Structured retail products: risk sharing or risk creation?

Otavio Bitu Bruno Giovannetti Bernardo Guimaraes

May 2024
Structured Retail Product: Example

If S&P500 grows by more than x % in 5 years:

the gross return is 1.2 + x %

Otherwise:
the gross return is 1.2
Structured Retail Product: Example

Structure:
Underlying assets: AAPL, FB, NFLX, traded at Nasdaq
4 observation dates (after 6 mo, 1 yr, 18 mo, 2 yrs)

For each nth observation date, if all 3 prices are above initial prices:
the product is terminated and the gross return is 1 + 0.0875n

Otherwise:
the product continues
if it continues until the final date, the gross return is 1.
The question

Structured retail products (SRPs):

Market has persistently increased all over the world
Growing attention from researchers
The question

Structured retail products (SRPs):

Market has persistently increased all over the world
Growing attention from researchers

Crucial question: when a financial institution issues an RSP,

is the institution sharing risk?
or is it generating unbacked-risk?
This paper

Framework to test risk-sharing vs risk-creation

Simple model
Test requires (only) information on expected risk and return
This paper

Framework to test risk-sharing vs risk-creation

Simple model
Test requires (only) information on expected risk and return

Empirical test with a novel dataset

Hand-collected data
The model
Model

2 periods
one bank (B)
measure-one continuum of individuals (I).
Model

2 periods
one bank (B)
measure-one continuum of individuals (I).
Agents consume Ci in the second period only.
Banks and retail investors are risk-averse
Their preferences are given by

Ui = E (Ci ) − γi Var (Ci )

Model

The risk-adjusted excess return of an SRP is

rj = αj + εj
Model

The risk-adjusted excess return of an SRP is

rj = αj + εj

Perfectly informed banks choose αj

Retail investors choose quantity q
Model

Retail investors believe

E (rj ) = αj + ∆
for ∆ ≥ 0, and
e 2 = σ̄ 2 + φ σ 2
σ
with σ̄ 2 ≥ 0 and φ ∈ (0, 1).
Model: results

Suppose the SRP is a hedge for the bank. Then

1 q is positive in the case ∆ = 0
Model: results

Suppose the SRP is a hedge for the bank. Then

1 q is positive in the case ∆ = 0
2 αj might be positive or negative
Model: results

Suppose the SRP is a hedge for the bank. Then

1 q is positive in the case ∆ = 0
2 αj might be positive or negative
3 αj is increasing in σ 2 for a given ∆
Model: results

Suppose the SRP creates risk. Then

1 q = 0 in the case ∆ = 0
Model: results

Suppose the SRP creates risk. Then

1 q = 0 in the case ∆ = 0
2 αj must be negative
Model: results

Suppose the SRP creates risk. Then

1 q = 0 in the case ∆ = 0
2 αj must be negative
3 αj is decreasing in σ 2 for a given ∆
Intuition

If a type of SRP creates risk:

Selling the product raises risk in the bank’s balance sheet
Issuers demand higher prices to sell riskier products they can’t hedge
Naive retail investors accept that
Same reason for why expected return of bets is negative
Intuition

If a type of SRP creates risk:

Selling the product raises risk in the bank’s balance sheet
Issuers demand higher prices to sell riskier products they can’t hedge
Naive retail investors accept that
Same reason for why expected return of bets is negative

If a type of SRP shares risk:

Higher risk akin to lower marginal cost
Empirical Analysis
Data

1847 SRPs
distributed by the largest Brazilian brokerage house
from 2017 to 2022

For each SRP, we

hand-collected its brochure from the brokerage house website
estimated excess return, volatility, volatility, and beta

Novel database with risk and return for many SRPs

Data: estimation of risk and expected return

For each SRP, we

estimate variances and covariances among assets and market index
(correlations are important)
simulate many paths for the underlying assets
compute market return and product’s return in each path
get average (expected) returns, beta and volatilities
Types of SRPs

1 return never negative, higher if an asset price grows

2 same as (1) but return never below a positive number
3 same as (1) but real return never negative
4 return higher if an asset price changes by much
5 (autocallable) return depends on joint paths of 3-4 asset prices
Expected returns

Expected excess return (%, per year)

# products Mean 1 pct 25 pct 50pct 75 pct 99 pct

Type 1 637 -2.15 -8.92 -5.78 -2.51 0.99 6.96

Type 2 500 -1.09 -5.45 -2.55 -1.11 0.09 7.66
Type 3 103 -1.21 -3.33 -2.59 -1.84 0.00 4.16
Type 4 128 -1.05 -5.10 -2.90 -1.59 0.63 3.65
Type 5 479 -2.56 -4.88 -3.12 -2.47 -1.90 -0.76
Risk and return

Type 1 Dependent variable: Expected excess return

(1) (2) (3) (4)

Total volatility 0.924*** 0.965***

(7.92) (8.68)
Market beta 0.062*** 0.064***
(20.98) (30.50)
Residual volatility 0.730*** 0.694***
(8.81) (29.59)
Constant -0.086*** -0.089*** -0.087*** -0.085***
(-12.95) (-11.44) (-17.74) (-54.33)
Bank-year F.E. No Yes No Yes
Obs 637 637 637 637
Adj-R2 72.30% 80.03% 83.41% 91.68%
Risk and return

Type 2 Dependent variable: Expected excess return

(1) (2) (3) (4)

Total volatility 0.660*** 0.631***

(9.16) (8.13)
Market beta 0.020 0.036***
(0.64) (5.02)
Residual volatility 0.605*** 0.534***
(4.68) (4.62)
Constant -0.038*** -0.036*** -0.036*** -0.037***
(-12.33) (-11.61) (-7.32) (-11.89)
Bank-year F.E. No Yes No Yes
Obs 500 500 500 500
Adj-R2 30.76% 76.85% 29.06% 78.19%
Risk and return

Type 3 Dependent variable: Expected excess return

(1) (2) (3) (4)

Total volatility 0.864*** 0.824***

(5.98) (23.24)
Market beta 0.086*** 0.069***
(10.96) (6.39)
Residual volatility 0.596*** 0.684***
(5.95) (23.26)
Constant -0.044*** -0.043*** -0.044*** -0.045***
(-7.49) (-29.59) (-7.45) (-30.47)
Bank-year F.E. No Yes No Yes
Obs 103 103 103 103
Adj-R2 62.5% 87.04% 74.44% 79.85%
Risk and return

Type 4 Dependent variable: Expected excess return

(1) (2) (3) (4)

Total volatility 0.794*** 0.829**

(32.19) (5.70)
Market beta 0.062* 0.035**
(2.78) (5.02)
Residual volatility 0.708*** 1.047***
(12.65) (19.20)
Constant -0.059*** -0.061*** -0.064*** -0.072***
(-12.82) (-6.89) (-9.75) (-8.76)
Bank-year F.E. No Yes No Yes
Obs 128 128 128 128
Adj-R2 59.16% 86.85% 53.05% 87.96%
Risk and return

Type 5 Dependent variable: Expected excess return

(1) (2) (3) (4)

Total volatility -0.256** -0.526***

(-2.58) (-4.74)
Market beta 0.024*** 0.024***
(3.90) (4.29)
Residual volatility -0.354** -0.537***
(-2.54) (-10.19)
Constant -0.014*** -0.001 -0.015** -0.006**
(-3.15) (-0.16) (-3.00) (-2.45)
Bank-year F.E. No Yes No Yes
Obs 479 479 479 479
Adj-R2 7.70% 34.60% 20.56% 41.37%
Risk and return

0.15 SRPs of Types 1, 2, 3, and 4 SRPs of Type 5 (Autocallable)

0.04
Expected exc. ret. orthogonal to beta
Expected exc. ret. orthogonal to beta
0.10

0.02
0.05

0
0

−0.02
−0.05 −0.10

−0.04

−0.05 0 0.05 0.10 .15 0.20 −0.02 0 0.02 0.04

Residual vol. orthogonal to beta Residual vol. orthogonal to beta
Concluding remarks

Other papers have shown SRPs are expensive

We show some SRPs are socially undesirable

Financial markets are there to help risk sharing

Some SRPs create unbacked risk

Policy implication