Tuesday, May 11, 2010

Annuities, Superannuation, and Geometric Progressions

After a dinner-table discussion about superannuation – I know: the Jones family is craaazily exciting – I came away with two problems I wanted to solve:
  1. If, after I retire, I would like to be able to spend X dollars/euro/whatever every year for N years, what lump sum do I need at retirement to achieve that?
  2. If I start saving now, how much do I need to put aside each year to reach that lump sum?
The first step is to ask yourself what level of income you'd like to have when you retire. My Dad suggested somewhere between 60-80% of your last income before retirement, which seems reasonable: it's not what you really need, but you're probably used to that level of income, and it's what you'd get on a generous pension (see, for example, Greece).
Alternatively, we can just name a figure – I'm calculating in current dollars here, so that should be fairly easy. Let's, an an example, set a goal of AU$60,000 per year for 30 years. (Call me aspirational. *grins*)

A stream of income like this is called an annuity, and calculating their value is reasonably easy; every Commerce student should have covered it introductory Finance. Pretend, for a moment, that you needed to pay someone else $60k, today – how much money would you need? Well, $60k, obviously. Now pretend you only had to pay this amount in one year, and could invest the money until then. How much money would you need today? Only enough so that what you currently have plus the interest you earn equals $60k. If you invest one (1) dollar today, in a year you will have (1+i) dollars, where i is the (nominal) interest rate, e.g. 0.06 or 6%. Working backwards, we can divide our debt from next year by (1+i) and find out how much we need right now; this is called the present value (PV) of the debt.
In algebraic notation, it looks like this:


If we had two years to invest before paying the $60k, we could take advantage of compound interest: the $1 we invest today would be worth (1+i) in one year, and that would be worth that much again the year after – (1+i)x(1+i), or (1+i)2. Algebraically:


By extension, a debt paid three years from now is worth (1+i)3th of its future value, and a debt paid n years from now can be discounted (the technical term for what we're doing) by (1+i)n.

It should hopefully only be a short leap to realising that, from our retirement savings, we will be paying off a debt each year – to ourselves. So the question becomes: if I want to pay myself $60k now, and $60k next year, and $60k the year after that, and so on for n years, how much do I need today? Easy: just take the amounts we calculated above in (1) and (2) and add them together: $60k for today, plus $60k divided by (1+i) for next year, plus ...


The payments quite clearly change from one year to the next at the same rate: to get from one year's payment to the one after it, you divide the first year's payment by a further (1+i), and that's it! That makes these payments a geometric progression, and the rate each one changes by – (1+i) – is the common ratio. So we can also express the amount we need to save as:

Now we get to be clever, and reduce the number of terms. For the sake of abstraction I'm just going to replace "$60,000" with "x", but we'll put it back in at the end:
Before you get too settled with equation (6), we should account for inflation: after all, we don't just want $60,000 every year, we want that amount of purchasing power every year; if the price of everything we buy goes up by 10%, we'd need $60,000 x (1+10%) = $66,000 to buy it all. Sparing you the messy details, (6) becomes (7):
Engineers, physicists and the like will despair at my use of π as a variable, but I am an economist and do not care: here and now, π represents the rate of inflation.

Let's bring this back into a format that you, gentle reader, can more easily utilise. Below you'll find a table containing four variables for you to enter, with the fifth row (in bold) containing equation 7 adapted to use cell references. Simple paste the table into the indicated cells and you should be good to go:

Row/ColumnAB
1Desired annuity60000
2Number of years30
3Expected inflation0.025
4Expected interest rate0.05
5Lump sum needed=B1*(1-(1+B3)^(B2+1)/(1+B4)^(B2+1))/(1-(1+B3)/(1+B4))

I've chosen what I called an "aspirational" target of $60k per year for 30 years – aspiration both because of the level (the average Australian salary is currently around $50,000) and the length (assume a retirement of around 65-70: 30 years beyond that would be a fair effort!). The Reserve Bank of Australia (RBA) has an inflation target of 2-3% over the medium term, so in this example I've pitched the expected inflation into the middle of that range, assuming the RBA will succeed. The expected interest rate is relatively conservative: I get more than that on my bank account. However, it's probably better to err on the side of caution when choosing these variables!

If you've copied the table into your spreadsheet and everything's worked, you should get a lump sum needed of $1,326,088 and change. This is the answer to Question 1, and will be the target of our next set of equations as we seek to answer Question 2.

This time, we're saving an amount every year of our working life – say, 40 years – so that, at the end, we've saved up $1.3 million (in current dollars). The money we invest at the start of of this period gathers more (compound) interest than the money we invest later on, so instead of discounting (dividing) our money figure x by the interest rate we'll multiply it. Conversely, inflation decreases the value of our savings, so that'll come into the equation as a divisor; essentially, the two terms (inflation and interest) have just changed places.

Equation 8 is basically Equation 5 with inflation added and the terms reversed because we're saving to achieve a future value (FV) rather than paying out from a present value.

Now the tricky part to this equation – and the reason I didn't posted this on, say, last Thursday – is accounting for wage increases. Why does that matter? Well, we've started off assuming that we're contributing a fixed amount x each year – but the reality is that, during our working lives, we can expect to earn less at the start and more at the end: it's going to be much easier to save money towards the end of our career than at the start.
The difficulty with accounting for wage increases is that their effect goes the other way to the inflation and the interest rate effects: your first salary will not have increased at all, but anything you save will be subject to a lot of inflation and interest; your last salary will (hopefully) have seen many wage increases, but the money you save from it won't earn any interest or be devalued by inflation. Observe:
The first term x · (1+w)n-1 is the amount (a percentage, perhaps) you're saving from the wage you've earned after 39 wage increases (n = 40 years). The second term is that same amount/percentage two years before retirement: only n-2=38 wage increases, but with one period of interest and one period of inflation effects. What took me a bit of time was converting that series into a neat expression


and figuring out the common ratio

Once you have it, it's not that difficult: (1+w) decreases in k (because we're counting last period to first period), so each progression needs to reduce that (1+w) term by one factor, but after working exclusively with terms that increased in k it took me a little while to figure out. On the plus side, I was proud when I did get there :-)

Anyway, multiplying both sides of (10) by [1-(11)], cancelling out, and then dividing again by [1-(11)] – analogue to what we did in Equation (5) to get to (6) – brings us to the final equation:

Bam. Isn't it pretty? A form only a parent could love, perhaps; be that as it may, it's still very useful! Technically, we can't actually answer Question 2 with the equation in its current form (it would need to be solved for x), but I promised you no more equations; instead I'll give you table with which to play around. You can figure out how much you'll save for various rates of contribution; once you've chosen values for inflation, interest, salary increase, and years worked, you can adjust the savings rate until you reach the lump sum calculated in Question 1:

Row/ColumnAB
11Starting annual salary50000
12Expected yearly salary increase0.025
13Years worked40
14Savings rate0.12
15Expected inflation0.025
16Expected interest rate0.07
17Lump sum at end=((B11*B14)*((1+B12)^(B13-1)))*(1-(((1+B16)^B13)/(((1+B15)^B13)*((1+B12)^B13))))/(1-(1+B16)/((1+B15)*(1+B12)))

I consider the starting values here to be reasonably realistic: the average current annual salary is $50,000; 40 years is a reasonable working life; the current superannuation contribution rate is 12% (in Australia); inflation is in the middle of the RBA's 2-3% band; a 7% interest rate is, I feel, conservative, taking the 10-year Government bond rate of ~5.5% as a baseline and looking to a "normal" long-term market investment return in the low double-digits. Yearly salary increases is not something I have a whole of knowledge about, but I've acted upon the assumption that we work in industries whose (relative) productive value does not decrease over time; that implies that our salaries would increase at least as much as the price level (i.e. by the rate of inflation).

This table is good for just plugging figures and seeing how big a difference various changes in the variables have – a change in the interest rate of 1-percentage point, for example, has almost as much effect as a 3-percentage point change in your savings rate. Lesson: choosing good investments is more important than just savings lots.

With the starting values, the expected savings lump sum is $918k-odd. That's obviously less than the $1.326m we're aiming for, but ratcheting up our savings rate to ~17.5% gets us there. If you think that's high, Germany has a compulsory superannuation contributions rate of roughly 20%; with the current Australian compulsory rate of 12%, another 6% of your post-tax income should do the trick.

Two final points:
  1. Yes, I have used "wage" and "salary" interchangeably here – it's bad, I'm sorry.
  2. I've not accounted for tax in my equations. This is an important omission. I'm not up to date on what the current tax scheme is for superannuation versus other forms on investment. The contributions (savings) themselves are tax free, as far as I know, as is the money you "pay" yourself once in retirement, but I don't know whether the interest you earn in the meantime is taxed, or at what rate. I have the vague recollection that there's a 15% tax in Australia when you withdraw your superannuation, but I'll have to look into that. If anyone knows the specifics, or can point me in the right direction, I'd welcome the assistance!
    Depending on where the tax falls it'll be more or less easy to integrate into the equation: a 15% tax on superannuation returns would simply involve replacing the term (1+i) with (1+i)·(1-t), which isn't too complex.

2 comments:

JB said...

Yo MJ, minimum compulsory Superannuation Guarantee is 9%, not 12%... So it's worse than you think...

But, there's the government co-contribution to consider. For the cool sum of $1k, the government will chip in an extra $1.5k. Fun, no?

Martin said...

Yeah, I know the (proposed) compulsory Superannuation Guarantee rise to 12% is only scheduled to come into full effect a few years from now (assuming it gets passed), but accounting for different levels of contribution would be a needless complication of the equation. The choice of what variables to put in (9%, 12%, 20% contribution) doesn't change the equation, just the final figure.

The government co-contribution is even trickier to model, since it's staggered according to your income, and there's no way of knowing how long or at what levels it will exist in the future. Plus, it makes the equation Australia-specific.
But you're right, for those on low to moderate incomes the Government co-contribution is a fabulous way to earn a risk-free 150% return on investment. If you can spare $1,000, you'd be a mug not to take the offer up.

Do you know off the top of your head, JB, how we tax super in Australia? Do we tax the contributions, the earnings, or the final payout?