The dilemma of advice

Some colleagues at Irrational Agency and I have recently been working on a framework around advice, and a key dilemma:

• Those who know less about finance, are more in need of advice
• But those who know less about finance, are less likely to realise they need advice

In some ways this is another form of Dunning-Kruger syndrome. But there's an interesting additional twist. Because advice is an information good, you can't make an informed decision about it before receiving it, because you don't know what the advice will be. And once you've received it, you may think you don't need it, because now you possess all the knowledge that the advice gave you!

So an element of trust is required to make the decision to buy advice, as well as an element of self-knowledge. Neither of these things are in abundant supply in the finance industry.

An outline of our analysis follows:

1. The context: more and better quality investment is needed. Consumers need to make better financial returns in order to finance retirement in an ageing society; and the economy (especially the UK economy) needs higher investment to increase growth beyond its current anaemic level.
2. The problem: The bounded rationality of retail investors presents a barrier to achieving higher investment outcomes.
1. Investors have cognitive biases that can reduce investment beneath its optimal level:
1. Risk aversion: many people don't like the risk of equity investment even if its average returns are higher
2. Loss aversion: many people are unwilling to accept any losses, and would rather sacrifice future growth in order to avoid the possibility of their capital being affected
3. Ambiguity aversion: nobody can say for sure what will happen to an investment, and this ambiguity is too much for some people.
2. Several cognitive limits also reduce people's likelihood to invest
1. Lack of comprehension: people don't understand the nature of investments, especially around concepts like diversification
2. Limited calculation capability: people aren't good at understanding the long-term compounding effect of higher or lower returns and fees
3. Lack of planning ability: people don't plan their future actions well
4. Lack of capability to commit to intrapersonal agreements with future self
3. Several possible solutions can help solve these challenges:
1. Personal cognitive tools: if investors can adopt new mindsets or can train themselves to overcome their cognitive barriers, they can make better choices. These tools might include:
1. Rules to understand compounding (e.g. the rule of 72)
2. Training tools such as Serious Games to help people become more resilient to risk and loss
2. Personally selected external tools: Investors might voluntarily adopt tools to help them overcome their own biases. For example:
1. automatic deposit into an investment account
2. term commitments that don't allow for withdrawals in case the investor becomes fearful
3. diversified low-fee funds that automatically diversify
3. Finally, institutions might offer externally imposed structures that move people towards optimal decisions, outside of their own free choice:
1. Tax and similar incentives: many countries offer short-term tax relief or a reduced overall tax burden in return for increased savings
2. Forced saving: taxation in return for guaranteed future income, which may or may not be directly tied to the amount saved
3. Cultural pressures e.g. "the housing ladder": cultural norms that push people towards saving more than they might otherwise be inclined to do
4. As a special case of the institutional/external offering, various forms of advice can change people's behaviour and encourage higher levels of investment or more effective investment choices.
1. Informal advice (non-professional, peer-to-peer): social information sharing or other pressures that encourage people to invest
2. Informal advice (journalistic, internet): information from media that is often free or low-cost, and generally comes without liability to the offerer, except for any risk to their reputation
3. Formal advice, in which an adviser personally talks to the investor and makes recommendations, which are typically different to what the investor would do on their own. These fall into three categories:
1. Independent and paid for by the investor
2. Independent and paid for by commission on products
3. Non-independent: offered alongside product
4. Guidance (non-tailored), in which guidance might be prepared for a class of investors or those with particular characteristics, but is not produced in consultation with the individual who is using it.
5. Guidance/advice (algorithmic/AI), in which an advice algorithm may be created which can produce customised recommendations based on an individual's data or answers to a questionnaire, but no individual adviser is involved.
4. A model of agency and influenced decisions
1. To explore the optimal combination of individual agency, intervention and advice, we can build a model that incorporates the likely outcomes of investment with and without intervention, as well as a factor addressing the process utility (positive or negative) that some individuals may experience from the interventions themselves.
2. Let us assume a set of investment instruments I = {i}, each of which has an average annual return r[i] and an expected variance v[i].
3. Each investor A has a 'natural' bundle of instruments B they would buy without intervention.
4. Each form of intervention can be defined as a function F:A,B->B. Executing F on investor A who would naturally choose B, results in A owning a different bundle of investments F(A,B).
5. We can calculate the total return and total variance of B and F(A,B), subtracting the former from the latter as an aggregate estimate of the impact of the intervention, rF and vF. The impact may be different depending on the characteristics of A.
6. Each intervention F also has a price, which we assume to be deducted from the value of the investment portfolio prior to purchase of a bundle.
7. Finally, F also has a process utility u(F,A) which is experienced by the investor. Sometimes, this may be a positive number (for example, if an investor feels supported by the adviser and after the process is happier and less stressed about the disposition of their investments). In other cases, it might be negative (for example, if a portion of income is deducted and invested on their behalf, and they object to the reduction in personal choice this represents).
8. The challenge for investors is that the impact of F, and in some cases the process utility u, are not observable by them in advance. To the extent that advice outcomes are correlated between investors, or that the intervention process is highly transparent, it may be possible for them to observe other people's outcomes and estimate the impact for themselves.
9. Given this, there is a probability 0 < p(F,A) < 1 that each intervention will be selected by an investor (even for mandatory interventions, investors may in theory arrange their affairs so that they are not subject to the intervention, for instance by reducing their income or moving offshore). The weighted impact of each intervention on total return can be defined as the sum of p rF across all investors. To evaluate the effectiveness of an intervention F at the systemic level, we would compare the impact of all interventions in a system with and without F; a single F might itself have a positive impact but may change the likelihood of investors adopting other interventions, so can't necessarily be studied in isolation.
5. Evidence
1. We have an expanding reference list informing the various items incorporated in the above model; please contact me if you're interested in seeing this or discussing further.
2. Detailed evidence on the estimation of the model and the impact of specific interventions will require primary data collection.
6. Conclusions
1. To determine the impact of adding a new intervention (for example a new type of advice, or a change in taxation) a model of consumer behaviour in the presence and absence of the intervention is required.
2. This model can be constructed from a combination of data on existing investor behaviour and newly collected data from investment simulation surveys.
3. A choice is also required: how much weight to place on process utility versus total return. Investors may value their autonomy and feel that the imposition of an intervention, even with a positive expectation value, is harmful to them. Alternatively, investors may have different views to government on the future returns of an instrument and therefore the impact of an intervention (for instance, a ban on cryptocurrency "to protect investors" might have a positive expectation value in the government's modelling, but keen crypto speculators may predict a negative expectation value).
4. Expectation values themselves can form part of an instrument, so there may be room to develop "intervention arbitrage" tools that can allow interventions to become a Pareto improvement.