1 / 34

Apoorva Javadekar - How Does Mutual Fund Reputation Affect Subsequent Fund Flows?

Apoorva Javadekar - Its about Studies link between fund performance and fund flows (flow-schedule) <br>Finds and rationalizes evidence of return chasing and convexity in fund flows <br>But not much is known about the importance of performance history (reputation) <br>

Download Presentation

Apoorva Javadekar - How Does Mutual Fund Reputation Affect Subsequent Fund Flows?

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. How Does Mutual Fund Reputation Affect Subsequent FundFlows? ApoorvaJavadekar BostonUniversity February 8,2016 Apoorva Javadekar (BostonUniversity) February 8,2016

  2. Introduction Motivation I: Why Study MutualFunds? Mutual Funds: Important Vehicle ofInvestment Manage 15 Tr$ Mutual funds owns 30% US equities Vs 20% direct holdings 46% of US household own mutual funds 1 Understand BehavioralPatterns: Investors learn about managerial ability through returns ⇒ fund flows shed light on learning, information processing capacities etc. 2 Fund Flows Affect Managerial RiskTaking 90% funds managers paid as a % ofassets ⇒ flow patterns can affect risk taking ⇒ impacts on asset prices 3 Apoorva Javadekar (BostonUniversity) February 8,2016

  3. Introduction Motivation II: ThePaper ExistingLiterature: Studies link between fund performance and fund flows(flow-schedule) Finds and rationalizes evidence of return chasing and convexity in fund flows But not much is known about the importance of performance history (reputation) 1 ThisPaper: Explore the role of reputation for fund flows How history up to t − 1 affect link between time t performance and time t + 1flows Can we explain theevidence? 2 Apoorva Javadekar (BostonUniversity) February 8,2016

  4. Introduction Role ofReputation Better understanding of managerial incentives: High reputation ⇒ Low P(Getting Fired) (Khorana; 1996, Kostovetsky; 2011) My sample: 30% of the fired managers belong to bottom 20% reputationrank But compensation too determine the incentives and flows affect compensation ⇒ Important to know how reputation affectflows But can reputation affect flows? Investor Heterogeneity ⇒ investor composition is history-specific ⇒ subsequent reactions to fund performance become history-specific Apoorva Javadekar (BostonUniversity) February 8,2016

  5. Introduction Agenda Empirical Evidence Model Testing model predictions in data Tests to check validity of modelmechanism Apoorva Javadekar (BostonUniversity) February 8,2016

  6. Introduction LiteratureReview Return Chasing and flowconvexity: Ippolito (1992), Sirri & Tuffano (1998), Chevallier & Ellison (1997) Lack PerformancePersistence: Carhart (1997), Bollen & Busse (2004) test short and medium term persistence Risk shifting due to convexflows: Brown, Harlow, Starks (1996), Basak (2012) TheoreticalModels: Berk & Green (2004): rationalizes lack of persistence and return chasing simultaneously using decreasing returns and competitive capital supply Lynch & Musto (2003): explains convexity using manager replacement Berk & Tonks (2007): repeat losers have insensitive flows to the left offlow-schedule Apoorva Javadekar (BostonUniversity) February 8,2016

  7. EmpiricalEvidence Variables FundFlows: qit−[qit−1×(1+rit)] FLOWit= q (1 +r ) it−1× it where rit denotes net of expense fund returns during time t and qit denotes fund assets at the end of time t. FundPerformance: Ranks within same ’investment objective’ based on raw net returns (Sirri & Tuffano;1998) Ranks based upon ’CAPM-Alpha’ (Berk & Binsbergen; 2014) Ranks are normalized to lie between [0, 1]interval. Current Performance (Perfit ): Based upon current yeart Reputation (reputeit ): Based upon 5 year window ending with current year t. Apoorva Javadekar (BostonUniversity) February 8,2016

  8. EmpiricalEvidence SummaryStatistics Apoorva Javadekar (BostonUniversity) February 8,2016

  9. EmpiricalEvidence Basic RegressionFramework Objective: Asses impact of reputation starting at time t on flow-schedule for the period t +1 Regression: 5 5 FLOWit+1 =a+φjQjit+ψj(Qjit×reputeit−1) j=2j=2 +(γ×reputeit−1)+controlsit+εit+1 Qjit denotes dummy for j th quantile ofPerfit Regression of t + 1 flows on time t recent performance given reputation starting at timet Regression for each quantile of Perfit to account for non-linearity (Chevallier & Ellison;1997) Apoorva Javadekar (BostonUniversity) February 8,2016

  10. EmpiricalEvidence Regression Output: FundFlows Panel A: Raw Returns Q2t −Q1t Q3t −Q1t Q4t −Q1t Q5t −Q1t reputet−1 reputet−1×(Q2t−Q1t) reputet−1×(Q3t−Q1t) reputet−1×(Q4t−Q1t) reputet−1×(Q5t−Q1t) 0.013 (0.011) 0.032*** (0.012) 0.050*** (0.014) 0.107*** (0.018) 0.083*** (0.015) 0.043** (0.019) 0.108*** (0.021) 0.149*** (0.026) 0.261*** (0.033) 0.034*** (0.006) 0.084*** (0.007) 0.124*** (0.007) 0.241*** (0.010) 0.037*** (0.006) 0.090*** (0.007) 0.130*** (0.007) 0.246*** (0.010) 0.202*** (0.013) Intercept 0.056 (0.037) -0.008 (0.035) -0.067* (0.035) AdjR2 0.176 0.208 0.215 Results Apoorva Javadekar (BostonUniversity) February 8,2016 10 /34

  11. EmpiricalEvidence MainResults RegressionTable Result 1: Significant return chasing effect ignoring reputation interactions and even after controlling for reputation Result 2: Return chasing effect is reduced by more than half after including reputationinteractions Result 3: All the interaction terms are large andsignificant Significant =⇒ (Qj − Q1|repute = high) > (Qj − Q1|repute = low). Large =⇒ Interaction effect more important than return chasing effect Result 4: Coefficients on Interaction term rise monotonically with performance ⇒ Flow-Schedule more sensitive for higher reputed funds Flow-schedule sensitive even at the lower end for high reputation fund. Apoorva Javadekar (BostonUniversity) February 8,2016

  12. EmpiricalEvidence Flow-ScheduleGraph Apoorva Javadekar (BostonUniversity) February 8,2016

  13. EmpiricalEvidence Example Best Fund: Q5t = 1 and reputet−1 =0.90 Worst Fund: Q1t = 1 and reputet−1 =0.10 ∆FLOW≡FLOW(Best)−FLOW(Worst)=40.8% Break-Up: Apoorva Javadekar (BostonUniversity) February 8,2016

  14. EmpiricalEvidence RobustnessChecks Change in Market Share as dependent variable (Spiegel &Zhang; 2012) Result Results valid across age and size categories Result Results valid even if recent performance is computed over alonger horizon Result Apoorva Javadekar (BostonUniversity) February 8,2016

  15. EmpiricalEvidence Model Apoorva Javadekar (BostonUniversity) February 8,2016

  16. Model Set-Up Manager with unknown skill α and generates gross returnas Rt=α+εt with . . 2 ε ε ∼ N 0,σ t Convex cost of active management: C (x ) = ηx2 Net ReturnProcess: 1)2. .(ht 1 ×qt − − rt = ht−1Rt − f −η qt−1 where ht−1 denotes actively managed share of assets during timet Apoorva Javadekar (BostonUniversity) February 8,2016

  17. Model Investors andBeliefs Investors: Unit mass of risk neutralinvestors µ fraction of Always Attentive (AA) 1 − µ fraction Occasionally Attentive (OA) Each period, P(attention|OA) = δ < 1 Have infinitely deep pockets Beliefs About Managerial Skill: At the end of timet 2 α ∼ N(φ , σ) t t ⇒ Et(α)=Et(Rt+1)=φt Apoorva Javadekar (BostonUniversity) February 8,2016

  18. Model MechanismI Equilibrium Condition When δ = 1: (Berk & Green; 2004) Et(rt+1|ht,φt)=0 Deep pockets ensure that fund receive required inflows Full attention ensures that no investor invests in negative NPV manager. Equilibrium Condition When δ <1: Et(rt+1|ht,φt)≤0 Deep pockets ensure that no positive expected NPV project exists Inattentive investors ⇒ capital outflows could be less than required to attain zero NPVcondition Inattention =⇒ Over-Sized funds relative to competitive benchmark. Apoorva Javadekar (BostonUniversity) February 8,2016

  19. Model MechanismII In equilibrium: Low reputation funds predominantly owned by OA-types Because AA-types are fast to move out of poor performing funds Implications ForFlows: Dampened outflows after yet another bad performance by low reputationfunds Over-Sized ⇒ Low required inflows after a good performance Implications forPersistence: Over-Sized ⇒ Low reputation funds must under-perform Apoorva Javadekar (BostonUniversity) February 8,2016

  20. Model Solution With δ <1 Initial Investor Composition: A investor’s ownership at t = 0is µ λ0= µ+(1−µ)δ s F ¸¸ Attentive raction In xonomy Ec Competitive Size and Flows: qt∗satisfy Et[rt+1|ht,q∗]=0 t and requiredflows e∗∗ t = qt − qt−1(1 +rt ) AttentiveCapital:    zt =  +(1 −λ )δ q (1 +r ) λt−1 t−1 t−1 t As ¸¸ Within x ttentiveF raction Fund Apoorva Javadekar (BostonUniversity) February 8,2016 20 /34

  21. Model InvestorComposition Outflows ⇒ λt <λt−1 If fund has enough attentivecapital: λt−1 AA’sContributionToOutflows=>λ λt−1+(1−λt−1)δ t−1 If zt < |e∗| ⇒ λt = 0 as every attentive investorliquidates t Inflows⇒ λt >λt−1 AA-type contribute λ0 of new capital and outflows reduce λ ⇒ λ0 is upper limit ofλt−1 λt is a weighted average of λ0 and λt−1 ⇒ λt ∈ (λt−1,λ0) Persistent outflows ⇒ High fraction of Inattentive Investors Apoorva Javadekar (BostonUniversity) February 8,2016

  22. Model Learning and FundFlows BeliefUpdates: . . 2 .rt−Et−1(rt). t−1 s =ω¸¸x σt−1 φt = φt−1+ σ2 h 2 t−1 +σε t−1 ⇒ ∆φt bigger for over-sized funds as Et−1(rt ) <0 Fund Flows: Let qt−1=q∗ ×(1+ψt−1) t−1 If capital adjustment is complete 2 .1+ωt 1.rt +ψt−1.. − 2f2 FFt= −1 (1 +ψ )(1 +r ) t−1 t In case zt is not enough to support outflows zt FF =− t q(1 +r) t t+1 Apoorva Javadekar (BostonUniversity) February 8,2016

  23. Model Fund FlowsContinued Limited Outflows: Low reputed funds ⇒ low λt−1 flow-schedule on the left tail DampenedInflows: =⇒Flat Over-Size Effect: Low reputed fund ⇒ ψt−1 > 0 =⇒ required inflows e∗ = q∗ − qt−1(1 + rt ) are smaller compared tocompetitively t t sizedfund Learning Effect: Et−1(rt ) < 0 ⇒ q∗ itself is pushed up for a givenrt t ⇒ e∗ is higher for a given rt t For reasonable parameter values, Over-Size effect dominates Learning effect Apoorva Javadekar (BostonUniversity) February 8,2016

  24. Model Flows With Various ParameterValues Apoorva Javadekar (BostonUniversity) February 8,2016

  25. Model Performance Persistence Apoorva Javadekar (BostonUniversity) February 8,2016

  26. Model CalibrationExercise Parameter Value Source Size Distortion ψt: 2 ∗ E (r)=−ηhqψ=−f ×ψt t t+1 t s¸¸x t t ¸¸ x s 1.76% −1.64% Apoorva Javadekar (BostonUniversity) February 8,2016

  27. Model Experiments To Validate ModelMechanism Heterogeneity in Investors ⇒ Heterogeneity in Flows What events damp thisheterogeneity? Managerial Replacement:⇒ media news, and other soft information ⇒ higher investor attention even from otherwise inattentive investors ⇒ dampened investor heterogeneity Large Front Loads Large front loads ⇒ potentially more attention by investors In both these cases, interaction between reputation and recent performance must lose itsimportance. Replacementfrontloads Apoorva Javadekar (BostonUniversity) February 8,2016

  28. Model ConcludingRemarks Return chasing gets stronger with reputation Persistence in poor performance for low-reputationfunds Simple model with inattentive investors explains the heterogeneity in flow-schedule Interesting to study risk shifting conditional on reputation Apoorva Javadekar (BostonUniversity) February 8,2016

  29. Model Thank You! Apoorva Javadekar (BostonUniversity) February 8,2016

  30. Model Regression With Change in MarketShare Panel A: Raw Returns Panel B:CAPM-Alpha Q2t −Q1t Q3t −Q1t Q4t −Q1t Q5t −Q1t reputet−1 reputet−1×(Q2t−Q1t) reputet−1×(Q3t−Q1t) reputet−1×(Q4t−Q1t) reputet−1×(Q5t−Q1t) 0.042 (0.026) 0.107*** (0.032) 0.258*** (0.033) 0.510*** (0.046) -0.125*** (0.046) -0.186** (0.079) -0.158*** (0.051) -0.167** (0.070) -0.048 (0.060) 0.326*** (0.098) 0.577*** (0.169) 0.811*** (0.124) 1.309*** (0.186) 0.061** (0.026) 0.131*** (0.036) 0.276*** (0.035) 0.490*** (0.047) -0.085* (0.044) -0.130* (0.071) -0.110** (0.053) -0.149** (0.069) -0.023 (0.066) 0.297*** (0.088) 0.517*** (0.166) 0.753*** (0.121) 1.195*** (0.186) Intercept -0.189 (0.240) -0.217 (0.223) -0.220 (0.231) -0.305 (0.221) AdjR2 0.062 0.088 0.055 0.077 Back toRobustness Apoorva Javadekar (BostonUniversity) February 8,2016

  31. Model Age And Size Robustness With RawReturns Panel A: Age Bins Panel B: Size Bins Young=1 0.004 (0.024) 0.029 (0.026) 0.057* (0.030) 0.157*** (0.039) 0.075*** (0.028) 0.042 (0.038) 0.126*** (0.042) 0.177*** (0.052) 0.268*** (0.066) Young=0 0.011 (0.012) 0.035*** (0.013) 0.046*** (0.014) 0.087*** (0.019) 0.095*** (0.018) 0.048** (0.021) 0.089*** (0.024) 0.125*** (0.026) 0.237*** (0.036) Small=1 -0.001 (0.015) 0.016 (0.017) 0.041* (0.023) 0.116*** (0.028) 0.056** (0.024) 0.058* (0.031) 0.164*** (0.036) 0.214*** (0.053) 0.323*** (0.057) Small=0 0.034** (0.016) 0.045*** (0.017) 0.048*** (0.017) 0.086*** (0.021) 0.093*** (0.020) 0.014 (0.026) 0.070** (0.028) 0.122*** (0.028) 0.246*** (0.037) Q2t −Q1t Q3t −Q1t Q4t −Q1t Q5t −Q1t reputet−1 reputet−1×(Q2t−Q1t) reputet−1×(Q3t−Q1t) reputet−1×(Q4t−Q1t) reputet−1×(Q5t−Q1t) Intercept 0.096 (0.122) 0.024 (0.057) -0.044 (0.041) -0.076** (0.039) AdjR2 0.209 0.234 0.181 0.268 Back toRobustness Apoorva Javadekar (BostonUniversity) February 8,2016

  32. Model Longer Horizon For RecentPerformance Panel A: Raw Returns Panel B:CAPM-Alpha Q2t −Q1t Q3t −Q1t Q4t −Q1t Q5t −Q1t reputet−2 reputet−2×(Q2t−Q1t) reputet−2×(Q3t−Q1t) reputet−2×(Q4t−Q1t) reputet−2×(Q5t−Q1t) 0.008 (0.008) 0.042*** (0.009) 0.074*** (0.009) 0.177*** (0.013) 0.158*** (0.014) 0.005 (0.015) 0.021 (0.016) 0.024 (0.018) 0.048* (0.028) 0.066*** (0.022) 0.022 (0.029) 0.063** (0.030) 0.117*** (0.031) 0.230*** (0.043) 0.001 (0.014) 0.017 (0.017) 0.035* (0.018) 0.034 (0.028) 0.040* (0.022) 0.079*** (0.027) 0.076** (0.030) 0.144*** (0.031) 0.257*** (0.044) 0.019** (0.008) 0.060*** (0.009) 0.101*** (0.009) 0.217*** (0.013) 0.039*** (0.008) 0.058*** (0.008) 0.123*** (0.010) 0.212*** (0.013) 0.029*** (0.008) 0.041*** (0.008) 0.097*** (0.010) 0.173*** (0.013) 0.156*** (0.014) Intercept 0.035 (0.036) -0.035 (0.036) -0.005 (0.036) 0.002 (0.036) -0.037 (0.037) -0.074** (0.036) Adj.R2 0.329 0.343 0.347 0.326 0.339 0.344 Back toRobustness Apoorva Javadekar (BostonUniversity) February 8,2016

  33. Model Regression With ManagerialReplacement Panel A: RawReturns Panel B:CAPM-α Back toExperiments Apoorva Javadekar (BostonUniversity) February 8,2016

  34. Model Regressions Across FeeStructures Panel A: RawReturns Panel B:CAPM-Alpha Back toExperiments Apoorva Javadekar (BostonUniversity) February 8,2016

More Related