Dynamic formation of investment strategies for DC pension plan participants: two new approaches Vadim Prudnikov USATU, Ufa, Russia Radon Workshop on FDynamic formation of investment strategies for DC pension plan participants: two new approaches Vadim Prudnikov USATU, Ufa, Russia Radon Workshop on Financial and Actuarial Mathematics for Young Researchers Linz, Austria, May 30 - 31, 2007
Dynamic formation of investment strategies for DC pension plan participants: two new approaches Vadim Prudnikov USATU, Ufa, Russia Radon Workshop on F
Dynamic formation of investment strategies for DC pension plan participants: two new approaches Vadim Prudnikov USATU, Ufa, Russia Radon Workshop on Financial and Actuarial Mathematics for Young Researchers Linz, Austria, May 30 - 31, 2007
1
Ufa, Russia
2
Ufa State Aviation Technical University
3
The investment in a pension fund
The investment – is very important field of any pension fund for DB plan participants (the actuarial rate) for DC plan participants (the pension value, attractiveness for new clients) The only relevant characteristic for DC plan participants is the length of the investment horizon
4
Overview of the problem
Funds propose to their participants a various number of portfolios Two opposite opinions occur regarding the question whether or not investment risk depends on the length of investment horizon
5
Procedure 1
Purpose: reproduce the distribution function of the daily logreturns of assets Process of the daily logreturns is defined by the following model: (1)
6
Procedure 2
Purpose: reproduce the autocorrelation function of the process of the daily logreturns According to the signal processing theory: to get the power spectrum of the signal to go to the gain-frequency characteristic of the signal to estimate the parameters of the filter given the gain-frequency characteristic, the order of the filter and the method of estimation to pass ‘white’ noise through this filter
7
First approach, optimization problem
Consider a participant who has intervals up to retirement, each interval has days long Criterion: (2) Constrains: (3) (4) (5) - is the minimally allowed proportion of participants’ portfolio invested in each asset - is the structure of "equal-weighted" portfolio
8
Estimating of ,
3 methods for efficiencies calculation, using : by direct statistical approach (6) by the simulation of the financial market trajectories (using procedure 1 or procedure 2) (7) - is the value of the quote of asset "i" if the market will go by trajectory "l" (8) (9)
9
Function A(t), features
- is the number of intervals in the investment horizon of the most young participants 2 opinions: - investment risk of a participant is independent of the length of his investment horizon: (10) - investment risk of preretired participants must be fewer than it is for young participants - is a droningly increased function, and (11) (12)
10
let to be a minimally allowed step in the change of the portfolio value invested in any particular asset to realize the second approach we need: to form the countable set of the investment strategies to simulate the matrix which contain trajectories of financial market probably motion, for days each let S(j,z) be the participants’ portfolio value at the moment of retirement while he applies investment strategy x(j) and financial market moves over trajectory z (19) (20)
12
Second approach, optimization
we treat all the strategies as ‘participants’ of Z independently provided heats we transform the matrix of S(j,z) to the matrix of places that the ‘participants’ have achieved, Places(j,z) we consider as optimal the investment strategy that has the minimal sum of the places over all the heats: (21) 2 opinions the investment risk of a participant is independent of the length of his investment horizon, then: optimization is provided at the start time only and the strategies that were once found applies continuously up to retirement the investment risk does depend on the participants’ investment horizon, then: optimization is provided in the beginning of every interval
13
The mechanism of efficiency checking (1)
group all participants by the length of their investment horizon up to retirement, accurate within one year T is the number of groups we obtained the youngest participants has T years or intervals up to retirement variant – any approach with specific parameters: for the first approach: the form of A(t) the value of the method of calculating the efficiencies for the second approach: procedure used to simulate matrix PP the option whether the optimization is accomplished periodically or at the start time only
14
The mechanism of efficiency checking (2)
input data matrix P, containing V trajectories of actual market motion during the period of days. Should be obtained by cutting from actual market trajectories parameters of the variants to be checked checking of the variants is provided for every group of participants: applying the variants of forming strategies over the trajectories of the matrix P providing the analysis of the empirical distribution of the value of the portfolio that could be accumulated up to retirement while using one or another variant
15
The mechanism of efficiency checking (3)
analysis is provided by means of the stochastic dominance criteria of the first, the second and the third order for every pair of variants separately we consider the variant X as better than variant Y if the following inequality holds: (22) (23) (24)
16
Data used for checking
US financial market indexes (Dow Jones Industrials, index of 13-week US Treasury notes) period from 18.09.1986 to 30.04.2007 (5 199 daily quotes for every index) other parameters were chosen as , , ,so each trajectory of P must therefore have 40*125+100 = 5 100 daily quotes the cutting: first trajectory starts 18.09.1986, and every next starts one day after the preceding (then, V is equal to 5 199 – 5 100 + 1 = 100).
17
Variant 1: Calculation the efficiencies using direct statistical approach Variant 2: Calculation by simulation the trajectories using procedure 1 Variant 3: Calculation by simulation the trajectories using procedure 2 0 36 33 69 0 0 0 0 0 34 0 34
20
Comments