The Portfolio of Euro Area Fund Investors and ECB Monetary Policy Announcements

Transcription

1 Johannes Bubeck Maurizio Michael Habib Simone Manganelli European Central Bank* The Portfolio of Euro Area Fund Investors and ECB Monetary Policy Announcements IBRN-BdF Conference Global Financial Linkages and Monetary Policy Transmission Banque de France, 30 June 2017 * The views expressed are those of the authors and do not necessarily reflect those of the European Central Bank.

2 Rubric The paper at a glance: what we do Identify ECB monetary policy shocks using intraday data on EA short-term and long-term interest rates around key events Study impact of major ECB monetary policy announcements on a portfolio of Luxembourg-based investment funds, broadly representative of euro-area investors, between 2012 and mid Differentiate among active portfolio reallocation (flows) and passive rebalancing (return or exchange rate effects) on a daily basis In particular, we include proxy of FX effect on portfolio reallocation 2

3 Rubric The paper at a glance: what we find Portfolio balance channel of ECB unconventional policies is muted Monetary policy announcements work through the traditional signaling channel, generating significant valuation effects Passive shift in the portfolio of EA investors into riskier assets, in particular European and Emerging Market equity funds and out of bond funds 3

4 Rubric Motivation Growing literature on the impact of conventional and Unconventional Monetary Policies (UMP) on asset prices (Borio and Zabai, 2016) Existing evidence focusing on (i) US monetary policies and US investors and (ii) impact on asset prices and macroeconomic variables More limited evidence on: o EA policies and EA investors o impact of UMP on portfolio flows and portfolio allocation Our contribution Channels of transmission of ECB UMP (signaling vs portfolio rebalancing): which one was at work? Disentangle active and passive fx and return effect on a daily basis 4

5 Rubric Main channels of (unconventional) monetary policy Signalling channel (through changes in monetary policy stance): affect expectations about future rates, the risk neutral components of interest rates (Bauer, 2014), and provide new information to investors affect the forward rates and bond prices do not affect arbitrageurs positions in equilibrium and risk premia Portfolio rebalancing (through purchases of long-term T-bonds): investors will be forced to hold smaller positions in long term bonds and bear less duration risk (preferred habitat, Vayanos, 2009 and Greenwood, 2014) decrease in risk premia and an increase in bond prices portfolio reallocation Signaling Portfolio balance Active reallocation X Passive reallocation (through price changes) 5

6 Rubric Investment funds and monetary policy Monetary policy easing associated with a shift towards riskier assets, not necessarily domestic equity, more likely foreign securities US monetary policy easing : - rotation of portfolio towards AE equity by asset managers (Cenedese et al.2015) or by underlying investors (Curcuru et al. 2015), but not necessarily to US equity (Kroencke et al. 2015) - rebalancing of portfolio towards non-us equity (Cenedese et al. 2015, Fratzscher et al and Kroencke et al. 2015) Unconventional ECB monetary policies until 2012: - larger flows to EM bond and equity funds and AE bond funds (Fratzscher et al. 2016) ECB APP: - no increase in portfolios flows to EM (Georgiadis and Gräb, 2016) 6

7 Rubric The construction of the portfolio Portfolio of funds classified according to their broad mandate: Equity and Bond: Western Europe (WE), Other developed economies (USAJP), Global (GLOB), Emerging Markets (EM) Focus on underlying investors and flows to mutual funds over the short-run (not on asset allocation by fund managers) Table 2: EPFR Portfolio, average total TNA EUR 554bn, Jan Jun 2016 averages Equity Bonds WE USAPJ GLOB EM WE USAPJ GLOB EM EPFR Portfolio TNA (EUR bn) Share (% of total TNA) 16% 12% 12% 21% 11% 3% 17% 9% Currency denomination of funds (% of total TNA) EUR 92% 17% 38% 14% 92% 12% 43% 23% USD 3% 67% 60% 83% 0% 83% 46% 73% Other 5% 17% 2% 2% 8% 5% 11% 4% Country allocation of funds (% of total TNA) EA 62% 2% 13% 1% 83% 5% 29% 0% US 0% 67% 46% 0% 1% 73% 34% 0% Other 38% 32% 42% 99% 16% 22% 37% 100% 7

8 Rubric EPFR data: investment funds Daily investment fund data from 01/01/2012 to 30/06/2016 on total assets (A), returns (r) and flows: Consistent in the original currency denomination A(t) = A(t-1) + A(t-1) * r(t) + flows(t) + fx(t) Consistent in USD terms A and r(t) are obtained in the currency share class (denomination) of the fund, but eventually stock/flow data are reported in USD EPFR includes an additional term (fx) to account for the impact of the USD exchange rate on stocks and flows All measures converted to euro to obtain an fx effect in euro Universe fixed as of 01/01/2012, no entry bias, but exit bias, which we control with internal consistency checks on stocks 8

9 Rubric Dependent variables Total change in portfolio allocation Active reallocation Passive rebalancing Flows (injections/redemptions) Return effect Valuation fx effect (through currency denomination of funds) 9

10 Rubric Decomposition of changes in portfolio weights Change in total net assets of fund i (A i,t ) driven by its total return R i,t =(1+r i,t ) on assets in the previous period, by flows (f i,t ) and the exchange rate adjustment (fx i,t ) Decomposition (extending Ahmed et al. 2016) 10

11 Rubric Passive reallocation due to return effects New weight of asset class i due only to gross returns (R) Initial weight of asset class i where: 11

12 Rubric Passive reallocation due to return effects Same initial balanced (50/50) equity/bond portfolio at time t-1: A E,t-1 =50 ; A B,t-1 =50 ; ΣA i,t-1 =100 but now equity returns (+5%) over-perform bond returns (+1%) and flows (and fx effects) are equal to 0 Initial allocation E B Total A i,t A i,t-1 / ΣA i,t-1 50% 50% 100% Flows f i,t Returns R i,t = 1+r i,t Final allocation A i,t A i,t / ΣA i,t 51% 49% 100% Passive reallocation Δw R i,t 1% -1% 12

13 Rubric Passive reallocation due to exchange rate effects New weight of asset class i due only to exchange rate valuation effects(fx) Initial weight of asset class i where: 13

14 Rubric Active reallocation due to flows Final weight of asset class i Passive fx effect Passive return effect Initial weight of asset class i 14

15 Rubric Active reallocation due to flows: a simple example Assume an initial balanced (50/50) equity/bond portfolio (in $): A E,t-1 =50 ; A B,t-1 =50 ; so that total portfolio ΣA i,t-1 =100 which attracts 2.5$ to equity funds, but only 0.5$ to bond funds, while returns (and fx effects) are equal to 0 Initial allocation E B Total A i,t A i,t-1 / ΣA i,t-1 50% 50% 100% Flows f i,t Returns R i,t = 1+r i,t Final allocation A i,t A i,t / ΣA i,t 51% 49% 100% Active reallocation Δw f i,t 1% -1% 15

16 Rubric Monetary policy surprises Monetary Policy Surprises identified through changes in: average short-term (1W, 1M, 3M) EA rates average DE-ES-IT long-term (10y) government bond yields Time window around ECB press conferences (14:30 Frankfurt time +/- 60 minutes) and other important announcements (based on Rogers et al and Altavilla et al. 2015) Sign inverted: positive surprise corresponds to lower yields Advantages: clean identification, approach similar to several other studies Disadvantage: we capture only short-term impact 16

17 Rubric Monetary policy surprises Change in EA short-term rates Change in EA 10-year yields 17

18 Rubric The Impact of Monetary Policy on EA Fund Investors Y i,t = Σ θ i,j (MP surprise) t-j + Σ β i,k (Y) i,t-k + Σ Γ i,j (Controls) i,t-j + ε i,t Dependent variables: Reallocation measures (Y) i Reallocation across i=1,..,8 fund categories (e.g. WE Bonds, ) Variables: Total reallocation, active reallocation, passive reallocation (joint estimation) In addition: flows (% of TNA) and benchmark returns Variable of interest: (MP surprise) baseline: impact up to 4 lags (test of Σθ i,j = 0) control also contemporaneous impact (test of θ i,0 = 0) Other controls: Other macroeconomic news (Citi Economic surprise index) Lags of the dependent variable 18

19 Rubric Results Impact of surprises to EA short-term (ST) rates (large shock by 2 St. Dev. ~ 3.5 bps) Full sample: beginning of 2012 until end-june 2016 Before APP until August 2014 APP since September 2014 Impact of surprises to EA long-term (10Y) yields (large shock by 2 St. Dev. ~ 10 bps) Full sample: beginning of 2012 until end-june 2016 Before APP until August 2014 APP since September

20 Rubric Impact of surprise change in EA ST rates: /2016 Equity Bonds WE USAPJ GLOB EM WE USAPJ GLOB EM Joint estimation with restriction on the sum of the fitted values Active reallocation * ** Passive reallocation return * * Passive reallocation FX ** 1.221** *** * *** Total reallocation * Separate estimation (equation by equation) Flows (% of TNA) Surprises Observations Benchmarks: Equity Bonds EUR/USD WE USA GLOB EM WE USA GLOB EM -31.0*** *** 33.3*** 34.6*** 33.0*** Note: ***, ** and * indicate statistical significance at the 1, 5 or 10 percentage level, respectively, of an F-Test of the impact of the monetary policy surprise up to 4 lags Reallocation out of non-euro area equity funds but result not particularly robust to different lag structure (0 or 2 lags) or across different time-periods Depreciation of the EUR vs USD (around 1% for large shock) and FX passive reallocation out of European (EUR denominated) funds 20

21 Rubric Impact of surprise change in EA ST rates: ECB 1 st phase Equity Bonds WE USAPJ GLOB EM WE USAPJ GLOB EM Joint estimation with restriction on the sum of the fitted values Active reallocation * Passive reallocation return *** *** * 1.313*** 0.497*** 1.932** 0.862** Passive reallocation FX ** 1.070** *** * *** Total reallocation ** * ** *** 2.051*** Separate estimation (equation by equation) Flows (% of TNA) * 1.801* Surprises Observations Benchmarks: Equity Bonds EUR/USD WE USA GLOB EM WE USA GLOB EM -19.1** *** 24.8** 24.4** 22.1** Note: ***, ** and * indicate statistical significance at the 1, 5 or 10 percentage level, respectively, of an F-Test of the impact of the monetary policy surprise up to 4 lags Reallocation into non-euro area developed bond markets: small impact on flows Positive impact on bond prices and passive reallocation into bond funds 21

22 Rubric Impact of surprise change in EA ST rates: ECB APP Equity Bonds WE USAPJ GLOB EM WE USAPJ GLOB EM Joint estimation with restriction on the sum of the fitted values Active reallocation Passive reallocation return 8.729*** ** ** ** Passive reallocation FX ** 2.598** ** ** Total reallocation 5.795** ** Separate estimation (equation by equation) Flows (% of TNA) Surprises Observations Benchmarks: Equity Bonds EUR/USD WE USA GLOB EM WE USA GLOB EM -67.3*** 100.8*** 65.6** 74.1*** 73.6* 15.2** 62.2** 68.2*** 68.0*** Note: ***, ** and * indicate statistical significance at the 1, 5 or 10 percentage level, respectively, of an F-Test of the impact of the monetary policy surprise up to 4 lags Sharp reaction of WE equity prices large positive passive return effect for WE equity funds EUR depreciates by 2.3% reallocation out of WE equity funds (-0.14% to large shock or almost 3 daily SD) Total reallocation: WE equity vs. WE bond 22

23 Rubric Impact of surprise change in EA 10Y yields: /2016 Equity Bonds WE USAPJ GLOB EM WE USAPJ GLOB EM Joint estimation with restriction on the sum of the fitted values Active reallocation * Passive reallocation return 1.390** * * * ** Passive reallocation FX *** 0.598** 0.163* 1.051*** *** Total reallocation ** 1.278** ** * Separate estimation (equation by equation) Flows (% of TNA) ** Surprises Observations Benchmarks: Equity Bonds EUR/USD WE USA GLOB EM WE USA GLOB EM -12.0** 17.7** 15.7** 15.2** 18.4*** 3.9*** 13.5** 13.9** 13.4** Note: ***, ** and * indicate statistical significance at the 1, 5 or 10 percentage level, respectively, of an F-Test of the impact of the monetary policy surprise up to 4 lags Active reallocation into EM equity: modest impact and not robust Large asset price and FX effects from the APP period (see next slides) 23

24 Rubric Impact of surprise change in EA 10Y yields: ECB 1 st phase Equity Bonds WE USAPJ GLOB EM WE USAPJ GLOB EM Joint estimation with restriction on the sum of the fitted values Active reallocation ** ** Passive reallocation return Passive reallocation FX * ** * Total reallocation Separate estimation (equation by equation) Flows (% of TNA) *** Surprises Observations Benchmarks: Equity Bonds EUR/USD WE USA GLOB EM WE USA GLOB EM Note: ***, ** and * indicate statistical significance at the 1, 5 or 10 percentage level, respectively, of an F-Test of the impact of the monetary policy surprise up to 4 lags Apart from active reallocation into EM, not much action from surprises to EA long-term yields in the first part of the sample 24

25 Rubric Impact of surprise change in EA 10Y yields: ECB APP Equity Bonds WE USAPJ GLOB EM WE USAPJ GLOB EM Joint estimation with restriction on the sum of the fitted values Active reallocation Passive reallocation return 3.377*** ** *** *** *** *** Passive reallocation FX *** 1.006*** *** * 0.170*** *** Total reallocation 1.500** ** *** * *** ** Separate estimation (equation by equation) Flows (% of TNA) * 2.238** Surprises Observations Benchmarks: Equity Bonds EUR/USD WE USA GLOB EM WE USA GLOB EM -25.9*** 40.6*** 29.5*** 33.6*** 38.4*** 7.8*** 28.3*** 29.8*** 28.9*** Note: ***, ** and * indicate statistical significance at the 1, 5 or 10 percentage level, respectively, of an F-Test of the impact of the monetary policy surprise up to 4 lags Surprises to long-term yields good proxy of balance sheet policies at the zerolower bound Reallocation out of bond funds and into equity funds (WE and EM) as a result of significant passive FX and returns effects 25

26 Rubric Robustness Distinguish normal surprises from tail events (observation below the 10 th and above the 90 th percentile of distribution of surprises) Major announcements drive the results Institutional vs. retail fund shares Different behaviour: institutional investors flow into EM equity Local Projections (Jorda, 2005) to study adjustment of total portfolio weights over longer-horizon of two-week Sample starting from 01/01/2014 Distinguish between positive (lower yields or rates) and negative (higher yields or rates) surprises No major differences for these robustness tests 26

27 Rubric Concluding remarks 27

28 Rubric Concluding remarks Portfolio balance channel of ECB unconventional monetary policies is generally muted ECB monetary policy worked through signaling channel: generating significant valuation effects (passive rebalancing) Passive shift in the portfolio of EA investors into riskier assets, in particular European and Emerging Market equity funds and out of bond funds Robust evidence that (unsophisticated) investors are affected by monetary policy mainly through the impact it has on asset prices by changing expectations of future interest rates Consistent with rational inattention and infrequent portfolio changes 28

29 Rubric Background slides 29

30 Rubric Literature large impact of ECB policies on markets The 1 st phase of ECB policies mainly confidence channel positive impact on stock markets, compression in EA coreperiphery spreads and euro up (Rogers et al. 2014); positive international spillovers on equity markets (Fratzscher et al. 2016) The Asset Purchasing Programme (APP) Persistently reduced sovereign bond-yields (Andrade et al. 2015); with spill-over to non-targeted assets such as corporates (Altavilla et al. 2015); again positive international equity spillovers but euro down (Georgiadis and Gräb, 2016) 30

31 Rubric Focus on Luxembourg We focus on Luxembourg based mutual funds to track the portfolio of EA investors: why Luxembourg? - Largest asset manager in the euro area: EUR 3.7 trillion AuM out of EUR 10.6 trillion in the EA (around 1/3) - Broadly representative of an average EA investor: around 75% of cross-border equity investment in Luxembourg originates from EA (IMF CPIS derived liabilities) - Good coverage in EPFR compared to other EA countries (27% of AuM as reported by ECB), higher for equity (36%), than for bond (19%) funds 31

32 Rubric Joint estimation of reallocation coefficients We exploit the following: portfolio weights sum to 1 across fund categories (while reallocation measures sum to 0) We impose an additional restriction on our fitted values and estimate the model jointly to increase the precision of our estimates Step 1: We stack the estimation equation and the constraint Step 2: We estimate the transformed equation 32

33 Rubric Retail vs. Institutional: ECB 1 st phase (10-year rates) Equity Bonds WE USAPJ GLOB EM WE USAPJ GLOB EM All Funds Active reallocation ** ** Flows (% of TNA) *** Retail fund shares Active reallocation ** Flows (% of TNA) *** Institutional fund shares Active reallocation ** Flows (% of TNA) ** Additional Information Surprises Observations Note: ***, ** and * indicate statistical significance at the 1, 5 or 10 percentage level, respectively, of an F-Test of the impact of the monetary policy surprise up to 4 lags tbd 33

34 Rubric Retail vs. Institutional: ECB APP (10-year rates) Equity Bonds WE USAPJ GLOB EM WE USAPJ GLOB EM All Funds Active reallocation Flows (% of TNA) * 2.238** Retail fund shares Active reallocation Flows (% of TNA) * Institutional fund shares Active reallocation ** Flows (% of TNA) ** 3.029** Additional Information Surprises Observations Note: ***, ** and * indicate statistical significance at the 1, 5 or 10 percentage level, respectively, of an F-Test of the impact of the monetary policy surprise up to 4 lags tbd 34

35 Rubric Additional dependent variables: benchmark returns Additional control: inclusion of benchmark indices and the USD/EUR as dependent variables 35