 xb h5 y4e ff3 fs3 fc0 sc0 ls31 ws81">Annex B: Robustness results <span class="_7 blank"></span>...................................................................................................<span class="_0 blank"></span><span class="ws88">........................... 53<span class="ff7 fsd ls5 ws7"> </span></span></div><div class="t m0 x4 h5 y4f ff3 fs3 fc0 sc0 ls49 ws7a">B.1<span class="ff7 fsd ls4a ws7"> </span><span class="ls31 ws81">Bivariate panel regression<span class="ws7e"> with time trends <span class="_8 blank"></span>...................................................................... <span class="_4 blank"></span>53<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y50 ff3 fs3 fc0 sc0 ls49 ws7a">B.2<span class="ff7 fsd ls4a ws7"> </span><span class="ls15 ws7e">Average dependent and in<span class="ls2d ws82">dependent variables <span class="_7 blank"></span>............................................................. <span class="_8 blank"></span>54<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y51 ff3 fs3 fc0 sc0 ls49 ws7a">B.3<span class="ff7 fsd ls4a ws7"> </span><span class="ls4e ws84">Short-term market rates as the dependent variable <span class="_3 blank"></span><span class="ls2d ws8a">...................................................... 56<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y52 ff3 fs3 fc0 sc0 ls49 ws7a">B.4<span class="ff7 fsd ls4a ws7"> </span><span class="ls31 ws81">Excluding periods after the world wars <span class="_5 blank"></span>............................................................................... <span class="_4 blank"></span>60<span class="ff7 fsd ls5 ws7"> </span></span></div><div class="t m0 x4 h5 y53 ff3 fs3 fc0 sc0 ls49 ws7a">B.5<span class="ff7 fsd ls4a ws7"> </span><span class="ls15 ws7e">Alternative expectations of <span class="ws81">inflation and GDP growth <span class="_7 blank"></span>................................................. <span class="_4 blank"></span>61<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y54 ff3 fs3 fc0 sc0 ls49 ws7a">B.6<span class="ff7 fsd ls4a ws7"> </span><span class="ls31 ws81">Savers\u2019 ratio as an independent variable <span class="_5 blank"></span>............................................................................ <span class="_4 blank"></span>63<span class="ff7 fsd ls5 ws7"> </span></span></div><div class="t m0 x4 h5 y55 ff3 fs3 fc0 sc0 ls49 ws7a">B.7<span class="ff7 fsd ls4a ws7"> </span><span class="ls32 ws8f">Time-varying retirement age interact<span class="ls15 ws7e">ions with demographic variable<span class="_0 blank"></span>s <span class="_7 blank"></span>................ <span class="_8 blank"></span>64<span class="ff7 fsd ls5 ws7"> </span></span></span></div><div class="t m0 x4 h5 y56 ff3 fs3 fc0 sc0 ls49 ws7a">B.8<span class="ff7 fsd ls4a ws7"> </span><span class="ls28 ws45">Productivity growth as an independent variable............................................................. <span class="_8 blank"></span>65<span class="ff7 fsd ls5 ws7"> </span></span></div><div class="t m0 x4 h5 y57 ff3 fs3 fc0 sc0 ls49 ws7a">B.9<span class="ff7 fsd ls4a ws7"> </span><span class="ls15 ws7e">Risk premium as an independent variable <span class="_5 blank"></span>......................................................................... <span class="_8 blank"></span>66<span class="ff7 fsd ls5 ws7"> </span></span></div><a class="l" data-dest-detail='[70,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:174.637000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[69,"XYZ",106,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:193.938000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[68,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:213.239000px;width:384.750000px;height:19.300000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[67,"XYZ",106,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:232.539000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[65,"XYZ",106,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:251.840000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[64,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:271.141000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[60,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:290.442000px;width:384.750000px;height:19.300000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[58,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:309.742000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[57,"XYZ",106,742,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:329.043000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[57,"XYZ",106,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:82.800000px;bottom:348.344000px;width:413.100000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[56,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:367.645000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[50,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:386.946000px;width:384.750000px;height:19.300000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[48,"XYZ",77,742,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:406.246000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[48,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:82.800000px;bottom:425.547000px;width:413.100000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[42,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:82.800000px;bottom:444.848000px;width:413.100000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[40,"XYZ",77,586,null]'><div class="d m1" style="border-style:none;position:absolute;left:82.800000px;bottom:464.149000px;width:413.100000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[38,"XYZ",77,277,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:483.450000px;width:384.750000px;height:19.300000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[30,"XYZ",77,613,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:502.750000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[29,"XYZ",106,506,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:522.051000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[25,"XYZ",106,552,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:541.352000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[25,"XYZ",106,712,null]'><div class="d m1" style="border-style:none;position:absolute;left:82.800000px;bottom:560.653000px;width:413.100000px;height:19.300000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[16,"XYZ",77,520,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:579.953000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[14,"XYZ",77,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:599.254000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[12,"XYZ",77,712,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:618.555000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[10,"XYZ",77,512,null]'><div class="d m1" style="border-style:none;position:absolute;left:111.150000px;bottom:637.856000px;width:384.750000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[10,"XYZ",77,645,null]'><div class="d m1" style="border-style:none;position:absolute;left:82.800000px;bottom:657.157000px;width:413.100000px;height:19.300000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[6,"XYZ",77,454,null]'><div class="d m1" style="border-style:none;position:absolute;left:82.800000px;bottom:676.457000px;width:413.100000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[5,"XYZ",106,778,null]'><div class="d m1" style="border-style:none;position:absolute;left:82.800000px;bottom:695.758000px;width:413.100000px;height:19.301000px;background-color:rgba(255,255,255,0.000001);"></div></a></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div> <div id="pf5" class="pf w2 hb" data-page-no="5"><div class="pc pc5 w2 hb"><div class="t m0 xd h17 y58 ff6 fs3 fc0 sc0 ls5 ws7"> </div><div class="t m0 xe h18 y59 ff3 fse fc0 sc0 ls52 ws94">WP685 Why so low for so l<span class="_1 blank"> </span>ong? A long view of<span class="_1 blank"> </span> real interest ra<span class="_1 blank"> </span>te determina<span class="_1 blank"> </span>tion <span class="_9 blank"> </span>1</div><div class="t m0 xe h16 y5a ff6 fsc fc2 sc0 ls5 ws7"> </div><div class="t m0 xe h7 y5b ff3 fs5 fc1 sc0 ls50 ws90">Introduction<span class="ff9 fs3 fc0 ls5 ws7"> </span></div><div class="t m0 xe h5 y5c ff3 fs3 fc0 sc0 ls4e ws95">Global real (inflation-adjusted) <span class="ls49 ws96">interest rates, shor<span class="_0 blank"></span>t and long, have b<span class="_0 blank"></span>een on a downward trend<span class="_0 blank"></span> </span></div><div class="t m0 xe h5 y5d ff3 fs3 fc0 sc0 ls11 ws97">throughout<span class="_0 blank"></span> much of the past 30 years and ha<span class="ls34 ws98">ve remained exceptionally low since the Great </span></div><div class="t m0 xe h5 y5e ff3 fs3 fc0 sc0 ls4e ws99">Financial Crisis (GFC). This has triggered a debat<span class="ls4d">e about the reasons for <span class="ls28">the decline. Invariably, </span></span></div><div class="t m0 xe h5 y5f ff3 fs3 fc0 sc0 ls32 ws9a">the presumption is that the evol<span class="ls2c ws9b">ution of real interest rates reflects changes in underlying </span></div><div class="t m0 xe h5 y60 ff3 fs3 fc0 sc0 ls35 ws9c">saving-investment determinants. These are seen<span class="ls53 ws7"> t<span class="_a blank"></span>o<span class="_a blank"></span> g<span class="_3 blank"></span>o<span class="_a blank"></span>v<span class="_3 blank"></span>e<span class="_a blank"></span>r<span class="_3 blank"></span>n<span class="_a blank"></span> v<span class="_3 blank"></span>a<span class="_a blank"></span>r<span class="_3 blank"></span>i<span class="_a blank"></span>a<span class="_3 blank"></span>t<span class="_a blank"></span>i<span class="_3 blank"></span>o<span class="_a blank"></span>n<span class="_3 blank"></span>s<span class="_a blank"></span> i<span class="_b blank"></span>n<span class="_a blank"></span> s<span class="_b blank"></span>o<span class="_b blank"></span>m<span class="_b blank"></span>e<span class="_b blank"></span> n<span class="_b blank"></span>o<span class="_b blank"></span>t<span class="_b blank"></span>i<span class="_b blank"></span>o<span class="_a blank"></span>n<span class="_b blank"></span>a<span class="_b blank"></span>l<span class="_b blank"></span> </span></div><div class="t m0 xe h5 y61 ff3 fs3 fc0 sc0 ls36 ws9d">\u201cequilibrium\u201d or natural real rate, defin<span class="_1 blank"> </span>ed as the real interest rat<span class="_1 blank"> </span>e that would prevail when </div><div class="t m0 xe h5 y62 ff3 fs3 fc0 sc0 ls19 ws8d">actual output equals potential outp<span class="_0 blank"></span>ut, <span class="ls5 ws9e">towards which market rates gravitate.<span class="_0 blank"></span> </span></div><div class="t m0 xa h5 y63 ff3 fs3 fc0 sc0 ls4e ws9f">The presumption that real interest rates <span class="_0 blank"></span>are so<span class="ls34 wsa0"> anchored is evident in<span class="ls4f wsa1"> two broad<span class="_0 blank"></span> analytical </span></span></div><div class="t m0 xe h5 y64 ff3 fs3 fc0 sc0 ls17 ws7">strands. </div><div class="t m0 xa h5 y65 ff3 fs3 fc0 sc0 ls31 ws81">The first focuses on <span class="ffa ls54 ws91">obse rv ed </span><span class="ls5 wsa2"> real interest rates and relates them d<span class="_0 blank"></span>irectly to the evolutio<span class="_0 blank"></span>n </span></div><div class="t m0 xe h5 y66 ff3 fs3 fc0 sc0 ls15 wsa3">of the factors that underpin th<span class="ls4e wsa4">e economy\u2019s saving-inve<span class="_0 blank"></span>stment balance (eg IMF (2014), Bean et </span></div><div class="t m0 xe h5 y67 ff3 fs3 fc0 sc0 ls28 wsa5">al (2015), Council of Economic Advisers (2015)). <span class="ls4c wsa6">One prominent v<span class="_0 blank"></span>ariant is<span class="ls4e wsa7"> the hypothesis that </span></span></div><div class="t m0 xe h5 y68 ff3 fs3 fc0 sc0 ls29 ws5c">persistently weak demand for capital, a rising pr<span class="ls55 wsa8">opens<span class="_1 blank"> </span>ity to sav<span class="_1 blank"> </span>e and lo<span class="_1 blank"> </span>wer tr<span class="_1 blank"> </span>end grow<span class="_1 blank"> </span>th have </span></div><div class="t m0 xe h5 y69 ff3 fs3 fc0 sc0 ls1c wsa9">brought about an era of \u201cs<span class="_0 blank"></span>ecular stagnation\u201d <span class="ls4c wsaa">(<span class="_0 blank"></span>Summers (2014, 2015)). Another variant argue<span class="_0 blank"></span>s </span></div><div class="t m0 xe h5 y6a ff3 fs3 fc0 sc0 ls32 wsab">that a higher propensity to save in e<span class="_0 blank"></span>merging <span class="ls2f wsac">market economies (EMEs), coupled with inves<span class="_0 blank"></span>tors\u2019 </span></div><div class="t m0 xe h5 y6b ff3 fs3 fc0 sc0 ls29 wsad">growing preference for safe a<span class="ls31 wsae">ssets, has boosted<span class="_0 blank"></span> the supply of<span class="ls4d wsaf"> saving worldwide (Bernanke </span></span></div><div class="t m0 xe h5 y6c ff3 fs3 fc0 sc0 ls4d wsb0">(2005), Broadbent (2014), Caballero et al (2008)).<span class="ls19 wsb1"> Most recently, <span class="_0 blank"></span>demographic changes have </span></div><div class="t m0 xe h5 y6d ff3 fs3 fc0 sc0 ls4e wsb2">been singled out (Carvalho et al<span class="_0 blank"></span> (2016), Gagnon et al (2016),<span class="_0 blank"></span> Rachel and Smith (2017)). This </div><div class="t m0 xe h5 y6e ff3 fs3 fc0 sc0 ls56 wsb3">strand typically does not consider in<span class="_1 blank"> </span>flation explicit<span class="ls57 wsb4">ly and links real interest<span class="_1 blank"> </span> rates directly to the </span></div><div class="t m0 xe h5 y6f ff3 fs3 fc0 sc0 ls35 wsb5">posited real-sector determinants.<span class="ls1b wsb6"> In effect, it assumes that ove<span class="_0 blank"></span>r the relevant horizon t<span class="_0 blank"></span>he </span></div><div class="t m0 xe h5 y70 ff3 fs3 fc0 sc0 ls4e ws84">observed (market) rate and the un<span class="ls31 ws81">observed natural rate coincide. </span></div><div class="t m0 xa h5 y71 ff3 fs3 fc0 sc0 ls49 wsb7">The second strand focus<span class="_0 blank"></span>es on the <span class="ffa ls31 ws77">equilibrium</span><span class="ls47 wsb8"> or <span class="ffa ls51 ws92">natur<span class="_0 blank"></span>al<span class="ff3 ls5 wsb9"> real rate, estimate<span class="_0 blank"></span>d as an </span></span></span></div><div class="t m0 xe h5 y72 ff3 fs3 fc0 sc0 ls28 wsba">unobserved variable in a filtering system (<span class="ls15 wsbb">eg Laubach and Williams (2015), Justiniano and </span></div><div class="t m0 xe h5 y73 ff3 fs3 fc0 sc0 ls32 wsbc">Primiceri (2010)). Typically, the natural rate is anchored to theory-prescribed variables, suc<span class="_0 blank"></span>h as </div><div class="t m0 xe h5 y74 ff3 fs3 fc0 sc0 ls36 wsbd">potential growth and ho<span class="ls34 wsbe">usehold preferences, which are them<span class="ls15 ws7e">selves unobserved, and inflation </span></span></div><div class="t m0 xe h5 y75 ff3 fs3 fc0 sc0 ls2a wsbf">plays a critical role in pinning down the natural <span class="ls31 wsc0">rate alongside the other la<span class="ls4f wsc1">tent system variables. </span></span></div><div class="t m0 xe h5 y76 ff3 fs3 fc0 sc0 ls4d wsc2">In Laubach and Williams (2003), for example, risi<span class="ls2f wsc3">ng inflation indicates that output is <span class="_0 blank"></span>above </span></div><div class="t m0 xe h5 y77 ff3 fs3 fc0 sc0 ls58 wsc4">potential and, correspondingly, that the actual in<span class="_1 blank"> </span>terest rate is below the natural rate; falling </div><div class="t m0 xe h5 y78 ff3 fs3 fc0 sc0 ls34 wsc5">inflation indicates the revers<span class="ls4d wsc6">e. These reflect the well known Phillips-curve and aggregate-</span></div><div class="t m0 xe h5 y79 ff3 fs3 fc0 sc0 ls58 wsc7">demand (IS) relationships that lie at the <span class="ls4f ws85">core of standard macroeconomic model<span class="_0 blank"></span>s. </span></div><div class="t m0 xa h5 y7a ff3 fs3 fc0 sc0 ls34 wsc8">Both strands share a couple of limitations. Th<span class="_1 blank"> </span><span class="ls4e wsc9">e bulk of the analysis examines the period </span></div><div class="t m0 xe h5 y7b ff3 fs3 fc0 sc0 ls2f wsca">since the mid-1980s,<span class="_0 blank"></span> when real interest rates<span class="_0 blank"></span> ha<span class="ls35 wscb">ve been declining. And <span class="ls4e wscc">neither<span class="_0 blank"></span> tests directly </span></span></div><div class="t m0 xe h5 y7c ff3 fs3 fc0 sc0 ls2f wscd">the hypotheses that <span class="ls4d wsce">the postulated saving-investment fr<span class="ls4e wscf">amework and/or the postulated </span></span></div><div class="t m0 xe h5 y7d ff3 fs3 fc0 sc0 ls2d wsd0">inflation determination process adequately ch<span class="ls15 wsd1">aracterise the data. These are regarded as </span></div><div class="t m0 xe h5 y7e ffa fs3 fc0 sc0 ls32 ws93">maintained<span class="ff3 ls28 wsd2"> hypotheses, be it in<span class="wsd3"> the underlying narrative and ca<span class="ls4f wsd4">libration of structural models </span></span></span></div><div class="t m0 xe h5 y7f ff3 fs3 fc0 sc0 ls4e wsa4">or in the filtering system<span class="ls15 wsa3">s. There is little by way of direct <span class="ls59 wsd5">estima<span class="_0 blank"></span>tion that te<span class="ls4d wsd6">sts the link between </span></span></span></div><div class="t m0 xe h5 y80 ffa fs3 fc0 sc0 ls2a ws78">observable<span class="ff3 ls32 wsd7"> variables, such as demographics, and real<span class="ls2b wsd8"> i<span class="_0 blank"></span>nterest rates. Notable exceptions<span class="_0 blank"></span> are </span></span></div><div class="t m0 xe h5 y81 ff3 fs3 fc0 sc0 ls32 wsd9">Hamilton et al (2015) and Lunsfo<span class="wsda">rd and West (201<span class="_0 blank"></span>7), who consi<span class="ls35 wsdb">der some such variables over </span></span></div><div class="t m0 xe h5 y82 ff3 fs3 fc0 sc0 ls2f ws7d">longer periods. </div><div class="t m0 xa h5 y83 ff3 fs3 fc0 sc0 ls56 wsdc">We aim to fill this gap by systematically exam<span class="_1 blank"> </span><span class="ls33 wsdd">ining the empirical link bet<span class="_1 blank"> </span>ween real interest </span></div><div class="t m0 xe h5 y84 ff3 fs3 fc0 sc0 ls5 wsde">rates and the posited determinants, <span class="_0 blank"></span>not just sinc<span class="ls2b ws4b">e the 1980s but also back in <span class="_0 blank"></span>history. Based on </span></div><div class="t m0 xe h5 y85 ff3 fs3 fc0 sc0 ls34 wsdf">data starting in the 19th cen<span class="_1 blank"> </span>tury for 19 economies, we find only a tenuous link between<span class="_1 blank"> </span> real </div><div class="t m0 xe h5 y86 ff3 fs3 fc0 sc0 ls4e wse0">interest rates and observable proxies for th<span class="ls35 wse1">e main saving-investment determinants. <span class="_0 blank"></span>Some </span></div><div class="t m0 xe h5 y87 ff3 fs3 fc0 sc0 ls4d wse2">variables, notably demographics, <span class="ls15 wsd2">do exhibit the expected relationsh<span class="ls2f wse3">i<span class="_0 blank"></span>p with real interest rates </span></span></div></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div> <div id="pf6" class="pf w2 hb" data-page-no="6"><div class="pc pc6 w2 hb"><img fetchpriority="low" loading="lazy" class="bi xf y88 w3 hc" alt="" src="https://files.passeidireto.com/7e6bc13c-bf65-4cc9-ae46-d7a88a02abc2/bg6.png"><div class="t m0 xf h19 y89 ffb fsb fc0 sc0 ls5 ws7"> </div><div class="t m0 xf h18 y59 ff3 fse fc0 sc0 ls52 ws94">2 <span class="_c blank"> </span>WP685 Why so low for so l<span class="_1 blank"> </span>ong? A long view of<span class="_1 blank"> </span> real interest ra<span class="_1 blank"> </span>te determina<span class="_1 blank"> </span>tion</div><div class="t m0 xf h16 y5a ff6 fsc fc2 sc0 ls5 ws7"> </div><div class="t m0 xf h5 y8a ff3 fs3 fc0 sc0 ls62 ws7">in <span class="_1 blank"> </span><span class="ffa ls63 wse4">some</span><span class="ls35 wse7"> subsamples, especially in the more recent on<span class="ls2d wse8">e. But there is little evidence of a stable </span></span></div><div class="t m0 xf h5 y8b ff3 fs3 fc0 sc0 ls2d wse9">relationship across subsamples. This applies <span class="ls2f ws7d">to both domestic and <span class="_0 blank"></span>global variables. </span></div><div class="t m0 x10 h5 y8c ff3 fs3 fc0 sc0 ls13 wsea">Going beyond the standard factors, we investigate whether monetary policy has persistent<span class="_1 blank"> </span> </div><div class="t m0 xf h5 y8d ff3 fs3 fc0 sc0 ls49 wseb">effects on real interest rates. In our lo<span class="_0 blank"></span>ng sa<span class="ls4e wscc">mple, monetary policy regimes, such as the gold </span></div><div class="t m0 xf h5 y8e ff3 fs3 fc0 sc0 ls58 wsec">standard, Bretton Woods and inflation targeting, go hand-in-hand with significant shift<span class="_1 blank"> </span>s in real </div><div class="t m0 xf h5 y8f ff3 fs3 fc0 sc0 ls30 wsed">interest rates<span class="_1 blank"> </span>. At a global level, we find that the in<span class="_1 blank"> </span>fluence of external fact<span class="_1 blank"> </span>ors on countries\u2019 real </div><div class="t m0 xf h5 y90 ff3 fs3 fc0 sc0 ls4e wse0">interest rates reflects the importa<span class="_0 blank"></span>nce of the fi<span class="ls2a wsee">nancially dominant countries\u2019 role as global </span></div><div class="t m0 xf h5 y91 ff3 fs3 fc0 sc0 ls5 wsef">monetary anchors rather than co<span class="_0 blank"></span>mmon variations<span class="ls2f wsf0"> in global saving<span class="_0 blank"></span>-investment determinants. </span></div><div class="t m0 xf h5 y92 ff3 fs3 fc0 sc0 ls15 wsf1">This suggests that co-movements in real interest rates <span class="ls2d wsf2">across countries are <span class="_0 blank"></span>more closely </span></div><div class="t m0 xf h5 y93 ff3 fs3 fc0 sc0 ls2f wsf3">related to the monetary policy of glob<span class="_0 blank"></span>al anchor<span class="ls64 ws7"> c<span class="_4 blank"></span>o<span class="_7 blank"></span>u<span class="_7 blank"></span>n<span class="_4 blank"></span>t<span class="_7 blank"></span>r<span class="_4 blank"></span>i<span class="_7 blank"></span>e<span class="_7 blank"></span>s<span class="_4 blank"></span> t<span class="_7 blank"></span>h<span class="_4 blank"></span>a<span class="_7 blank"></span>n<span class="_7 blank"></span> t<span class="_4 blank"></span>o<span class="_7 blank"></span> f<span class="_4 blank"></span>a<span class="_7 blank"></span>c<span class="_7 blank"></span>t<span class="_4 blank"></span>o<span class="_7 blank"></span>r<span class="_4 blank"></span>s<span class="_7 blank"></span> s<span class="_7 blank"></span>u<span class="_4 blank"></span>c<span class="_7 blank"></span>h<span class="_4 blank"></span> a<span class="_7 blank"></span>s<span class="_7 blank"></span> a<span class="_4 blank"></span> g<span class="_7 blank"></span>l<span class="_4 blank"></span>o<span class="_7 blank"></span>b<span class="_4 blank"></span>a<span class="_7 blank"></span>l<span class="_7 blank"></span> </span></div><div class="t m0 xf h5 y94 ff3 fs3 fc0 sc0 ls2f ws7d">saving glut. </div><div class="t m0 x10 h5 y95 ff3 fs3 fc0 sc0 ls15 wsf4">Overall, our results raise ques<span class="ls2b wsf5">tions about the prevailing paradigm of real interest rate </span></div><div class="t m0 xf h5 y96 ff3 fs3 fc0 sc0 ls2d wsf6">determination. The saving-investment framew<span class="ls65 wsf7">ork may not serve as a reliable guide for </span></div><div class="t m0 xf h5 y97 ff3 fs3 fc0 sc0 ls4e ws95">understanding real interest rate developments. <span class="ls58 wsf8">And inflation may not be a sufficiently reliable<span class="_1 blank"> </span> </span></div><div class="t m0 xf h5 y98 ff3 fs3 fc0 sc0 ls28 wsf9">signal of where real interest rates are relati<span class="ls2a ws35">ve to some unobserved natural level. Monetary </span></div><div class="t m0 xf h5 y99 ff3 fs3 fc0 sc0 ls31 wsfa">policy, and financial factors more generally, ma<span class="ls65 wsfb">y have an important bearin<span class="_1 blank"> </span>g on persistent </span></div><div class="t m0 xf h5 y9a ff3 fs3 fc0 sc0 ls35 wsfc">movements in real interest rates. </div><div class="t m0 x10 h5 y9b ff3 fs3 fc0 sc0 ls4e wsfd">The rest of the paper is organi<span class="ls29 wsfe">sed as follows. Section 1 provi<span class="ls35 wsff">des an overview of existing </span></span></div><div class="t m0 xf h5 y9c ff3 fs3 fc0 sc0 ls32 wsff">approaches to explaining real interest rates, <span class="ls13 ws100">highlighting their limitat<span class="_1 blank"> </span>ions. Section 2 analyses<span class="_1 blank"> </span> </span></div><div class="t m0 xf h5 y9d ff3 fs3 fc0 sc0 ls13 ws101">the relationship between real interest rates an<span class="ls15 ws7e">d a standard set of real-sector determinants for </span></div><div class="t m0 xf h5 y9e ff3 fs3 fc0 sc0 ls2b ws102">a cross section of countries over a long<span class="_0 blank"></span> time <span class="ls36 ws9d">span. Section 3 explores the possible role of </span></div><div class="t m0 xf h5 y9f ff3 fs3 fc0 sc0 ls36 ws103">monetary factors. The final section concludes. <span class="ls4e ws104">The Annexes provide detailed information about </span></div><div class="t m0 xf h5 ya0 ff3 fs3 fc0 sc0 ls5a ws105">the data and <span class="_0 blank"></span>robustness tests.<span class="_0 blank"></span> </div><div class="t m0 xf h7 ya1 ff3 fs5 fc1 sc0 ls5b ws106">1. <span class="_d blank"> </span>Real interest rate determination: an overview of approaches </div><div class="t m0 xf h5 ya2 ff3 fs3 fc0 sc0 ls13 ws107">Prevailing approaches to explaining real interest<span class="_1 blank"> </span> rates are premised on the notion that the </div><div class="t m0 xf h5 ya3 ff3 fs3 fc0 sc0 ls15 wsa3">desired (ex ante) supply of saving and the desir<span class="ls35 ws108">ed (ex ante) demand for investment determi<span class="_0 blank"></span>ne </span></div><div class="t m0 xf h5 ya4 ff3 fs3 fc0 sc0 ls29 ws109">some notional equilibrium real int<span class="_1 blank"> </span>erest rate consistent with full employment or output at </div><div class="t m0 xf h5 ya5 ff3 fs3 fc0 sc0 ls4e ws10a">potential, also known as the \u201cnatural rate\u201d. Th<span class="ls14 ws10b">is <span class="_0 blank"></span>notion ta<span class="_1 blank"> </span>kes root in the \u201cloana<span class="_1 blank"> </span>ble funds\u201d </span></div><div class="t m0 xf h5 ya6 ff3 fs3 fc0 sc0 ls2f ws10c">framework, where saving-investm<span class="ls15 ws10d">ent determinants drive the deman<span class="ls28 ws10e">d for, and supply of, funds </span></span></div><div class="t m0 xf h5 ya7 ff3 fs3 fc0 sc0 ls33 ws10f">that pin down the market-clearing interest rate (in equilibriu<span class="_1 blank"> </span>m at the marginal product of </div><div class="t m0 xf h1a ya8 ff3 fs3 fc0 sc0 ls28 wse5">capital).<span class="fs4 ls5 ws3c v1">1</span><span class="ls35 wsfc v0"> The framework therefore focuses on <span class="_0 blank"></span>the <span class="ls66 ws110">determin<span class="_1 blank"> </span>ants of sav<span class="_1 blank"> </span>ing and in<span class="_1 blank"> </span>vestment<span class="_1 blank"> </span>. </span></span></div><div class="t m0 x10 h5 ya9 ff3 fs3 fc0 sc0 ls15 ws111">On the saving side, the stan<span class="ls2a ws112">dard building block is groun<span class="ls4d ws113">ded on households\u2019 optimising </span></span></div><div class="t m0 xf h5 yaa ff3 fs3 fc0 sc0 ls65 ws114">intertemporal consumption decision<span class="_1 blank"> </span>s, as captur<span class="ls35 ws115">ed by the Euler equation. The derived saving<span class="_0 blank"></span> </span></div><div class="t m0 xf h5 yab ff3 fs3 fc0 sc0 ls35 ws116">function depends positively on unobserv<span class="ls28 ws117">ed intertemporal prefer<span class="ls4d ws118">ences and expected<span class="_0 blank"></span> </span></span></div><div class="t m0 xf h5 yac ff3 fs3 fc0 sc0 ls65 ws119">consumption growth (or output growth in equ<span class="_1 blank"> </span>ilibrium). With household heterogeneity, </div><div class="t m0 xf h5 yad ff3 fs3 fc0 sc0 ls31 ws11a">demographic variables and income distribution al<span class="ls2b ws11b">so come into play. A <span class="ls28 ws11c">higher life expectancy </span></span></div><div class="t m0 xf h5 yae ff3 fs3 fc0 sc0 ls34 ws11d">influences life-cycle decisions, raising desired savi<span class="_1 blank"> </span><span class="ls29 ws11e">ng and lowering the equilibrium real int<span class="_1 blank"> </span>erest </span></div><div class="t m0 xf h5 yaf ff3 fs3 fc0 sc0 ls2f ws52">rate. A higher dependency ratio lowers saving and <span class="ls28 ws4d">raises the real interest<span class="ls32 ws11f"> rate as <span class="_0 blank"></span>the working-</span></span></div><div class="t m0 xf h5 yb0 ff3 fs3 fc0 sc0 ls4f ws120">age population saves more than yo<span class="_0 blank"></span>unger <span class="ls28 wsf9">and older cohorts. Population growth<span class="ffa ls5c ws7"> <span class="ff3 ls19">influences </span></span></span></div><div class="t m0 xf h1b yb1 ff6 fsd fc0 sc0 ls5 ws7"> <span class="_1 blank"> </span> <span class="_3 blank"></span> </div><div class="t m0 xf h1c yb2 ff6 fs9 fc0 sc0 ls5 wse6">1<span class="ff3 fsa ls5d ws121 v3"> <span class="_0 blank"></span> <span class="_e blank"> </span>See Wicksell (1898) and Woodford (2003). As discussed <span class="ws122">in detail in Borio and Disy<span class="ls5e">atat (2011, 2015), <span class="_1 blank"> </span>despite the </span></span></span></div><div class="t m0 x11 h11 yb3 ff3 fsa fc0 sc0 ls67 ws123">term, this market for fun<span class="_1 blank"> </span>ds is in fact a market for <span class="_1 blank"> </span>goods and bears <span class="_1 blank"> </span>no relationship to the<span class="_1 blank"> </span> flow of financing that<span class="_1 blank"> </span> </div><div class="t m0 x11 h11 yb4 ff3 fsa fc0 sc0 ls1f ws69">actually underpins economic ac<span class="ls5f ws124">tivity. In fact, in contrast <span class="ls24 ws125">to current usage (eg Mankiw (<span class="ls68 ws126">2013)), even in the original </span></span></span></div><div class="t m0 x11 h11 yb5 ff3 fsa fc0 sc0 ls20 ws127">literature, \u201cloanable funds\u201d was not used as sy<span class="_0 blank"></span>nonymous <span class="ls3b ws128">with \u201csaving\u201d, as credit al<span class="ls60 ws129">so played a key role (eg </span></span></div><div class="t m0 x11 h11 yb6 ff3 fsa fc0 sc0 ls61 ws12a">Robertson (1934) and Ohlin (1937)). </div><a class="l" data-dest-detail='[6,"XYZ",0,149,-32768]'><div class="d m1" style="border-style:none;position:absolute;left:114.000000px;bottom:298.974000px;width:3.493000px;height:8.423000px;background-color:rgba(255,255,255,0.000001);"></div></a></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div> <div id="pf7" class="pf w2 hb" data-page-no="7"><div class="pc pc7 w2 hb"><img fetchpriority="low" loading="lazy" class="bi xe yb7 w3 hc" alt="" src="https://files.passeidireto.com/7e6bc13c-bf65-4cc9-ae46-d7a88a02abc2/bg7.png"><div class="t m0 xd h17 y58 ff6 fs3 fc0 sc0 ls5 ws7"> </div><div class="t m0 xe h18 y59 ff3 fse fc0 sc0 ls52 ws94">WP685 Why so low for so l<span class="_1 blank"> </span>ong? A long view of<span class="_1 blank"> </span> real interest ra<span class="_1 blank"> </span>te determina<span class="_1 blank"> </span>tion <span class="_9 blank"> </span>3</div><div class="t m0 xe h16 y5a ff6 fsc fc2 sc0 ls5 ws7"> </div><div class="t m0 xe h5 y8a ff3 fs3 fc0 sc0 ls63 ws12c">both the demographic dyna<span class="ls2a ws12d">mics and the capital-labour ratio, result<span class="_1 blank"> </span>ing in offsetting effects on </span></div><div class="t m0 xe h1d y8b ff3 fs3 fc0 sc0 ls15 ws12e">interest rates (Carvalho et el (2016)).<span class="fs4 ls5 ws3c v1">2</span><span class="ls4e ws10a v0"> Higher income<span class="_0 blank"></span> inequality in<span class="ls32 ws12f">creases saving, as richer </span></span></div><div class="t m0 xe h5 yb8 ff3 fs3 fc0 sc0 ls69 ws130">households have a higher ma<span class="ls4c ws131">rginal pr<span class="_0 blank"></span>opensity to save<span class="_0 blank"></span>. </span></div><div class="t m0 xa h5 yb9 ff3 fs3 fc0 sc0 ls4e ws132">On the investment side, firm profit maximi<span class="ls51 ws133">satio<span class="_0 blank"></span>n and the resul<span class="_0 blank"></span>tin<span class="ls29 ws134">g demand function for </span></span></div><div class="t m0 xe h5 yba ff3 fs3 fc0 sc0 ls2a ws135">capital point to the relevance of factors su<span class="_1 blank"> </span>ch as the relative supply of labour and capital,<span class="_1 blank"> </span> </div><div class="t m0 xe h5 ybb ff3 fs3 fc0 sc0 ls2a ws136">population growth, investment profitability, produ<span class="_1 blank"> </span>ctivity growth and the relative price of </div><div class="t m0 xe h5 ybc ff3 fs3 fc0 sc0 ls28 ws137">capital to that of output. C<span class="_0 blank"></span>heaper physical capi<span class="ls65 ws138">tal, eg from technological advances, means that </span></div><div class="t m0 xe h5 ybd ff3 fs3 fc0 sc0 ls4f ws139">less investment is needed to main<span class="ws13a">tain the same level of producti<span class="_0 blank"></span>on. Provided this income </span></div><div class="t m0 xe h5 ybe ff3 fs3 fc0 sc0 ls1b ws13b">effect always dominates, a<span class="_0 blank"></span>s typi<span class="ls2d ws82">cally assumed, the relative price of capital should go hand-in-</span></div><div class="t m0 xe h5 ybf ff3 fs3 fc0 sc0 ls1b ws13b">hand with hig<span class="_0 blank"></span>her desired investment, an<span class="_0 blank"></span>d hence higher real interest ra<span class="_0 blank"></span>tes. </div><div class="t m0 x12 h5 yc0 ff3 fs3 fc0 sc0 ls63 ws13c">If economies are financially integrated, the equivalent global variables matter as well. For </div><div class="t m0 xe h5 yc1 ff3 fs3 fc0 sc0 ls28 ws13d">example, the saving glut hypoth<span class="ls35 ws13e">esis (Bernanke (2005)<span class="_0 blank"></span>) posits that desired saving in emerging </span></div><div class="t m0 xe h5 yc2 ff3 fs3 fc0 sc0 ls2f ws13f">markets has put downward pressure on r<span class="_0 blank"></span>eal rate<span class="ls15 ws140">s globally. Similarly, a greater demand for safe </span></div><div class="t m0 xe h5 yc3 ff3 fs3 fc0 sc0 ls32 ws141">assets (Caballero et al (2008, 2016)) may help explain <span class="_0 blank"></span>declining risk-free rates. More generally, </div><div class="t m0 xe h1e yc4 ff3 fs3 fc0 sc0 ls15 ws7e">a higher risk premium may lower desired<span class="ls4e ws84"> investment and raise desi<span class="_0 blank"></span>red saving.<span class="fs4 ls5 ws3c v1">3</span><span class="ls5 ws7 v0"> </span></span></div><div class="t m0 xa h5 yc5 ff3 fs3 fc0 sc0 ls4e ws142">The corresponding explanations<span class="ls33 ws143"> for declining and persistently low real interest rates </span></div><div class="t m0 xe h5 yc6 ff3 fs3 fc0 sc0 ls29 ws144">follow essentially two approaches<span class="ls6e ws145">. The first, wh<span class="_1 blank"> </span>ich focuses on <span class="ffa ls58 ws12b">observed</span><span class="ls6f ws146"> re<span class="_1 blank"> </span>al interest<span class="_1 blank"> </span> rates an<span class="_1 blank"> </span>d </span></span></div><div class="t m0 xe h5 yc7 ff3 fs3 fc0 sc0 ls2f ws147">relates them directly to the evolution of th<span class="ls32 ws12f">e factors that underpin <span class="ls15 ws148">the economy\u2019s saving-</span></span></div><div class="t m0 xe h5 yc8 ff3 fs3 fc0 sc0 ls65 ws149">investment balance, comes in two variants. On<span class="_1 blank"> </span>e <span class="ls36 ws14a">is largely narrative: it tells plausible st<span class="_1 blank"> </span>ories </span></div><div class="t m0 xe h5 yc9 ff3 fs3 fc0 sc0 ls36 ws14b">relating real interest rates to it<span class="_1 blank"> </span>s determinants, typically based on informal inspection<span class="_1 blank"> </span> of the </div><div class="t m0 xe h5 yca ff3 fs3 fc0 sc0 ls4e ws14c">behaviour of the relevant <span class="_0 blank"></span>variables (eg IMF (2014)<span class="ls28 ws14d">, Bean et al (2014<span class="_0 blank"></span>), Eichengreen (2015) and </span></div><div class="t m0 xe h5 ycb ff3 fs3 fc0 sc0 ls15 ws14d">Council of Economic Advisers (2015)). The other is calibration: th<span class="ls32 ws14e">is systematically uses theory </span></div><div class="t m0 xe h5 ycc ff3 fs3 fc0 sc0 ls1b ws14f">to identify factors behind shi<span class="_0 blank"></span>fts in real interest r<span class="_0 blank"></span>ate trends, and data to cali<span class="_0 blank"></span>brate the </div><div class="t m0 xe h5 ycd ff3 fs3 fc0 sc0 ls2d ws80">corresponding structural models (<span class="ls28 ws150">eg Gagnon et al (2016), Carval<span class="ls31 ws87">ho et al (2016), Eggertsson et </span></span></div><div class="t m0 xe h5 yce ff3 fs3 fc0 sc0 ls4e wsa7">al (2017), Rachel and Smith (2017), Thwaites (2015), <span class="ls31 ws151">Vlieghe (2017)). In this<span class="ls34 ws152"> variant of the first </span></span></div><div class="t m0 xe h5 ycf ff3 fs3 fc0 sc0 ls31 ws153">approach, theory dictat<span class="ls63 ws154">es the relationships and the data<span class="ls15 ws155"> are only used to <span class="_0 blank"></span>gauge their </span></span></div><div class="t m0 xe h5 yd0 ff3 fs3 fc0 sc0 ls29 ws156">quantitative importance conditional on<span class="_1 blank"> </span> the theory<span class="ls31 ws157"> being true. The second <span class="ls33 ws158">approach is filtering: </span></span></div><div class="t m0 xe h5 yd1 ff3 fs3 fc0 sc0 ls36 ws159">this recovers equilibrium real interest<span class="_1 blank"> </span> rates st<span class="ls58 ws15a">atistically by anchoring t<span class="_1 blank"> </span>hem to some economic </span></div><div class="t m0 xe h5 yd2 ff3 fs3 fc0 sc0 ls58 ws15b">relationships, notably the link between economic slack an<span class="_1 blank"> </span>d inflation \u2013 the Phillips curve (eg </div><div class="t m0 xe h5 yd3 ff3 fs3 fc0 sc0 ls4e ws132">Laubach and Williams (2003, 2015),<span class="_0 blank"></span> Holston et al<span class="ls15 ws15c"> (2016), Justiniano and <span class="_0 blank"></span>Primiceri (2010), Del </span></div><div class="t m0 xe h5 yd4 ff3 fs3 fc0 sc0 ls2b ws15d">Negro et al (2017) and Johannsen and M<span class="_0 blank"></span>ertens (<span class="ls4e ws15e">2016)). Table 1 provides a summary of selected </span></div><div class="t m0 xe h5 yd5 ff3 fs3 fc0 sc0 ls19 ws7">studies. </div><div class="t m0 xa h5 yd6 ff3 fs3 fc0 sc0 ls28 ws15f">How far does the resulting empirical eviden<span class="ls32 ws160">ce support the hypoth<span class="ls2f ws161">esis that saving-</span></span></div><div class="t m0 xe h5 yd7 ff3 fs3 fc0 sc0 ls58 ws162">investment imbalances have driven real interest <span class="_1 blank"> </span><span class="ls36 ws163">rates to such low levels? Existing st<span class="_1 blank"> </span>udies, in </span></div><div class="t m0 xe h5 yd8 ff3 fs3 fc0 sc0 ls34 ws164">our view, have provided estimates of the extent<span class="ls15 ws165"> to which savi<span class="_0 blank"></span>ng-inves<span class="ws4e">tment determinants can </span></span></div><div class="t m0 xe h5 yd9 ff3 fs3 fc0 sc0 ls2f ws166">explain real interest rate movements <span class="ffa ls31 ws7">conditional <span class="_1 blank"> </span></span><span class="ls49 ws167">on the theory, but not <span class="ls28 ws168">convincing evidence </span></span></div><div class="t m0 xe h5 yda ff3 fs3 fc0 sc0 ls4d ws169">supporting the under<span class="ls15 ws16a">lying theory itself. Too <span class="_0 blank"></span>much of<span class="ls29 ws16b"> the theory has bee<span class="_0 blank"></span>n embedded in </span></span></div><div class="t m0 xe h5 ydb ff3 fs3 fc0 sc0 ls28 ws150">maintained hypotheses and <span class="ls36 ws16c">thus its validity has not been<span class="_1 blank"> </span> subject to a test. </span></div><div class="t m0 xa h5 ydc ff3 fs3 fc0 sc0 ls56 wsdc">This conclusion is most obvious for th<span class="_1 blank"> </span>e narrative variant of the first approach, which ne<span class="_1 blank"> </span>ver </div><div class="t m0 xe h5 ydd ff3 fs3 fc0 sc0 ls32 ws16d">quite tests the saving-inve<span class="_0 blank"></span>stment framework. Rather<span class="ls13 ws16e">, it uses it to see wh<span class="ls4c ws16f">at factors appear<span class="_0 blank"></span> to </span></span></div><div class="t m0 xe h5 yde ff3 fs3 fc0 sc0 ls36 ws170">be more consistent with the downward trend in rates. And since it largely relies on informal </div><div class="t m0 xe h5 ydf ff3 fs3 fc0 sc0 ls29 ws3a">inspection of bilateral relationships graph<span class="_1 blank"> </span>ically, it <span class="ls2d ws82">is not that hard to find some that appear to </span></div><div class="t m0 xe h1b ye0 ff6 fsd fc0 sc0 ls5 ws7"> <span class="_1 blank"> </span> <span class="_3 blank"></span> </div><div class="t m0 xe h1c ye1 ff6 fs9 fc0 sc0 ls5 wse6">2<span class="ff3 fsa ls70 ws171 v3"> <span class="_e blank"> </span>Lower population growth raise<span class="_1 blank"> </span>s the old-a<span class="_1 blank"> </span>ge dependen<span class="_1 blank"> </span>cy ratio, increasing<span class="_1 blank"> </span> the equilibrium<span class="_1 blank"> </span> real interest<span class="_1 blank"> </span> rates. But </span></div><div class="t m0 x13 h11 ye2 ff3 fsa fc0 sc0 ls71 ws172">it also raises th<span class="_1 blank"> </span>e capital-to-labour ratio<span class="_1 blank"> </span> and lowers t<span class="_1 blank"> </span>h<span class="ls6a ws173">e marginal product of capital. The <span class="_1 blank"> </span>net effect on the </span></div><div class="t m0 x13 h11 ye3 ff3 fsa fc0 sc0 ls40 ws174">equilibrium re<span class="_1 blank"> </span>al interest rate <span class="_1 blank"> </span>is a priori ambi<span class="_1 blank"> </span>guous. </div><div class="t m0 xe h1c ye4 ff6 fs9 fc0 sc0 ls5 wse6">3<span class="ff3 fsa ls5d ws175 v3"> <span class="_6 blank"></span> <span class="_e blank"> </span>The risk premium is<span class="_0 blank"></span> defined as the difference between the cost of capital and the ris<span class="_0 blank"></span>k-free rate. Risk premium </span></div><div class="t m0 x13 h11 ye5 ff3 fsa fc0 sc0 ls5f ws176">shifts may originate from a repricing of risks (eg due to demand for safe as<span class="ls6b ws177">sets, as in Gourinchas and Rey (2016) </span></div><div class="t m0 x13 h11 ye6 ff3 fsa fc0 sc0 ls6c ws178">and Del Negro et al (2017)) or changes in<span class="ls6b ws179"> underlying risks (eg prod<span class="ls3b ws17a">uctivity gro<span class="_1 blank"> </span>wth uncertaint<span class="ls6d ws17b">y, as in Marx et al </span></span></span></div><div class="t m0 x13 h11 ye7 ff3 fsa fc0 sc0 ls6a ws17c">(2017) or Vlieghe (2017)). </div><a class="l" data-dest-detail='[7,"XYZ",0,126,-32768]'><div class="d m1" style="border-style:none;position:absolute;left:450.300000px;bottom:571.073000px;width:3.493000px;height:8.424000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[7,"XYZ",0,164,-32768]'><div class="d m1" style="border-style:none;position:absolute;left:274.560000px;bottom:755.993000px;width:3.493000px;height:8.424000px;background-color:rgba(255,255,255,0.000001);"></div></a></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div> <div id="pf8" class="pf w2 hb" data-page-no="8"><div class="pc pc8 w2 hb"><img fetchpriority="low" loading="lazy" class="bi xf ye8 w3 hc" alt="" src="https://files.passeidireto.com/7e6bc13c-bf65-4cc9-ae46-d7a88a02abc2/bg8.png"><div class="t m0 xf h19 y89 ffb fsb fc0 sc0 ls5 ws7"> </div><div class="t m0 xf h18 y59 ff3 fse fc0 sc0 ls52 ws94">4 <span class="_c blank"> </span>WP685 Why so low for so l<span class="_1 blank"> </span>ong? A long view of<span class="_1 blank"> </span> real interest ra<span class="_1 blank"> </span>te determina<span class="_1 blank"> </span>tion</div><div class="t m0 xf h16 y5a ff6 fsc fc2 sc0 ls5 ws7"> </div><div class="t m0 xf h5 y8a ff3 fs3 fc0 sc0 ls49 ws17e">hold for at least p<span class="_0 blank"></span>art of the time. This t<span class="_0 blank"></span>ype of an<span class="ls57 ws17f">alysis is best interpreted as a first look at the </span></div><div class="t m0 xf h5 y8b ff3 fs3 fc0 sc0 ls15 wsd2">data and as a basis for a more in<span class="ls28 wsd3">-depth evaluation. Nor, in all fa<span class="ls4e wse2">irness, does it pretend to be </span></span></div><div class="t m0 xf h5 ye9 ff3 fs3 fc0 sc0 ls4c ws180">more than that. </div><div class="t m0 x10 h5 yea ff3 fs3 fc0 sc0 ls4e wsa7">Calibration based on structural models \u2013 <span class="_0 blank"></span>the se<span class="ls13 ws181">cond variant of the first approach<span class="_1 blank"> </span> \u2013 takes </span></div><div class="t m0 xf h5 yeb ff3 fs3 fc0 sc0 ls32 ws182">the narrative approach much furthe<span class="ws183">r. It quantifies the effect that<span class="ls58 ws184"> specific saving and investment </span></span></div><div class="t m0 xf h5 yec ff3 fs3 fc0 sc0 ls31 ws185">factors would have within a fully specified theore<span class="ls58 ws186">tical model that is calib<span class="ls4e wsc9">rated to fit the data </span></span></div><div class="t m0 xf h5 yed ff3 fs3 fc0 sc0 ls4d ws187">as closely as possible. Hence, it provides in<span class="ls58 ws188">formation about the relative importance of the<span class="_1 blank"> </span> </span></div><div class="t m0 xf h5 yee ff3 fs3 fc0 sc0 ls63 ws189">different factors in a general equilibrium setting, while at the same time addressing the </div><div class="t m0 xf h5 yef ff3 fs3 fc0 sc0 ls63 ws164">challenges raised by regi<span class="ls2c ws18a">me changes and expectations. Neverth<span class="_1 blank"> </span>eless, just as with the informal </span></div><div class="t m0 xf h5 yf0 ff3 fs3 fc0 sc0 ls58 ws18b">approach, the validity of the <span class="_1 blank"> </span>underlying theory <span class="_1 blank"> </span><span class="ls28 ws168">is not tested. Moreover, the models typically </span></div><div class="t m0 xf h5 yf1 ff3 fs3 fc0 sc0 ls31 ws18c">include parameters that are <span class="ls35 ws18d">poorly identified and have no<span class="_0 blank"></span><span class="ls29 ws18e"> clear benchmark values. The </span></span></div><div class="t m0 xf h5 yf2 ff3 fs3 fc0 sc0 ls32 ws18f">resulting large number of degrees of freedom comp<span class="_0 blank"></span>licate the evaluation of the final resul<span class="_0 blank"></span>ts: </div><div class="t m0 xf h5 yf3 ff3 fs3 fc0 sc0 ls28 wsd2">there is a risk that the importan<span class="ls4f ws190">ce of particular fac<span class="_0 blank"></span>tors may be overstated or sp<span class="_0 blank"></span>ecific periods </span></div><div class="t m0 xf h1f yf4 ff3 fs3 fc0 sc0 ls4f wse4">\u201coverfitted\u201d.<span class="ff6 fs4 ls5 ws17d v1">4</span><span class="ls5 ws7 v0"> </span></div><div class="t m0 x10 h5 yf5 ff3 fs3 fc0 sc0 ls2a ws191">The filtering approach faces similar challenges. He<span class="ls15 ws140">re, the role of a priori restrictions on the </span></div><div class="t m0 xf h5 yf6 ff3 fs3 fc0 sc0 ls15 ws192">data is critical. In particular, one typical key <span class="ls57 ws193">maintained hypothesis is <span class="ls73 ws194">that<span class="_1 blank"> </span> inflation provides </span></span></div><div class="t m0 xf h5 yf7 ff3 fs3 fc0 sc0 ls28 ws13d">the right signal to identify cyc<span class="ls35 ws195">lical deviations of the market rate<span class="ls36 ws159"> from the natural rate. All else </span></span></div><div class="t m0 xf h5 yf8 ff3 fs3 fc0 sc0 ls58 ws196">equal, if, say, inflation increases, it is inferred that output is above potential (Phillips curve), </div><div class="t m0 xf h5 yf9 ff3 fs3 fc0 sc0 ls2c ws197">which in turn implies that<span class="_1 blank"> </span> the market rate is be<span class="ls63 ws198">low the natural rate (IS curve). And yet, the link </span></div><div class="t m0 xf h5 yfa ff3 fs3 fc0 sc0 ls29 ws199">between economic slack and inflat<span class="ls47 ws19a">ion has proved<span class="_0 blank"></span> rather weak <span class="ls63 ws19b">and elusive over the <span class="_0 blank"></span>years, </span></span></div><div class="t m0 xf h5 yfb ff3 fs3 fc0 sc0 ls63 ws19c">making any firm inferences suspect (ie Forbes<span class="ws58"> et al (2017), Stock<span class="_0 blank"></span> and Watson (2007), Borio </span></div><div class="t m0 xf h5 yfc ff3 fs3 fc0 sc0 ls35 ws19d">(2017a)). Indeed, recent work has found that fi<span class="_0 blank"></span>na<span class="ls32 ws19e">ncial cycle proxies capture cyclical outp<span class="_0 blank"></span>ut </span></div><div class="t m0 xf h5 yfd ff3 fs3 fc0 sc0 ls2f wsca">variations better than inflation <span class="ls15 ws111">(Borio et al (2017), Kiley (2015)), <span class="_0 blank"></span><span class="ls4f ws19f">yielding natural interest rate </span></span></div><div class="t m0 xf h5 yfe ff3 fs3 fc0 sc0 ls51 ws1a0">estimates that are somew<span class="_0 blank"></span>hat hi<span class="ls63 ws1a1">gher and decline by le<span class="ls15 ws1a2">ss (Juselius et al (2017)). Moreover, </span></span></div><div class="t m0 xf h5 yff ff3 fs3 fc0 sc0 ls29 ws1a3">filtering approaches typically relate the un<span class="_1 blank"> </span>o<span class="ls2f ws1a4">bserved natural rate to other unobservable </span></div><div class="t m0 xf h5 y100 ff3 fs3 fc0 sc0 ls35 ws1a5">variables in the system, such as<span class="ls31 ws1a6"> potential growth and preferences, giving rise to many degrees </span></div><div class="t m0 xf h5 y101 ff3 fs3 fc0 sc0 ls35 ws195">of freedom when fitting the story. T<span class="_0 blank"></span>hus, the ma<span class="ls4d ws7f">intained hypothesis ends up having a decisive </span></div><div class="t m0 xf h5 y102 ff3 fs3 fc0 sc0 ls35 wsd9">influence on the end-result (see Lubik and Matthes (2015) for a si<span class="_0 blank"></span>milar critique). And as with </div><div class="t m0 xf h5 y103 ff3 fs3 fc0 sc0 ls34 wsbe">calibration, the risk of \u201coverfitting\u201d in any given sample is material. </div><div class="t m0 xf h1b y104 ff6 fsd fc0 sc0 ls5 ws7"> <span class="_1 blank"> </span> <span class="_3 blank"></span> </div><div class="t m0 xf h1c y105 ff6 fs9 fc0 sc0 ls5 wse6">4<span class="ff3 fsa ls1f ws1a7 v3"> <span class="_1 blank"> </span> <span class="_f blank"> </span>To be more precise, in calibration, the researc<span class="_0 blank"></span>her choses values for both the stru<span class="_0 blank"></span>ctural parameters and unobserved<span class="_0 blank"></span> </span></div><div class="t m0 x11 h11 y106 ff3 fsa fc0 sc0 ls74 ws1a8">shock processe<span class="_1 blank"> </span>s to mimic <span class="_1 blank"> </span>some key fe<span class="_1 blank"> </span>atures of the data<span class="_1 blank"> </span>. These commo<span class="_1 blank"> </span>nly include stea<span class="_1 blank"> </span>dy-state rati<span class="_1 blank"> </span>os between </div><div class="t m0 x11 h11 y107 ff3 fsa fc0 sc0 ls75 ws1a9">variables, seco<span class="_1 blank"> </span>nd mome<span class="_1 blank"> </span>nts of selected va<span class="_1 blank"> </span>riables an<span class="_1 blank"> </span>d so on. Yet the <span class="_1 blank"> </span>key features t<span class="_1 blank"> </span>ypicall<span class="_1 blank"> </span>y only constitute<span class="_1 blank"> </span> a small </div><div class="t m0 x11 h11 y108 ff3 fsa fc0 sc0 ls3b ws17a">subset of the mo<span class="_1 blank"> </span>del\u2019s full im<span class="_1 blank"> </span>plications for the<span class="_1 blank"> </span> data and <span class="ls71 ws1aa">there<span class="_1 blank"> </span> is less discip<span class="_1 blank"> </span>line in th<span class="ls43 ws1ab">e<span class="_1 blank"> </span> remaining direct<span class="_1 blank"> </span>ions. This </span></span></div><div class="t m0 x11 h11 y109 ff3 fsa fc0 sc0 ls67 ws1ac">gives the investi<span class="_1 blank"> </span>gator considerable<span class="_1 blank"> </span> degrees of<span class="_1 blank"> </span> freedom wh<span class="ls72 ws1ad">en fitting the<span class="_1 blank"> </span> features of interest<span class="_1 blank"> </span> at the expense of<span class="_1 blank"> </span> </span></div><div class="t m0 x11 h11 y10a ff3 fsa fc0 sc0 ls76 ws1ae">general model<span class="_1 blank"> </span> fit. Equally<span class="_1 blank"> </span> problematic<span class="_1 blank"> </span> is the hi<span class="_1 blank"> </span>gh reliance<span class="_1 blank"> </span> on persiste<span class="_1 blank"> </span>nt shock pro<span class="_1 blank"> </span>cesses or <span class="_1 blank"> </span>unobserved st<span class="_1 blank"> </span>ochastic </div><div class="t m0 x11 h11 y10b ff3 fsa fc0 sc0 ls77 ws1af">trends. With<span class="_1 blank"> </span> sufficie<span class="_1 blank"> </span>ntly man<span class="_1 blank"> </span>y such p<span class="_1 blank"> </span>rocesse<span class="_1 blank"> </span>s, the mode<span class="_1 blank"> </span>l can <span class="_1 blank"> </span>generate a<span class="_1 blank"> </span> perfect<span class="_1 blank"> </span> fit with<span class="_1 blank"> </span>out an i<span class="_1 blank"> </span>ncrease <span class="_1 blank"> </span>in </div><div class="t m0 x11 h11 y10c ff3 fsa fc0 sc0 ls3d ws1b0">predictive power<span class="_1 blank"> </span> \u2013 a ca<span class="_1 blank"> </span>se of \u201coverf<span class="_1 blank"> </span>itting\u201d. </div><a class="l" data-dest-detail='[8,"XYZ",0,169,-32768]'><div class="d m1" style="border-style:none;position:absolute;left:132.420000px;bottom:589.914000px;width:3.603000px;height:8.625000px;background-color:rgba(255,255,255,0.000001);"></div></a></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div> <div id="pf9" class="pf w2 hb" data-page-no="9"><div class="pc pc9 w2 hb"><img fetchpriority="low" loading="lazy" class="bi x14 y10d w4 h20" alt="" src="https://files.passeidireto.com/7e6bc13c-bf65-4cc9-ae46-d7a88a02abc2/bg9.png"><div class="t m0 xd h17 y58 ff6 fs3 fc0 sc0 ls5 ws7"> </div><div class="t m0 xe h18 y59 ff3 fse fc0 sc0 ls52 ws94">WP685 Why so low for so l<span class="_1 blank"> </span>ong? A long view of<span class="_1 blank"> </span> real interest ra<span class="_1 blank"> </span>te determina<span class="_1 blank"> </span>tion <span class="_9 blank"> </span>5</div><div class="t m0 x14 h16 y5a ff6 fsc fc2 sc0 ls5 ws7"> </div><div class="t m0 xa h5 y10e ff3 fs3 fc0 sc0 ls56 ws1b6">All this highligh<span class="_1 blank"> </span>ts the importan<span class="_1 blank"> </span>ce of confro<span class="ls81 ws1b7">nt<span class="_1 blank"> </span>ing t<span class="_1 blank"> </span>he hyp<span class="_1 blank"> </span>othes<span class="_1 blank"> </span>is mor<span class="_1 blank"> </span><span class="ls58 ws1b8">e directly with the </span></span></div><div class="t m0 x14 h5 y10f ff3 fs3 fc0 sc0 ls36 ws1b9">data, examining systematically the relati<span class="ls2d ws1ba">onship between real <span class="ls82 ws1bb">int<span class="_1 blank"> </span>erest<span class="_1 blank"> </span> rate<span class="_1 blank"> </span>s and <span class="ffa ls1c ws1b1">observable</span><span class="ls5 ws7"> </span></span></span></div><div class="t m0 x14 h5 y110 ff3 fs3 fc0 sc0 ls4e ws104">variables. And yet, there are very few studi<span class="_0 blank"></span>es that <span class="ls2c ws1bc">do this. Much of this work examines an earlier </span></div><div class="t m0 x14 h5 y111 ff3 fs3 fc0 sc0 ls4e ws1bd">period \u2013 the surge of real<span class="_0 blank"></span> interest rates in<span class="ls4c ws1be"> the early <span class="_0 blank"></span>1980s (Blanchard and Summers<span class="_0 blank"></span> (1984), </span></div><div class="t m0 x14 h5 y112 ff3 fs3 fc0 sc0 ls4f ws1bf">Barro and Sala-i-Martin (1990), Orr et a<span class="_0 blank"></span>l (1995)). Hardly any have covered the more recent </div><div class="t m0 x14 h5 y113 ff3 fs3 fc0 sc0 ls4d ws1c0">phase of declining rates. An ex<span class="ls15 ws1c1">ception is Lunsford and<span class="_0 blank"></span> West (2017), who focus on the United </span></div><div class="t m0 x14 h5 y114 ff3 fs3 fc0 sc0 ls2b ws1c2">States for the period <span class="_0 blank"></span>1890-2015 and evaluate<span class="_0 blank"></span> the <span class="ls15 ws1c3">bivariate correlation between real interest<span class="_0 blank"></span> </span></div><div class="t m0 x14 h5 y115 ff3 fs3 fc0 sc0 ls2f ws1c4">rates and a number of factors. The authors find we<span class="_0 blank"></span><span class="ls63 ws1c5">ak evidence overall, particularly for variables </span></div><div class="t m0 x14 h5 y116 ff3 fs3 fc0 sc0 ls4d wsaf">representing aggregate growth <span class="ls11 ws1c6">(GDP, consump<span class="_0 blank"></span>tion, total<span class="_0 blank"></span> factor productivity (<span class="_0 blank"></span>TFP)), th<span class="_0 blank"></span>ough </span></div><div class="t m0 x14 h5 y117 ff3 fs3 fc0 sc0 ls4c ws1c7">they do find some supp<span class="_0 blank"></span>ort for demographic va<span class="ls2f ws1c8">riables (for <span class="_0 blank"></span>the weak explanatory power of<span class="_0 blank"></span> </span></div><div class="t m0 x14 h5 y118 ff3 fs3 fc0 sc0 ls4d ws1c9">output growth, see also Hamilton et al (<span class="ws1ca">2015)). Our paper c<span class="_0 blank"></span>omplements this work by </span></div><div class="t m0 x14 h5 y119 ff3 fs3 fc0 sc0 ls2b ws1cb">considering a wider set of countries, conducti<span class="ls33 ws1cc">ng joint-specification analysis to allow for </span></div><div class="t m0 x14 h5 y11a ff3 fs3 fc0 sc0 ls2a ws1cd">interactions between explan<span class="_1 blank"> </span>atory variables, <span class="ls33 ws1ce">and exploring the role of monetary policy. </span></div><div class="t m0 x14 h21 y11b ff3 fsd fc0 sc0 ls78 ws1cf">A summary of selected studies on <span class="ls79 ws1d0">real interest rate determination <span class="_10 blank"> </span><span class="fs3 ls36 ws16c v4">Table 1</span></span></div><div class="t m0 x15 h22 y11c ff3 fsf fc0 sc0 ls83 ws1d1">Study Methodology <span class="_11 blank"></span>Coverage </div><div class="t m0 x16 h22 y11d ff3 fsf fc0 sc0 ls84 ws1d2">Key factors<span class="ff9 ls5 ws7"> </span></div><div class="t m2 x17 h22 y11e ff3 fsf fc0 sc0 ls85 ws1d3">Growth & productivity<span class="ff9 ls5 ws7"> </span></div><div class="t m2 x18 h22 y11f ff3 fsf fc0 sc0 ls86 ws1b2">Demographi<span class="_0 blank"></span>cs<span class="ff9 ls5 ws7"> </span></div><div class="t m2 x19 h22 y120 ff3 fsf fc0 sc0 ls87 ws1d4">Relative price of capital<span class="ff9 ls5 ws7"> </span></div><div class="t m2 x1a h22 y121 ff3 fsf fc0 sc0 ls88 ws1b3">Inequality<span class="ff9 ls5 ws7"> </span></div><div class="t m2 x1b h22 y122 ff3 fsf fc0 sc0 ls88 ws1d5">Global saving glut<span class="ff9 ls5 ws7"> </span></div><div class="t m2 x1c h22 y123 ff3 fsf fc0 sc0 ls83 ws1d6">Demand for safe assets<span class="ff9 ls5 ws7"> </span></div><div class="t m2 x1d h22 y124 ff3 fsf fc0 sc0 ls89 ws1d7">Risk premium<span class="ff9 ls5 ws7"> </span></div><div class="t m2 x1e h22 y125 ff3 fsf fc0 sc0 ls8a ws1b4">Others<span class="ff9 ls5 ws7"> </span></div><div class="t m0 x14 h22 y126 ff3 fsf fc0 sc0 ls7a ws1d8">IMF <span class="_12 blank"></span>(2014) <span class="_13 blank"> </span>Narrative <span class="_14 blank"> </span>Global <span class="_15 blank"> </span>X X X <span class="_5 blank"></span>X <span class="_16 blank"> </span> </div><div class="t m0 x14 h22 y127 ff3 fsf fc0 sc0 ls8b ws1d9">Bean et al (2015) <span class="_17 blank"> </span>Narrative <span class="_18 blank"> </span>Global <span class="_19 blank"> </span> <span class="_1a blank"> </span>X <span class="_1b blank"> </span>X <span class="_1b blank"> </span>X <span class="_1b blank"> </span>X <span class="_1b blank"> </span>X <span class="_1a blank"> </span> <span class="_1c blank"> </span> </div><div class="t m0 x14 h22 y128 ff3 fsf fc0 sc0 ls83 ws1d6">Eichengreen (2015) <span class="_1d blank"> </span>Narrati<span class="ls7b ws1da">ve <span class="_1e blank"> </span>US <span class="_1f blank"> </span>X <span class="_20 blank"></span>X <span class="_3 blank"></span>X X <span class="_21 blank"> </span> <span class="_16 blank"> </span> </span></div><div class="t m0 x14 h22 y129 ff3 fsf fc0 sc0 ls7a ws1d8">CEA <span class="_12 blank"></span>(2015) <span class="_22 blank"> </span>Narrative <span class="_14 blank"> </span>Global <span class="_15 blank"> </span>X <span class="_20 blank"></span>X <span class="_21 blank"> </span> X <span class="_20 blank"></span>X <span class="_20 blank"></span>X <span class="_1f blank"> </span> </div><div class="t m0 x14 h22 y12a ff3 fsf fc0 sc0 ls88 ws1d5">Goodhart and Pradhan (2017) <span class="_23 blank"> </span>Narrative <span class="_18 blank"> </span>Global <span class="_19 blank"> </span> <span class="_1a blank"> </span>X <span class="_1a blank"> </span> <span class="_24 blank"> </span> <span class="_24 blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_1c blank"> </span> </div><div class="t m0 x14 h22 y12b ff3 fsf fc0 sc0 ls8c ws1db">Gagnon et al (2016)<span class="_0 blank"></span> <span class="_26 blank"> </span>Calibration <span class="_27 blank"> </span>US <span class="_1c blank"> </span> <span class="_1a blank"> </span>X <span class="_1a blank"> </span> <span class="_25 blank"> </span> <span class="_24 blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_1c blank"> </span> </div><div class="t m0 x14 h22 y12c ff3 fsf fc0 sc0 ls83 ws1d6">Carvalho et al (2016) <span class="_28 blank"> </span>Calibration <span class="_1e blank"> </span>Global <span class="_19 blank"> </span> <span class="_1a blank"> </span>X <span class="_1a blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_24 blank"> </span> <span class="_24 blank"> </span> <span class="_1c blank"> </span> </div><div class="t m0 x14 h22 y12d ff3 fsf fc0 sc0 ls8d ws1dc">Rachel and Smith (2<span class="_0 blank"></span>017) <span class="_29 blank"> </span>Calibr<span class="ls7c ws1dd">ation <span class="_2a blank"> </span>Global <span class="_2b blank"> </span>X X X X X <span class="_21 blank"> </span> <span class="_2c blank"> </span>X <span class="_16 blank"> </span> </span></div><div class="t m0 x14 h22 y12e ff3 fsf fc0 sc0 ls8e ws1de">Thwaites <span class="_2d blank"></span>(2015) <span class="_2e blank"> </span>Calibration <span class="_2f blank"> </span>Global <span class="_15 blank"> </span> <span class="_3 blank"></span>X <span class="_20 blank"></span> <span class="_1f blank"> </span> </div><div class="t m0 x14 h22 y12f ff3 fsf fc0 sc0 ls88 ws1df">Vlieghe <span class="_2d blank"></span>(2017) <span class="_30 blank"> </span>Calibration <span class="_31 blank"> </span>UK <span class="_32 blank"> </span>X <span class="_20 blank"></span> <span class="_20 blank"></span>X <span class="_32 blank"> </span> </div><div class="t m0 x14 h22 y130 ff3 fsf fc0 sc0 ls8f ws1e0">Marx et al (2017) <span class="_33 blank"> </span>Calibration <span class="_1e blank"> </span>Global <span class="_2f blank"> </span>X <span class="_1b blank"> </span>X <span class="_1a blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_24 blank"> </span> <span class="_1a blank"> </span>X <span class="_34 blank"> </span> </div><div class="t m0 x14 h22 y131 ff3 fsf fc0 sc0 ls7d ws1e1">Eggertsson et al (2017) <span class="_35 blank"> </span>Calibration <span class="_27 blank"> </span>US <span class="_34 blank"> </span>X <span class="_36 blank"> </span>X <span class="_1b blank"> </span>X <span class="_1b blank"> </span>X <span class="_1a blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_1c blank"> </span> </div><div class="t m0 x14 h22 y132 ff3 fsf fc0 sc0 ls8c ws1db">Del Negro et al (2017)<span class="_0 blank"></span> <span class="_37 blank"> </span>Filtering <span class="_38 blank"> </span>US <span class="_34 blank"> </span>X <span class="_1a blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_24 blank"> </span> <span class="_24 blank"> </span> <span class="_1a blank"> </span>X <span class="_34 blank"> </span> </div><div class="t m0 x14 h22 y133 ff3 fsf fc0 sc0 ls90 ws1e2">Laubach <span class="_2d blank"></span>and <span class="_2d blank"></span>Williams <span class="_2d blank"></span>(2003) <span class="_39 blank"> </span>Filtering <span class="_3a blank"> </span>US <span class="_32 blank"> </span>X <span class="_20 blank"></span> <span class="_1f blank"> </span> </div><div class="t m0 x14 h22 y134 ff3 fsf fc0 sc0 ls83 ws1d6">Holston et al (2016) <span class="_3b blank"> </span>Filtering <span class="_3c blank"> </span>Four advanced<span class="_2b blank"> </span>X <span class="_1a blank"> </span> <span class="_25 blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_24 blank"> </span> <span class="_24 blank"> </span> <span class="_1c blank"> </span> </div><div class="t m0 x14 h22 y135 ff3 fsf fc0 sc0 ls88 ws1d5">Justiniano and Primiceri (2010) <span class="_3d blank"> </span>Filtering <span class="_3e blank"> </span>US <span class="_34 blank"> </span>X <span class="_1a blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_24 blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_2a blank"> </span> </div><div class="t m0 x14 h22 y136 ff3 fsf fc0 sc0 ls84 ws1d2">Clarida (2017) <span class="_3f blank"> </span>Filtering <span class="_3c blank"> </span>Fo<span class="ls89 ws1e3">ur <span class="_2d blank"></span>advanced<span class="_2b blank"> </span>X <span class="_3 blank"></span> <span class="_2b blank"> </span> </span></div><div class="t m0 x14 h22 y137 ff3 fsf fc0 sc0 ls89 ws1d7">Gourinchas and Rey (2016) <span class="_2f blank"> </span>Predictive regression <span class="_40 blank"> </span><span class="ls88 ws1e4 v5">Four <span class="_2d blank"></span>advanced<span class="_41 blank"> </span> </span></div><div class="t m0 x1f h23 y137 ff3 fsf fc0 sc0 ls7e ws7"> <span class="fs10 ls91 ws1b5 v6">Consumption</span></div><div class="t m0 x20 h24 y138 ff3 fs10 fc0 sc0 ls7f ws7">-to-wealth </div><div class="t m0 x14 h22 y139 ff3 fsf fc0 sc0 ls92 ws1e5">Hamilton et al (2015) <span class="_42 blank"> </span>Long-run correlation <span class="_43 blank"> </span>Global <span class="_2f blank"> </span>X <span class="_1a blank"> </span> <span class="_25 blank"> </span> <span class="_24 blank"> </span> <span class="_24 blank"> </span> <span class="_24 blank"> </span> <span class="_25 blank"> </span> <span class="_2a blank"> </span> </div><div class="t m0 x14 h23 y13a ff3 fsf fc0 sc0 ls88 ws1d5">Lunsford and West (2017) <span class="_44 blank"> </span>Long-run correlation <span class="_31 blank"> </span><span class="ls80 ws1e6 v5">US <span class="_16 blank"> </span>X X X X X <span class="_2c blank"> </span> <span class="_21 blank"> </span>X <span class="_45 blank"> </span></span><span class="fs10 ls93 ws7 v6">Money </span></div><div class="t m0 x21 h24 y13b ff3 fs10 fc0 sc0 ls94 ws7">growth </div></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div> <div id="pfa" class="pf w2 hb" data-page-no="a"><div class="pc pca w2 hb"><img fetchpriority="low" loading="lazy" class="bi xf y13c w3 hc" alt="" src="https://files.passeidireto.com/7e6bc13c-bf65-4cc9-ae46-d7a88a02abc2/bga.png"><div class="t m0 xf h19 y89 ffb fsb fc0 sc0 ls5 ws7"> </div><div class="t m0 xf h18 y59 ff3 fse fc0 sc0 ls52 ws94">6 <span class="_c blank"> </span>WP685 Why so low for so l<span class="_1 blank"> </span>ong? A long view of<span class="_1 blank"> </span> real interest ra<span class="_1 blank"> </span>te determina<span class="_1 blank"> </span>tion</div><div class="t m0 xf h16 y5a ff6 fsc fc2 sc0 ls5 ws7"> </div><div class="t m0 x10 h5 y8a ff3 fs3 fc0 sc0 ls32 ws1ef">Indeed, a common premise of all the traditional a<span class="ls2b ws1f0">pproaches is <span class="_0 blank"></span>that real interest rates over </span></div><div class="t m0 xf h5 y8b ff3 fs3 fc0 sc0 ls31 ws1f1">long horizons are determined <span class="ffa ls34 ws1e7">exclusively</span><span class="lsab ws7"> b<span class="_3 blank"></span>y<span class="_20 blank"></span> r<span class="_3 blank"></span>e<span class="_3 blank"></span>a<span class="_20 blank"></span>l<span class="_3 blank"></span> f<span class="_20 blank"></span>a<span class="_3 blank"></span>c<span class="_20 blank"></span>t<span class="_3 blank"></span>o<span class="_20 blank"></span>r<span class="_3 blank"></span>s<span class="_20 blank"></span>.<span class="_3 blank"></span> M<span class="_20 blank"></span>o<span class="_3 blank"></span>n<span class="_3 blank"></span>e<span class="_20 blank"></span>t<span class="_3 blank"></span>a<span class="_20 blank"></span>r<span class="_3 blank"></span>y<span class="_20 blank"></span> p<span class="_3 blank"></span>o<span class="_20 blank"></span>l<span class="_3 blank"></span>i<span class="_3 blank"></span>c<span class="_20 blank"></span>y<span class="_3 blank"></span> e<span class="_20 blank"></span>x<span class="_3 blank"></span>e<span class="_20 blank"></span>r<span class="_3 blank"></span>t<span class="_20 blank"></span>s<span class="_3 blank"></span> o<span class="_3 blank"></span>n<span class="_20 blank"></span>l<span class="_3 blank"></span>y<span class="_20 blank"></span> a<span class="_20 blank"></span> </span></div><div class="t m0 xf h5 ye9 ff3 fs3 fc0 sc0 ls33 ws1f2">transitory influence, which can be entirely ig<span class="ls32 ws1f3">nored (narrative and calibration analysis) or </span></div><div class="t m0 xf h5 y13d ff3 fs3 fc0 sc0 ls32 wsbc">filtered out (filtering analysis). <span class="ls36 ws5b">The maintained assumption is that monet<span class="_1 blank"> </span>ary policy is neutral in </span></div><div class="t m0 xf h5 y13e ff3 fs3 fc0 sc0 ls35 ws1f4">the long run.<span class="_0 blank"></span> For example, Del Negro et al (2017, p<span class="_0 blank"></span> 1) describe the natural ra<span class="_0 blank"></span>te as \u201c\u2026 the </div><div class="t m0 xf h5 y13f ff3 fs3 fc0 sc0 ls33 ws1f5">counterfactual rate that<span class="_1 blank"> </span> would be observed \u2018in the absence\u2019 of monetary policy\u201d. Still, in our </div><div class="t m0 xf h5 y140 ff3 fs3 fc0 sc0 ls19 ws1f6">view, the notion that in a comp<span class="_0 blank"></span><span class="ls34 ws1f7">lex monetary economy it is po<span class="ls65 ws1f8">ssible to<span class="_1 blank"> </span> cleanly delineate a </span></span></div><div class="t m0 xf h5 y141 ff3 fs3 fc0 sc0 ls4e ws1f9">\u201cmonetary veil\u201d from the underlyin<span class="ls32 ws1fa">g real drivers is an exceeding<span class="ls28 ws1fb">ly strong presumption. T<span class="_0 blank"></span>his </span></span></div><div class="t m0 xf h5 y142 ff3 fs3 fc0 sc0 ls17 ws1fc">presumption has<span class="_0 blank"></span> not been sufficiently s<span class="_0 blank"></span>crutinis<span class="ls58 ws1fd">ed. In Section 3 we explore its validity and </span></div><div class="t m0 xf h5 y143 ff3 fs3 fc0 sc0 ls5a ws105">usefulness in the c<span class="_0 blank"></span>urrent context.<span class="_0 blank"></span> </div><div class="t m0 xf h7 y144 ff3 fs5 fc1 sc0 ls18 ws1fe">2. <span class="_d blank"> </span>Real interest rate determination: the role of real factors </div><div class="t m0 xf h5 y145 ff3 fs3 fc0 sc0 ls1c ws1ff">We next exploit long his<span class="_0 blank"></span>torical data and cross-co<span class="_0 blank"></span>untry variation to tes<span class="_0 blank"></span>t for long-term </div><div class="t m0 xf h5 y146 ff3 fs3 fc0 sc0 ls31 ws200">relationships between real interest rates an<span class="ls28 ws201">d the saving-investment determinants suggested </span></div><div class="t m0 xf h5 y147 ff3 fs3 fc0 sc0 ls4d ws7f">by theory. We impose no prior <span class="ls2f ws202">restrictions on these <span class="ls28 ws45">relationships and allow the data to speak </span></span></div><div class="t m0 xf h5 y148 ff3 fs3 fc0 sc0 ls2a ws203">about their nature and stability. Before tu<span class="_1 blank"> </span>rning to the data and estimation<span class="_1 blank"> </span>, we outline the </div><div class="t m0 xf h5 y149 ff3 fs3 fc0 sc0 ls35 wsfc">essential elements of our empirical approach.<span class="_0 blank"></span> </div><div class="t m0 xf h25 y14a ff3 fs11 fc1 sc0 ls95 ws204">2.1 <span class="_46 blank"> </span>Essential elements of the empirical strategy<span class="ff5 ls5 ws7"> </span></div><div class="t m0 xf h5 y14b ff3 fs3 fc0 sc0 ls17 ws146">As noted, the sta<span class="_0 blank"></span>ndard saving-investme<span class="_0 blank"></span>nt fram<span class="ls2f ws205">ework relies on the assumption that money<span class="_0 blank"></span> is </span></div><div class="t m0 xf h5 y14c ff3 fs3 fc0 sc0 ls47 ws206">neutral \u201cin the long r<span class="_0 blank"></span>un\u201d, so that only re<span class="_0 blank"></span>al factor<span class="ls4e ws207">s drive real interest rates. Of course, strictly </span></div><div class="t m0 xf h5 y14d ff3 fs3 fc0 sc0 ls4f ws208">speaking, the \u201clong run\u201d is an <span class="_0 blank"></span>analytical concept.<span class="ls82 ws209"> It is the r<span class="_1 blank"> </span>esu<span class="_1 blank"> </span>lt of a th<span class="_1 blank"> </span>ought<span class="_1 blank"> </span> experi<span class="_1 blank"> </span>ment<span class="_1 blank"> </span>: it </span></div><div class="t m0 xf h5 y14e ff3 fs3 fc0 sc0 ls13 ws20a">refers to a situation in which all the variables <span class="ls4e wsb2">in the system, most nota<span class="ls4d">bly prices, <span class="_0 blank"></span>have been </span></span></div><div class="t m0 xf h5 y14f ff3 fs3 fc0 sc0 ls34 ws164">allowed to adjust (a st<span class="ls35 ws20b">eady state). For the empi<span class="ls2b ws20c">rical analysis, that concept has to be translated </span></span></div><div class="t m0 xf h5 y150 ff3 fs3 fc0 sc0 ls2d ws82">into calendar time. </div><div class="t m0 x22 h5 y151 ff3 fs3 fc0 sc0 ls29 ws3a">Concretely, this can be done as follows: </div><div class="t m0 x23 h26 y152 ffc fs3 fc0 sc0 ls5">\ue74e</div><div class="t m0 x24 h27 y153 ffc fs10 fc0 sc0 ls96">\uebe7<span class="fs3 lsac v7">=\ue74e</span></div><div class="t m0 x25 h28 y154 ffc fs10 fc0 sc0 ls97">\u2217<span class="fs3 ls5 ws1e8 v3">(<span class="v8">\ue73a</span></span></div><div class="t m0 x26 h27 y153 ffc fs10 fc0 sc0 ls98">\uebe7<span class="fs3 ls99 ws1e9 v7">;\ue003<span class="_4 blank"></span>\ue7da</span></div><div class="t m0 x8 h26 y155 ffc fs3 fc0 sc0 ls9a">)<span class="lsad v8">+\ue74e</span></div><div class="t m0 x27 h29 y153 ffc fs10 fc0 sc0 ls5">\uebe7</div><div class="t m0 x28 h29 y156 ffc fs10 fc0 sc0 lsae ws1ea">\uebc6\uebc9 <span class="ff3 fs3 ls17 ws44 v5"> <span class="_47 blank"> </span> <span class="_47 blank"> </span> <span class="_47 blank"> </span> (1) </span></div><div class="t m0 xf h26 y157 ff3 fs3 fc0 sc0 ls4f ws7">where <span class="_1 blank"> </span><span class="ffc ls5">\ue74e</span></div><div class="t m0 x4 h2a y158 ffc fs10 fc0 sc0 ls98">\uebe7<span class="ff3 fs3 ls36 ws20d v7"> is the real interest rate, <span class="ffc ls9b">\ue74e</span></span><span class="ls9c v2">\u2217</span><span class="ff3 fs3 ls58 ws20e v7"> is the equilibrium real interest rat<span class="_1 blank"> </span>e, which is a function of </span></div><div class="t m0 xf h26 y159 ff3 fs3 fc0 sc0 ls1b ws20f">saving-investment factor<span class="_0 blank"></span>s, <span class="ffc ls5">\ue73a</span></div><div class="t m0 x29 h2b y15a ffc fs10 fc0 sc0 ls98">\uebe7<span class="ff3 fs3 ls4f ws210 v7">, with parameters <span class="ffc ls9d">\ue7da<span class="ff3 ls32 ws19e">, and </span><span class="ls5">\ue74e</span></span></span></div><div class="t m0 x2a h29 y15a ffc fs10 fc0 sc0 ls5">\uebe7</div><div class="t m0 x2b h29 y15b ffc fs10 fc0 sc0 lsae ws1ea">\uebc6\uebc9 <span class="ff3 fs3 ls2b ws102 v5"> captures movements in <span class="_0 blank"></span>the real </span></div><div class="t m0 xf h2c y15c ff3 fs3 fc0 sc0 ls1c ws211">rates due to monetary p<span class="_0 blank"></span>olicy. We assume that the <span class="_0 blank"></span><span class="ffc ls9e">\ue74e<span class="fs10 ls97 v1">\u2217</span><span class="ls5 ws1e8 v9">(<span class="v8">\u2219</span>)<span class="ff3 ls58 ws212 v8"> function is approximately linear, so </span></span></span></div><div class="t m0 xf h2d y15d ff3 fs3 fc0 sc0 lsaf ws7">that <span class="ffc ls9f">\ue74e<span class="fs10 ls97 v1">\u2217</span><span class="ls5 ws1e8 v9">(<span class="v8">\ue73a</span></span></span></div><div class="t m0 x2c h27 y15e ffc fs10 fc0 sc0 ls98">\uebe7<span class="fs3 ls99 ws1e9 v7">;\ue003<span class="_4 blank"></span>\ue7da</span></div><div class="t m0 x2d h26 y15f ffc fs3 fc0 sc0 lsa0">)<span class="lsb0 v8">=\ue73a</span></div><div class="t m0 x2e h27 y15e ffc fs10 fc0 sc0 ls98">\uebe7<span class="fs3 lsa1 v7">\ue7da</span><span class="lsa2 v2">\uf1f1</span><span class="ff3 fs3 ls69 ws213 v7">. The more importan<span class="_0 blank"></span>t assumption i<span class="_0 blank"></span>s that monetar<span class="_0 blank"></span>y policy does no<span class="_0 blank"></span>t have </span></div><div class="t m0 xf h26 y160 ff3 fs3 fc0 sc0 ls2a ws214">lasting real effects on the real rate, which can be written formally as <span class="ffc ls5">\ue74e</span></div><div class="t m0 x2f h29 y161 ffc fs10 fc0 sc0 ls5">\uebe7</div><div class="t m0 x30 h2e y162 ffc fs10 fc0 sc0 lsae ws1eb">\uebc6\uebc9 <span class="fs3 ls31 ws1ec v5">~\ue72b (0)<span class="ff3 ls58 ws186">. We relax this </span></span></div><div class="t m0 xf h5 y163 ff3 fs3 fc0 sc0 ls19 ws8d">assumption in Section 3.<span class="_0 blank"></span> </div><div class="t m0 x22 h5 y164 ff3 fs3 fc0 sc0 ls63 ws215">Given that monetary policy does not have last<span class="ls73 ws194">ing effect<span class="_1 blank"> </span>s, Equation (1) implies th<span class="_1 blank"> </span>at any </span></div><div class="t m0 xf h5 y165 ff3 fs3 fc0 sc0 ls4d ws216">low-frequency movements or pe<span class="ls31 ws153">rmanent changes in <span class="ls2f ws217">the real rate reflect solely saving-</span></span></div><div class="t m0 xf h2f y166 ff3 fs3 fc0 sc0 ls2f ws218">investment factors.<span class="ff6 fs4 ls5 ws17d v1">5</span><span class="ls4d wsd6 v0"> For instance, bo<span class="_0 blank"></span>th the real interest <span class="ls35 ws1a5">rate and the saving-investment factors </span></span></div><div class="t m0 xf h5 y167 ff3 fs3 fc0 sc0 ls58 ws219">display dynamics which are statistically hard to<span class="ls4e wse2"> distinguish from a unit-root process over <span class="_0 blank"></span>the </span></div><div class="t m0 xf h5 y168 ff3 fs3 fc0 sc0 ls29 ws21a">full sample. If (1) is true, this should in<span class="_1 blank"> </span> and of itself yield a lot of statistical power to<span class="_1 blank"> </span> estimate </div><div class="t m0 xf h26 y169 ffc fs3 fc0 sc0 ls9d">\ue7da<span class="ff3 ls65 ws21b">. But if (1) is not true, it could also generate <span class="ls2b ws21c">\u201cspuriously\u201d strong corr<span class="ls32 ws21d">elation between <span class="_0 blank"></span>the real </span></span></span></div><div class="t m0 xf h30 y16a ff3 fs3 fc0 sc0 ls35 ws21e">interest rate and the saving-investm<span class="ls4f ws21f">ent factors in specific subsampl<span class="_0 blank"></span>es.<span class="ff6 fs4 ls5 ws17d v1">6</span><span class="ls4d ws220 v0"> To the extent that such </span></span></div><div class="t m0 xf h1b y16b ff6 fsd fc0 sc0 ls5 ws7"> <span class="_1 blank"> </span> <span class="_3 blank"></span> </div><div class="t m0 xf h31 y16c ff6 fs9 fc0 sc0 ls5 wse6">5<span class="ff3 fsa ls3f ws221 v3"> <span class="_e blank"> </span>Note t<span class="_1 blank"> </span>hat (1) is not a reduce<span class="_1 blank"> </span>d form. The reduce<span class="_1 blank"> </span>d form of <span class="ffc ls5">\ue74e</span></span></div><div class="t m0 x27 h32 y16d ffc fs12 fc0 sc0 lsa3">\uebe7<span class="ff3 fsa ls3b ws63 v6">, given (1) and the above a<span class="_1 blank"> </span>ssu<span class="ls77 ws222">mptions, i<span class="_1 blank"> </span>s a stationar<span class="_1 blank"> </span>y </span></span></div><div class="t m0 x11 h33 y16e ff3 fsa fc0 sc0 lsa4 ws223">autoregressive representation ar<span class="_0 blank"></span>ound <span class="ffc ls5 ws1ed">\ue74e<span class="fs12 va">\ue750</span></span></div><div class="t m0 x31 h34 y16f ffc fs12 fc0 sc0 lsa5">\u2217<span class="ff3 fsa ls1e ws224 v5"> with steady-state <span class="ffc ls5">\ue74e</span></span></div><div class="t m0 x32 h35 y170 ffc fs12 fc0 sc0 lsa6">\uebe7<span class="fsa lsb1 v6">=\ue74e</span></div><div class="t m0 x33 h36 y171 ffc fs12 fc0 sc0 lsa7">\u2217<span class="fsa ls5 ws1ed vb">(<span class="v8">\ue73a</span></span></div><div class="t m0 x2b h35 y170 ffc fs12 fc0 sc0 lsa3">\uebe7<span class="fsa lsa8 ws1ee v6">;\ue003<span class="_4 blank"></span>\ue7da</span></div><div class="t m0 x34 h33 y172 ffc fsa fc0 sc0 ls5 ws1ed">)<span class="ff3 lsb2 ws7 v8">. </span></div><div class="t m0 xf h31 y173 ff6 fs9 fc0 sc0 ls5 wse6">6<span class="ff3 fsa ls3b ws225 v3"> <span class="_48 blank"> </span>Under the unit-root assumption, estimates of <span class="ffc lsa9">\ue7da<span class="ff3 lsb3 ws226"> would be s<span class="_1 blank"> </span>uper-consis<span class="_1 blank"> </span>tent, ie they wo<span class="_1 blank"> </span>uld converge a<span class="_1 blank"> </span>t the rate of<span class="_1 blank"> </span> </span></span></span></div><div class="t m0 x11 h33 y174 ffc fsa fc0 sc0 lsaa">\ue736<span class="ff3 ls76 ws227">. However, in thi<span class="_1 blank"> </span>s case, spuriou<span class="_1 blank"> </span>s correlation<span class="_1 blank"> </span> can also a<span class="_1 blank"> </span>rise. Of cour<span class="_1 blank"> </span>se, unit-roots a<span class="_1 blank"> </span>re best see<span class="_1 blank"> </span>n as convenien<span class="_1 blank"> </span>t </span></div><a class="l" data-dest-detail='[10,"XYZ",0,112,-32768]'><div class="d m1" style="border-style:none;position:absolute;left:383.940000px;bottom:166.734000px;width:3.603000px;height:8.625000px;background-color:rgba(255,255,255,0.000001);"></div></a><a class="l" data-dest-detail='[10,"XYZ",0,140,-32768]'><div class="d m1" style="border-style:none;position:absolute;left:162.900000px;bottom:219.894000px;width:3.603000px;height:8.625000px;background-color:rgba(255,255,255,0.000001);"></div></a></div><div class="pi" data-data='{"ctm":[1.000000,0.000000,0.000000,1.000000,0.000000,0.000000]}'></div></div>
Compartilhar