1 Disclaimer

2 Defining the index I’m looking for

3 Slicing the pie

4 The contenders

5 The two fund solution - 0.14% expense ratio

6 Calculating the U.S./International Split

A MSCI Barra’s view of the world

B FTSE’s view of the world

C Deep history

C.1 The three fund solution - 0.19% expense ratio

C.2 A four fund solution - 0.17% expense ratio

C.3 Cooking the books

C.3.1 Percentage of world market accounted for by the US - 41%

C.3.2 Percentage of all-world accounted for by ex US Large and Mid Cap - 53%

C.3.3 Percentage of all-world accounted for by ex US Small Cap - 6.0%

C.3.4 Percentage of all-world accounted for by EAFE large and mid cap - 41.9%

C.3.5 Percentage of all-world accounted for by Emerging Stocks - 11.1%

1 Safety first

2 Short term cash

3 Which bank is safe?

4 Credit union or bank?

5 Finding a credit union to put emergency cash in

1. Find Weiss List of top rated credit unions in Washington State, e.g. credit unions that got a B+ or higher. (13 entries)
2. Combine with Bankrate’s list of credit unions in Washington State with 4 or higher rating (added another 11 I didn’t already have)
3. Bauer’s list was hard to work with so what I did was just look up the ratings for the 24 banks I got from steps 1 and 2 and add in Bauer’s ratings.
4. Strike any bank that didn’t get at least a B+ from Weiss, a 4 from Bankrate and a 4 from Bauer (Reduced list from 24 to 5).
5. Run quickly through list to find the credit unions whose eligibility criteria my family meets (Reduced list from 5 to 1)
6. While the one remaining bank doesn’t have a local branch, that’s o.k. because I’m just using it for savings. But it does allow us to open an account remotely. So all is good. Just to be paranoid I checked the remaining bank out with the BBB and they got an A+.

6 What about daily banking?

7 What about other options?

7.1 Treasury only money market fund

7.2 Treasury bonds

7.3 Short term government bonds funds

7.4 Zero-Percent Certificate of Indebtedness

Contents

1 How much does a single year of college cost today?

Living in Washington State the obvious college to think about is the University of Washington. UW’s total costs for the 2010-2011 academic year for an instate student living away from home are $21,033. In my opinion UW is currently the only world class university in Washington State and I wouldn’t want to pin all of our daughter’s hopes on one university. So we’ll have to face the reality that our daughter may go out of state. Since I went to UCLA let’s see what UCLA would cost for the 2008-2009 academic year for an out of state student - $50,214 for an out of state student living in an off campus apartment. And remember, the UCLA figure is what it costs for an out of state student to go to UCLA today, right now, this very moment. Not, say, oh 15 years from now.

2 How quickly are college costs going up?

The most comprehensive source of information I could find on college pricing is [Board(a)]. But the raw data is available as spreadsheet in [Board(b)], I used Table 5. Please note that when staring at the table below the numbers in table 5 are adjusted for inflation.

Unfortunately there is a problem with the data. The 2009-2010 school year saw the single largest year-on-year increase in public four-year college tuition in the entire data set (which admittedly only goes back to 1979). The increase for private colleges was one of the largest ever. This does odd things to the numbers. For example, if I calculate the 10 year geometric mean increase from 1999 through 2008 I get 3.03% for public colleges and 1.89% for private colleges. Quite a change from the numbers above. So do college tuition increases during the greatest recession in recent memory where colleges, public and private, were essentially set loose on their own with their endowments pummeled really serve as a good basis for calculating what’s going to happen over the next 15 years? Beats me.

My guess is that 2010-2011 is going to be even worse, I know the UC system on its own increased tuition by 32%. Yes, 32%. So maybe these sky high rises are just the future and we need to get used to it? Or more likely at some point the whole system collapses because as the old saying goes, anything that can’t go on forever, won’t.

But in the meantime I need to estimate the future college cost increases. So I’m going to take the 15 year number, for no particularly good reason and I’m going to average it. In other words I’ll assume that for the next 15 years college costs will go up at a rate of (0.0317 + 0.0252)/2 = 2.85%.

3 How much will the bill be?

The first question is - how long will our daughter take to get her undergraduate degree? I haven’t had much success in finding solid data on how long it takes to graduate from college but I did find one data point from 2000. The table on page 46 of [for Education Statistics(2003)]. It claims that students on average took 55 months to graduate although students who went to only one college took 51 months on average but even that number can be split into 53 months for a public school and 47 months for a private college. I don’t have any trend data on these figures and no clue what anyone is projecting for the future. Still the message here is that in 2000 public school students took about 4.5 years to graduate and private school students took 4 years. Given the outrageous costs of college the only way I can see the number of years substantially increasing is if students are attending school part time to try and make ends meet. So I’m going to guess that my daughter is probably going to go to college for 4 years. Or, put another way, given what we expect to be paying 15 years from now she better not need more than 4 years.

In terms of the cost basis I’m going to pick the UCLA out-of-state figure, $50,214 and the 2.85% growth rate. As my daughter is now 4 years old she has 15 years left until she goes to college. The only problem is that the school year (which determines when I have to cough up the money) doesn’t follow the calendar year. Schools start around September and run through June or so. This means my first check to college should be due around September of 2024.

The 2024 figure comes out because my daughter will turn 5 in 2011 and start kindergarten during the 2011-2012 school year. She will have 12 years of school after Kindergarten. So 2011+12 = 2023. In other words, she will graduate high school during the 2023-2024 school year. Which means she will start college during the 2024-2025 school year.

So the goal is to have enough money to pay for her first year of college by September of 2024. In reality I could stretch it out since most of her college expenses will be paid by the quarter/semester (tuition, fees, books and on-campus housing) or month (general living expenses and off-campus housing). But trying to be that exact over a 15 year period seems like useless specificity.

So, to keep myself mildly insane (as opposed to completely wacko) I’ll assume we’ll save money on a September-September basis. Or, in accountant speak, our college ’fiscal’ year starts in September.

And yes, this always drives me nuts at work. ”The project will be delivered by 2008”, ”Wait, is that calendar or fiscal year?”, ”AAAAHHHH!!!!!”

So, let’s see, $50,214 base figure, 2.85% real rate of growth (e.g. after inflation), for 15 years to save until her first year of college, with 4 years of college total, the total bill will be $319,499.38. (See the appendix for details)

Oy.

Keep in mind, that the previous figure is ’real’ not ’nominal’ money. Or in other words, this is ’inflation adjusted’. So $320,300 is how much I’m estimating it will cost to send our daughter to college in 2010 dollars. By way of comparison, sending our daughter to UCLA today for 4 years would cost $209,606.90.

So, I guess we can look on the bright side. Sending our daughter to college 15 years from now will ’only’ cost $109,892.48 more (in 2010 dollars) then it would to send her to college today. Oy

A Calculating the cost of college

The math for calculating a compounding entry is thankfully very straight forward:

In the example I gave above the CurrentCost is $50,214 for 2010-2011 and GrowthRate is 2.85%. Let’s say my daughter was going to college in one year. In that case the cost would increase over one year and be $50,214*(1+0.0285) = $51,645.10. In other words the YearsToGrow value would be 1 because there was only one year in which the growth would compound. So if my daughter is going to college in 15 years from now then the YearsToGrow value is 15.

So the math for calculating my daughter’s expected cost of going to college is:

The cost for sending my daughter to college in 2010-2011 would be:

References

Contents

1 Where should the money go?

The IRS has kindly summarized all the various options available to get tax benefits for spending money on education in [IRS(2009)].

That various tax credits and deductions one can take when spending taxable money on education all come with strict limitations on the income one can earn and still qualify for the tax credits. Marina and I certainly plan to exceed those limits by the time our daughter gets to college so we can’t rely on their use for our planning. U.S. Savings Bonds are also not a viable option for similar income cap reasons. The Coverdale IRA doesn’t work because its contribution limits are low given the costs of college and eventually we would hope the income limits would also be an issue. The IRA option isn’t a terribly good one for us because we need our IRA money to pay for our own retirements. If we have to choose between saving for our own retirement and paying for our daughter’s education then we will choose our retirement. It would be unfortunate if our daughter has to take on debt to go to college, it would be catastrophic for all involved if our daughter has to spend the rest of her life supporting her parents.

Although we probably can’t use the various tax deductions for taxable money we can still choose to invest for our daughter’s college education using taxable money, alternatively we can use a 529 plan. I’m not a huge fan of 529 plans. I don’t like the fact that if we don’t use all the money we have to pay a penalty above and beyond the deferred taxes to get it back. I also don’t like the fact that the investment options for 529s tend to be both limited and expensive. But the fact is that money we put into the 529 plan grows and exits completely tax free. This means that all the income on the principal we deposit in a 529 plan will never be subject to taxation. To put this benefit into perspective imagine our marginal tax bracket is T% and our college investments makes an I% return. In that case if we put the money in a taxable account our return after N years would be (1 + I * (1 - T))^N where as in a 529 our return would be (1 + I)^N. For example, if our marginal tax bracket was 25% and we got a 5% return on our investment over a 17 year period then we would get a - 1  23% or so higher return on investment from a 529 than a taxable plan. That’s the kind of benefit that is too expensive to give up.

So we’ll be using a 529 plan.

References

Contents

1 CollegeSure – buying one year at a time

The cheapest possible way to buy the CollegeSure CDs we need to pay for our daughter's education is to simply buy all 4 years worth now. Using the prices I got on 1/25/2008 the total cost would be $153,891.70 to buy 4 years, 1 unit a year. In theory we could raid our retirement investments to try and scrape up that much cash but I would rather leave our retirement funds alone. Especially since we can, if worst comes to worse, take out loans later to pay for our daughter's college but we really don't have a lot of options for rebuilding our retirement savings. Besides if we lock up the money in college savings now and our daughter doesn't go to college or goes to a cheaper college then we will have to pay all sorts of penalties to get the money out.

Given the various uncertainties I want to take my standard approach and spread the risk equally across the various years from now until our daughter is done with college. My standard approach is to figure out a constant sum of money, in inflation adjusted terms, to invest every year from now until our daughter is done with college. I don't claim the strategy is perfect but it's all I've got.

2 How many units to buy

So the bottom line seems to be that we will calculate the cost of buying 4 CDs (one for each year of college). The first step is to figure out how many units to buy. As previously explained one unit equals the average cost of one year's private college. But UCLA, my reference school, costs more than the average private college.

So it seems I should plan on purchasing 1.23 units for each school year. Obviously this relationship can change over time and as it does I'll have to adjust the target units to purchase.

3 Defining the problem

But the real issue that caused me grief is – how should we calculate how much money to put into each CD? Given that I want to save the same amount of money each year (in inflation adjusted terms) I need to figure out how much money each year to spend on which CDs so that at the end we will have the right number of units for each year our daughter is going to college and will have spent the same amount of money each year on buying CDs.

The first constraint is that we spend the same amount of money, in inflation adjusted terms, each year buying CDs. This can be expressed as:

Where P is the year that a CD is purchased, u is the college year the CD is intended to be used for, F is the first year of college, L is the last year of college and S represents the amount of money spent in year P on a CD to be used for year of college u. The goal then is for any given year of purchasing P the sum of the money spent on the CDs should equal to B where B, in inflation adjusted terms, is a constant. We also require that S always be 0 or higher since we are not going to short the CDs.

The second constraint is a function of our goal, to have enough the right number of CollegeCD units for each year of college our daughter will attend school.

Where α is the first year we would have purchased CDs and where DU represents how many units a dollar would buy in CollegeSure CDs purchased in year p for use in school year U. NU represents the number of units needed for year U. The summation term stops at U-1 because we can't buy a CD in the same college year it is to be used in. So the goal of the previous summation is that for any given college year U the total number of units purchased for that year will equal NU_U.

The previous is really just a series of linear equations. The trick then is to find the values for S that when put into the equations will meet both of the previous requirements. The only problem is that in the interesting case we have more variables than equations. Let's pretend that it's 2008 and our daughter was going to start school in 2012 and finish school in 2013. In that case we could express the system to be solved as:

Note: The empty spaces are 0s, I have left them out to make the matrix easier to read.

DU is known, this is the actual or estimated number of units a $1 could purchase in the given purchase year for the given school year. So DU_8,12 represents how many units of CollegeSure CDs $1 could buy in the year 2008 for use in the school year 2012. DU can be assumed to express its value in constant dollars so we don't have to worry about adjusting for inflation. NU is also known, for example, in the previous section I calculated that NU would be 1.23 for both years. The trick then is find values for S so that when multipled by the left hand matrix the result is that for each year the spending equals to some constant B and we have exactly NU units worth of CDs available for the 2012 and 2013 school years. In this case we have 9 variables but only 7 equations which in this case happens to mean that there are many possible legal solutions. So how the heck do we find the 'optimal' values for S that will produce the lowest possible B and still result in having NU units for each year of college? This assumes, btw, that there even is a solution. In some cases there is no exact solution and we will have to violate the constraints to some extent in order to get a result.

4 Reverse Proportional Approach

While playing with the math the best solution I could come up with for calculating the best purchase plan is what I call the reverse proportional approach. In this approach at each purchase year I spend the budget in proportion to how much money it would cost to purchase all the units needed for that school year. I start this process at the last purchase year and work backwards. The following equations describe the approach:

NeededDollars can be negative so I put in a check that returns 0 to cover that possibility in my code.

To calculate my purchase budget I start with an initial budget that I'm sure will be larger than I need (see BMAX later for details) and then start a linear search trying to find a value for B that produces a purchase budget that buys NU units worth of CDs for each college year.

I've tried a number of other approaches to finding a good purchase plan including Monte Carlo simulations and genetic algorithms but neither approach was able to produce a better solution than the reverse proportional approach. Unfortunately this means nothing in the universal scheme of things. Just because I can't find a better approach then the reverse proportional approach doesn't mean one doesn't exist. It just means that I, a complete amateur, couldn't do better.

5 Calculating DU

The DU function returns, for a given year of purchase and year of college, how many units $1 will buy. To calculate this value I start off with the cost of buying CollegeSure CDs quoted to me on 1/24/2008 and 1/25/2008. The question I asked CollegeSure was "If I gave you $1,000 today for a CollegeSure CD that will expire in year X how many units will I have bought?" For CDs purchased in 2008 this is the values they gave me:

The previous table assumes that the year of purchase is 2008. I'm not going to worry about DU where the year of purchase is before 2008 since we aren't going to travel back in time. But I do have to worry about the value of DU when the year of purchase is greater than 2008. In that case I will assume that the values in the table given above hold but are adjusted by the yearly increase in college costs. This is a complete blind guess but I have to assume some sort of cost increase, above inflation, each year to compensate for the ever increasing cost of college.

Where Table indexes into the previous table assuming that the first entry has the index value of 0. CI is the college cost increase and represents how much we expect college costs to go up each year. Previously I had calculated this value as 2.6% after inflation. I believe that CollegeSure CDs are actually available for maturities up to 22 years but I only asked for the values for up to 20 years which results in u-p having to be less than or equal to 19 in order to index into the table.

Also note that all calculations assume that one is buying on the date that the data in the table was recorded and that all future purchases will also be on that date. I haven't put in any kind of corrective when the purchase date and the table date is off. My intended solution to this problem is to get a new quote when I'm ready to buy. Yes, this has obvious issues if my buy dates are all different but given the timescale and uncertainties involved I don't care.

6 Calculating BMAX

Calculating a good initial guess for the upper bound in the linear search turns out to matter, a lot. If the value is lower than the actual minimal value then a correct result cannot be found. If the value is too much higher than the minimum value then it will take a long time for the linear search to find the result (a real issue in Javascript where the performance is typically pretty awful). While I was working on genetic algorithms to find the best portfolio I came up with a value called BMAX which represents the maximum theoretical yearly budget that would be guaranteed to spend enough money to buy the necessary number of units. The trick was that I wanted a BMAX that would be reasonably close to the 'best' value so as to reduce the search space.

To calculate B_MAX I used an approximation. Let's pretend I only needed to buy a single year's worth of college units. In that case solving the matrix above is trivial. Every year's budget (e.g. S) will be the exact same value. So I just need to calculate what budget, B (which all the S's equal to) will purchase 1 unit.

In other words the budget needed every year from year α to year U-1 for college year U (whichever that is) equals B_U iff we are just saving for a single year. But I can use B_U as a crude way of calculating B_MAX.

7 Using the calculator

The calculator implements the reverse proportional approach. The calculator is fairly self explanatory. The key issue is that the data used to drive it is hopelessly out of date. I will contact the CollegeSure CD folks to get an update on the prices. If you have your own update you can download the file and do a search on the variable UnitsForDollars. It is an array that contains how many units a $1 can buy for a 1 year, 2 year, etc. CD.

In future years we can use the calculator by adjusting the units for each college year so they don't include the number of units we purchased this year. E.g. if we have already purchased Y units for college year X then we would enter NU_X -Y units in the calculator for year X.

Contents

Picking the right financial assets to save for college is a question of matching our needs to our financial assets. In this case we need to craft a portfolio designed to deliver a certain amount of money at a specified point in time.

1 No stocks

I like stocks. I know I shouldn't. I've studied enough of the international history of stock market movements to know that they are insanely risky and are most certainly not safer in the long run. But I like them. So when I started on this journey I was focused on what stocks and bonds to buy even though I knew this made little sense. The trap I fell into was not matching my assets to my liabilities. For example, if I am saving money for a fixed expense that will come due at a fixed point in the future does it make any sense to use an investment that makes absolutely no guarantees what so ever about how much it will return and when? If I was suggesting, say, using the income or some portion of a stock investment to pay for college that would be one thing. But because we are using a 529 plan all the money in the plan needs to go to education and therefore using stocks means believing that we can liquidate the entire portfolio at a fixed date and expect a predicable return. That simply isn't how stocks work so using stocks to save for college in a 529 plan just doesn't make sense to me.

2 Bond mutual funds

In many ways bonds would be a good match for college savings. They promise to pay a certain amount of money at a certain time. This is much better than, say, stocks. The problem is that I can't find any 529s that allow one to directly invest in traditional bonds. Instead the 529s offer bond mutual funds. A bond mutual fund is a different beast than buying and holding individual bonds to maturity. For one thing bond mutual funds make no promise about how much they will pay out. E.g. if I buy a 5% bond and hold it until maturity then I'm promised by the bond issuer to receive my 5% (subject to the risks explained in the TIPS bond mutual funds section below). A bond mutual fund makes no such promise. So fundamentally bond mutual funds aren't all that different than stocks in the sense that nobody is promising anything.

If our daughter is going to college in N years and we buy a bond with a N year maturity then in N years the bond matures, we get the principal and we move on with our lives. The amount of principal we get is set when we bought the bond. But if we need to sell the bond early, before its maturity, then we face interest rate risk. If market rates are higher than the bond's interest rate then we will have to sell the bond at a loss and lose some of our principal. This particular risk isn't an issue if we hold the bond to maturity.

But since bond mutual funds are essentially expiration free bonds whenever we need to pull out our money we have to face the very real risk of losing 'principal' (our initial investment) due to recent market movements. The risk of bad market movements taking a huge chunk out of the bond's value is quite real. For example, the Lehman Long Term Government Bond Index (which covers the kind of bonds I would imagine we would use to save for college) in 1999 went down 10.26% (after adjusting for inflation) in one year. This is a 10.26% loss in value due completely to interest rate risk (since the bonds themselves were still paying a positive return). Bonds have lots of risks but taking a 10% hair cut on the value of one's principal is (modulo general credit risk) not typically one of them.

Just to add insult to injury not only don't bond mutual funds have a well defined maturity but they also don't have a defined interest rate either. Normally one buys a bond at X% and gets X% repayments. It's not quite that simple since, outside of zero coupon bonds, there is re-investment risk. But with a bond mutual fund there is no defined interest rate which makes planning a nightmare. So my inclination is to avoid bond mutual funds as being incompatible with saving for a fixed expense at a fixed time.

BTW, much as with stocks, bond mutual funds can be an outstanding investment tool, especially given their benefits around diversification and a concurrent reduction in credit risk. I'm not panning stocks or bond mutual funds, I am however of the belief that they aren't the best investments to use to save for fixed expenses.

2.1 TIPS bond mutual funds

If I absolutely had to pick a bond mutual fund to invest in I would pick a Treasury Inflation Protected Securities (TIPS) bond mutual fund. TIPS are issued by the United States Government and have one special feature, their principal is adjusted for inflation. This means that (in theory anyway) TIPS don't suffer from inflation risk. Also, because they are issued by the U.S. Government, one can fairly say that they don't suffer from credit risk either¹ . TIPS however do suffer from interest rate risk, this shows up both in terms of calculating re-investment value of interest earned while we hold the mutual fund as well as possible risk when selling the mutual fund.

To help me estimate the likely return on TIPS I took a look at my favorite TIPS bond mutual fund, the Vanguard TIPS fund. It's maturity is 10 years (meaning the average bond held by the fund will expire in 10 years). To help me get some idea of what kind of return I could be looking at I looked at the historical yearly returns for 10 year TIPS bonds provided by the United States Federal Reserve. Unfortunately the data only goes back to 2003 even though TIPS were introduced in 1997. The geometric average return from 2003-2006 would be 2.0%. Still, 4 years isn't much data to go on. So I decided to take the 10 year Treasury bond return, adjust for inflation, and see what that would have returned. In this case the government provides data going back to 1962. After adjusting for inflation the geometric average return from 1962-2006 was 2.77%. But to be fair we will be saving money for 18 years. So I decided to do something extremely unscientific and calculate the geometric average return for all overlapping 18 year periods and then take the arithmetic average of the result. The mean return of the 18 year periods was 3.09% with the worst 18 year period (1963-1980) returning 0.93% and the best 18 year period (1981-1998) returning 4.90%.

For me, for some completely irrational reason, 2% is a magic number. 8 of the 28 18 year periods would have returned less than 2.55% which, given the fund expense of the Vanguard funds in a 529 plan, is what I would need to get a 2% return after expenses. Of course all these numbers are just that, numbers. There is no way for me to know, even using these numbers, what actual return I would have gotten since the bonds the mutual fund will buy aren't purchased based on some mythical yearly rate but based on the market rate on the date they are purchased.

In the end I'm going to fudge and estimate that a TIPS mutual fund would return around 2%, which means a 1.45% return in a 529 account thanks to mutual fund expenses.

3 Bond like investments available in 529 plans

I can't really find a way to buy traditional bonds through a 529 plan but there are some bond like creatures available. I did a search for bond 'like' offerings in 529 plans available to residents of the state of Washington.

The CollegeSure CD is a very interesting beast and I will examine it on its own later in this article.

The TIAA-CREF Guaranteed Return plan gets described in various ways but I believe the basic deal is that it will return 3% and might return more. Each year it looks like TIAA-CREF declares an interest rate and guarantees it for that year. Acacia Principal Plus Account works the same way with a 0.15% fee on the return. A 3% return isn't enough to even keep up with inflation and while their declared rates have been higher it doesn't look like they have yet to go over 4%. This doesn't seem like a good idea.

The Louisiana plan provides a state guarantee that any money invested in the principal protection option cannot lose principal but there is no guarantee of the interest rate. Illinois, Maine and Texas are similar except the sponsoring state appears not to guarantee that principal won't be lost. The North Carolina fund is run by the state but it isn't clear to me if it is guaranteed by the state. These options might be interesting for holding money just before spending it on college bills but without any idea of the likely return it doesn't seem a useful option for long term savings.

The Fifth Third 529 CDs are super long term CDs, up to 12 years, with a guaranteed interest rate. Unfortunately the interest rate does not adjust for inflation, neither the normal kind nor college price inflation. When I wrote this they were offering APYs of 4% interest rates for up to 19 years which isn't bad for a FDIC insured investment but I'm way too skittish to sign up for that return given the future possibilities of inflation. So this option won't work for me either.

4 CollegeSure CD

The CollegeSure CD plan offered by the College Savings Bank is the most interesting option in my opinion. It uses the Independent College 500 Index (IC500) calculated by the College Board to track changes in private four year college costs in the U.S. The CD then guarantees a return that will match the increase in the IC500 index. Of course anything that sounds too good to be true usually is and this is no exception. The first thing to notice in their disclosure statement [Fund(2007)] is the number 1.50%. This number is mentioned on page 2 and what it represents is the profit margin. The way the CollegeSure CD works is that every year the CD grows in value by the same rate of growth as the IC500 minus 1.5%. However the CollegeSure CD is sold at a premium to its face value so this is supposed to make up for the 1.5% return that the College Savings Bank is taking for itself.

Still, a guaranteed return, especially once that tracks the IC500 is nothing to sneeze at. So this brings up the next question – how sure can I be that the College Savings Bank will meet its obligations? The College Savings Bank is very quick to point out that their CDs are FDIC insured. This means that if the bank goes belly up the FDIC will guarantee not only the initial principal but all interest earned up to the point that the bank went bankrupt. There are still a few problems. First, it wasn't clear to the College Saving Bank if their CDs would qualify for FDIC insurance so they got opinions from the FDIC saying that their CDs should be insured. Note I say 'should'. Section I of [Fund(2007)] discloses the fact that the FDIC has never actually said they absolutely would insure the CDs only that given the FDIC's current interpretations of various regulations that weren't really written to cover the CollegeSure CDs it was the opinion of the FDIC that the CDs should be covered. But an opinion isn't a promise. The only real way to be sure that the CollegeSure CDs are covered are either for explicit rules to be passed covering them or for College Savings Bank to default and to see if the FDIC covers them.

But even if I assume that the CDs are covered there is still the problem that if the College Savings Bank can't meet its promises then I could find myself in a long painful struggle with the FDIC to get my money back. More to the point, how can I be sure that the College Savings Bank can meet its promises? Section VI. explains who the College Savings Bank is, the answer being – they are a wholly owned subsidiary of Pacific Lifecorp, a life insurance company.

A.M. Best, Moody's, Fitch and Standard and Poor's rate Pacific Lifecorp.

All the rating agencies use the same BB, BBB, A, AA and AAA rating scale for 'investment grade'. So an A is good but not great. The AAAL rating is just for the CollegeSure CD itself, independent of Pacific Lifecorp. The rating is actually AAA with an L meaning that the AAA only applies to the amount covered by FDIC insurance. In other words S&P, I think, is really saying "we think the FDIC will insure this product." Just to add to the fun the College Savings Bank itself makes its money by investing in 'high quality' adjustable rate mortgages (no you didn't read that wrong).

4.1 Penalty for early withdrawal

If we should need or want to cash in the CD early we face a number of penalties. The first penalty is that the guarantee that the CD will match the IC500's growth only applies if we hold the CD to maturity. If we cash it in early then we get the return equal to the IC500 – 1.5%. This guarantee matters because if college inflation is less than was estimated by the College Savings Bank when it issued a CD then the 1.5% profit margin will leave less value in the CD than is needed to buy one year of college. If the CD is held to maturity then College Savings Bank will pay off the difference, but if the CD is sold early then you get what you get. In addition any CD terminated within the first 3 years faces a 10% penalty, which then drops to 5% until the last year (for CDs that had an original maturity of greater than 3 years) when it is 1%.

5 Age Based Funds

There are also 'age appropriate' mutual funds. These funds are keyed to certain maturity dates and are supposed to use a combination of stocks and bonds that shift over time. The idea generally being to start off 'risky' with lots of stocks and as our daughter gets closer to college the fund automatically switches into bonds. As [Bodie and Clowes(2003)] points out this approach doesn't make sense even in theory much less practice. The funds act as mechanical clocks, at certain points they change the stock/bond mix in a precalculated way. But what if the stock market has a huge run up, would it make sense to change sooner to bonds? What if stocks, early on, do horribly, would it make sense to stay in stocks longer? These aren't choices these funds can make. More to the point, as I discussed previously both stocks and bond mutual funds have serious risk consequences that make me generally not favor them for savings purposes.

6 Comparing the Top Two Contenders

The two assets that really caught my attention are TIPS mutual funds and CollegeSure CDs. [Bodie and Clowes(2003)] in particular directly recommends using CollegeSure CDs. To help me choose what to do I want to get a sense of how much money we will have to spend for one versus the other.

CollegeSure CDs are purchased in units where one unit equals the average cost of one year's tuition, fees, room and board at a private college in the US as determined by the IC500. The idea being that if I buy 1 unit worth of CD this year then in X years when the CD matures I will get back the cash equivalent of 1 unit at current prices. I called up the College Savings Bank and found out that for the 2007 school year one unit has $35,272 worth of value.

In the case of TIPS my choice would be the Nevada 529 plan which uses the Vanguard TIPS fund with an expense fee of 0.55% (which is more than twice the 0.2% fee Vanguard charges for the fund outside of a 529 plan). This is the cheapest reputable TIPS fund I could find in a 529 plan.

Below is the output of comparing the costs of buying 1 unit of a CollegeSure CD as quoted by CollegeSure (note that prices fluctuate daily) and how much I believe I would have to buy in TIPS mutual fund shares to get the same result. In the case of TIPS I list two numbers, a straight calculation of how much I would have to save in TIPS to have the expected money needed to pay for college and also that number plus 15%. The 15% is my attempt to compensate for the fluctuations in TIPS interest rates so as to give me a guarantee roughly equal to CollegeSure's. See A for details on how I calculated the TIPS savings value and B for details on where the 15% number came from.

7 And the Winner Is?

Given the rampant uncertainties in all the various numbers I call the CollegeSure CD cost and the TIPS w/15% extra protection cost a tie. So the decision really comes down to – which one will let us sleep better at night? Should we pick the CollegeSure CD and worry about Pacific Lifecorp or the College Savings Bank going bust or should we use TIPS and worry about college costs going up faster than the 2.6% used in our calculations? In the end we picked the CollegeSure CDs. Having a 'guarantee' that a dollars worth of college savings we put in now will be worth a dollar in the future is more enticing then worrying about Pacific Lifecorp/College Savings Bank's financial state. Yes, I know, famous last words. But in finance you role the dice and take your chances.

References

A Comparing CollegeSure CDs and TIPS Mutual Funds

By way of background I will assume that college costs are going to go up by my previously estimated 2.6% a year. The cheapest TIPS mutual fund I could find is Nevada's Vanguard Fund which charges a 0.55% fee. Also, as previously mentioned, I will assume that the TIPS fund will make a real yearly return of 2%. So to calculate how much money I would have to save today in order to have one unit of value in N years the formula would be:

where S is the amount to be saved today and is the variable to be solved for. TI is the TIPS interest rate. TE is the expense ratio for the TIPS fund. U is the cost of a single unit in today's dollars. CI is the inflation rate for college costs. N is the number of years from now until our daughter goes to college. In our case TI = 2%, TE = 0.55%, U = $35,272 (e.q. IC500 unit cost for 2007), CI = 2.6% and N is range of values from 17 to 20. Note however that I'm being lazy. The CollegeSure CDs will expire in the middle of the year (when school starts) and not at the end of the year. So to make a fair comparison N should be at least calculated in months and the interest rates should be translated to monthly values. But I'm too lazy to do that work so my estimates will all be a bit off. But given all the other uncertainties this doesn't unduly worry me.

B How Risky Are Bond Mutual Funds?

In the previous table I calculated the returns on various Lehman Brothers bond indexes. I admit the data is for a short period of time but finding bond index data has proven to be nearly mission impossible, the data above was taken from benchmarks that Vanguard uses for its bond index funds. If we had used, say, long term bonds to pay for our daughter's college fund then on the good side we would have seen a 6% real return but on the bad side we would have had to deal with watching up to 10% of the value of the portfolio vanish within a one year period because of interest rate changes. To be fair this was the worst year. The best year was 1995 when the index return 27.34% (and yes, that is inflation adjusted).

What I wondered is – how much more would we have to save in order to reasonably insulate ourselves from mark to market risk (the risk that interest rates will go against us when we need to liquidate our bond mutual fund position)? To answer this question I looked at duration. Duration is a measure of the interest rate sensitivity of a bond or bond mutual fund. A bond with a 2.2 year duration will be expected to go up by 2.2% if general interest rates go up 1% and down 2.2% if general interest rates go down 1%. At the time I write this the duration of the Vanguard TIPS mutual fund is 6.6 years. If that is the duration when I need to sell the bond then an interest rate change of 1% would change the value of my mutual fund shares by 6.6%.

To help me get some idea of what kind of volatility this implies in practice I got the results for 10 year treasury bonds, which seem to have a maturity roughly equal to the TIPS bonds from 1962-2006 and calculated the average yearly interest rate change. The simple average actually came out to 0.02% and standard deviation came out to 1.01%. The worst single year return was -2.95% which with a 6.6 year duration would mean a worst case scenario of losing 19.47% of the value of the portfolio in one year. But here's the really interesting part. If yearly interest rate changes do follow a normal distribution then given the mean and standard deviation a -2.95% change is 2.92 standard deviations out. To put this in perspective if the yearly interest rate changes follow a normal distribution then one would only expect to see a return that close to three standard deviations about twice every 1000 years or so. So my guess is, this puppy ain't normal. Of course I just proved beyond any shadow of a doubt that I'm no statistician. There are well defined tests to see if a distribution is normal within a certain confidence interval but it's been too many years between me and college statistics for me to remember how to do this test.

Lacking a better idea I'll randomly pick 2 standard deviations out or 2.02% which means 2.02 * 6.6 = 14.52%. So this means that if I want to use a TIPS bond fund to invest for our daughter's college education we need to 'over save' by 15% to make sure we can ride out any fluctuations in the interest rates in the market.

C But Doesn't Modern Portfolio Theory make stocks more reasonably predictable, at least in the aggregate?

My own encounter with personal finance really began with [Bernstein(2000)] and his explanation of modern portfolio theory (MPT). This naturally led to [Malkiel(1996)] which also extols the power and virtues of MPT. I mention MPT mostly because it provides a model to describe the expected behavior of stocks and so can be used to try and answer my basic questions about the suitability of an asset for savings. The only problem is that I am now convinced that MPT's core assumption isn't even approximately correct. This assumption is that stock returns can be estimated with a normal distribution. I later found out that it has been known since at least the 1960s that this simplification is wildly wrong. Thanks to work by folks like Mandelbrot (nicely summarized in [Mandelbrot and Hudson(2004)]) we know that stock returns do not follow a normal distribution. This devastates MPT because it means that the distributions and correlations that underly MPT's logic, and give me some basis of answering the basic risk/return questions, are just plain not true.

The situation is actually worse than it sounds because while the evidence seems, to me at least, to be clear that stocks don't fit a normal distribution I have not yet been able to find any compelling evidence as to what distribution stocks do follow. In fact, I'm beginning to seriously suspect that stocks do not follow any particular distribution on a consistent basis. Put another way, let's pretend stock returns did follow a normal distribution. But what if the shape of the distribution itself randomly changed in unpredictable but significant ways over time? Knowing that whatever distribution was used is normal would suddenly not be terribly useful.

I have personally reached a point where I have no clue what returns or volatility one should expect from stocks. What I am sure of, as excellent books like [Bodie and Clowes(2003)] explain, is that stocks are insanely volatile and that they are in fact not safer in the long run. That is, the probability of something 'really bad' happening to stock returns increases the longer one holds stock. Yes, one can argue that the opposite is also true, the longer one holds stocks the more likely something really good would happen but this actually goes to the heart of savings. Savings, in my mind, is about meeting a particular goal at a particular time. I don't know that I care that I exceed the goal, the priority is to at least meet it which means in the case of savings I'm more worried about loss than gain.

Given stock's unpredictable returns and high volatility there is absolutely no way to be sure that stocks will actually return what we need when we need it. In fact the only thing one seems to be able to be sure of with stocks is that there isn't much to be sure of. Nor are stocks even intending to provide any guarantee. The implicit deal with stocks is that in return for buying a piece of the company one gets to hold on for the ride and while promises are made (although not necessarily kept) that the ride will be, on average, better than it is worse, no promises are even attempted in terms of promising any kind of return over any kind of period. In other words stocks don't even pretend to be predictable.

So while I think there is a lot of logic in purchasing stocks as an investment² buying stocks for a fixed savings goal like college looks little better than buying a lottery ticket. So I will be avoiding stocks in our college portfolio.

Contents

1 Vetting the Players

In deciding which insurance companies to work with I want the answer to one question – Will the company pay off when an insured event occurs? It turns out that the answer to this question comes in two parts:

• Is the company financially able topay off what it owes? To answer this question I check the "long-term" rating with major credit rating agencies A.M. Best (aa or aaa), Moody's (Aaa or Aa) and Standard and Poor's (AAA or AA). The values in parenthesis are the ratings each company uses to identify a company in outstanding financial health. I usually have to search a bit to find the 'composite' rating for the entire corporation.
• Does the company pay off claims fairly or does it try to screw its customers?
  • First I go to Washington State Insurance Commissioner's website and navigate to Publications->Auto->Auto Insurance Annual Complaint Comparison Guide. This lists all the companies licensed to sell auto insurance in Washington state and assigns them a 'complaint index' based on how many complaints the commissioner's office gets about the company as a function of their size in Washington State. There is a similar report for home insurance, just go to the "insurance type" drop down and select "homeowners". I try to find companies that do well on both lists. Note that the same company can appear multiple times on the list because different people get assigned to different subsidiaries of the company and those subsidiaries can and do have very different ways of treating their customers.
  • I then go to Better Business Bureau and see what kind of complaints the company has. Large companies will have an "HQ" entry where the BBB collects complaints from all over the country. The BBB usually shares a general summary of the company (e.g. Satisfactory or Unsatisfactory) as well as a break down of how many complaints of what type the company gets and how the company resolves those complaints.
  • I also check J.D. Power and Associates Homeowners Insurance [Center(2006b)] and Automobile Insurance [Center(2006a)] ratings which give an 'over all experience' rating of up to 5 points.
  • Finally I'll search the Internet for the company's name with words like "sucks" or "rip off" or "cheat" and see what comes up.

2 The Line Up

Once upon a time I would have spent a bunch of time figuring out which companies to talk to. This typically saved me a lot of money. But now that I have a child I just don't have time the way I used to. So I cheated. I included Allstate because they are my current insurer and I have had good experience with them. I picked State Farm because their over all scores are actually pretty good and I used to have them and I liked them as well. Met life ended up on the list via an independent insurance agent. Amica ended up on the list because I decided to add J.D. Power's insurance surveys and they had won both the home and auto for the last several years running.

3 Auto Insurance

I cover the details in my previous article on buying auto insurance. Only two things have changed since I wrote that article.

1. Marina and I decided not to renew our AAA membership because it was too expensive (and we didn't save enough with AAA discounts) and now use emergency repair services from our insurance company
2. Although we occasionally check on-line prices for auto insurance we generally want to get all of our insurance coverage from one company

But a brief highlight of what we look for:

1. Our liability (both normal and under insured) – I have no idea what to set this to but it's set for us if we get umbrella insurance
2. We take a $1,000 deductible on collision/comprehensive (generally the highest on offer)
3. We don't carry collision/comprehensive on cars whose value is in the $2,000 range
4. We don't carry personal injury protection (PIP) or other insurer provided medical insurance (we both have health insurance so we don't need the extra coverage)
5. We make sure to point out that both Marina and I drive very little (under 7,500 miles a year) which usually gets us an extra discount

We also get a free copy of our C.L.U.E. Report to make sure we know what the auto insurance companies know about us including accidents, claims, etc.

4 Home Owners Insurance

For the basics we look for:

All Perils Coverage

Home insurance can either cover "named perils" or "all perils". Named perils only protects against the specific things the insurance coverage lists. All perils covers everything that isn't explicitly excluded like wars, floods, etc.

Insured Home Value

In the end home owners insurance is about, in the worst case, rebuilding the house from the ground up if it's destroyed. We have no idea how much that would actually cost so we have to rely on the insurance companies to give us an estimate. Yes, we can go investigate cost per square foot for constructions in our area, I actually tried that while I was buying our house to get a sense for how big the housing bubble was and frankly it didn't work. Nobody was willing to give me an estimate, there were always too many caveats. Unfortunately the price that houses sell for aren't a useful guide because they include the cost of building the house, the cost of the land the house is on and the developer's total profit. So in the end I had to rely on the prices I got across multiple quotes combined with Dwelling Replacement Cost Coverage (see next).

Building Ordinance & Law Coverage/Dwelling Replacement Cost Coverage

Changes in the law can significantly increase the cost of rebuilding or repairing a house. Stricter building codes can mean that when repairing damage not only does the repair have to meet the new codes but so does anything in the house the comes near the repair, this can cause serious price increases. So we make sure the policy has 'extra' coverage in case we don't have enough insurance. 25%-30% "overage" coverage is pretty typical. Note that "guaranteed" coverage isn't typically guaranteed, there is almost always a blanket limit somewhere you can't go over.

Personal Property Replacement Cost Coverage

By default when a piece of personal property is stolen or destroyed and we use insurance to replace it all we will get from the insurance company is the value of the item minus depreciation. So in the case of our 10 year old TV set if it were stolen we would get the cash value of a 10 year old TV set, which isn't much. In practice this means we would not be able to buy a new TV set. To address this problem home insurance policies (and rental policies) offer replacement cost coverage. With this coverage if an item has to be replaced the insurance company will give us enough money to buy a new item.

Loss Of Use Coverage

If the house is damaged badly enough that we can't live there while it's being repaired then we will need money to live somewhere. Coverage tends to get quoted either for a period of time (E.g. 12 or 24 months) or a dollar amount (e.g. 10% of the insured value of the property).

Personal Liability Protection

Beats me what's 'enough', but this is set for us if we get umbrella insurance.

Sewer/Drain Backup

Home Owners insurance doesn't usually cover water related damage but many will have a rider to coverage water damage from sewer/drain backup. This is about the only serious source of water damage for any of the homes I've lived in so I like to have it.

Extended Personal Property Coverage

Personal property coverage tends to be really limited for more expensive items like art or jewelry. For these more expensive items we have to get special riders that identify the property directly and covers it. I don't bother however with items like computers that (at least in our case) are worth much less than our deductible.

Deductible

We tend to go for the highest deductible we can. Currently that's around $5,000. Some companies offered higher deductibles but the savings in premium for these higher deductibles have been in the $10-$20/year range so we didn't bother.

Earthquake Coverage

As explained in more detail below I decided to get earthquake coverage. We took a 10% deductible. It's high enough to lower our costs but low enough to not leave us 'naked'. Allstate offered us earthquake coverage viaGeoVera but they use a 'one pot' deductible that ended up meaning that in practice the dollar value of our deductible was literally twice what it would be under a straight "10% of the value of the house" deductible.

Flood Insurance

As mentioned below floods are not a real issue where we live so we decided to skip flood insurance.

Mold Insurance

We live in new construction that in the time we have lived here has shown itself to be of good quality so we don't believe the probability of mold issues is high enough to be worth addressing. As such we skipped mold coverage.

4.1 Earthquake and Flood Insurance

That the Pacific Northwest is subject to devastating earthquakes is a closed question[U.S. Department of the Interior(2003)]. The Seattle area in particular is criss crossed with faults. But just because a bad thing can happen doesn't necessarily mean it will. The one question I've never been able to get a decent answer to is – what is the probability of a devastating earthquake hitting the Redmond area in the next 50 or 100 years? The chapter titled "When and Where Will the Next Big Earthquake Occur" [Noson et al.(1998)Noson, Qamar, and Thorsen] is as close as I have found to a prediction and if I'm reading it right it predicts 6.5/7.0 quakes every 35 to 110 years and even larger quakes every few hundred years.

But even that information isn't necessarily useful. For example, the 2001 Nisqually quake was 6.8 and while it was scary the damage to my home in Redmond and Redmond in general was pretty much zero. It didn't even shake the items off my shelves at home. The reason for the low level of damage is that the Nisqually earthquake happened very deep below the ground so by the time it reached Redmond the over all impact had been significantly reduced.

But I did find [Kahn et al.(2001)Kahn, Gamble, McConaghy, and Srull]prepared for the city of Redmond that outlined a number of risks that the city faced. The top of the list was earthquakes. I can't say that the reasoning in the report gave me a lot of confidence but it's the best I have. The report also explained that floods, although somewhat common, tend to be minor and aren't likely to occur anywhere near where I live. So I'll skip flood insurance. One new fact that I did learn from the report is that the area in which I specifically live is subject to landslides. Thankfully the area is thickly forested with deep ground cover but still, it's good to know. I will need to make sure that our earthquake coverage also covers landslides.

5 Umbrella Insurance

The general rule of thumb seems to be to have enough liability insurance to at least cover one's net worth. I suspect this rule comes from the famous American saying "I'll sue you for everything you're worth." But the rule doesn't make any sense to me.

For example, if we get sued for say, slander (all puns intended), our liability insurance would have to cover both the cost of hiring attorneys as well as the cost of any eventual damages if we lose. I can't come up with any rational reason why the cost of the lawsuit and the cost of the settlement should be in anyway related to our net worth. Damages are, at least in theory, supposed to be based on the damage done, not based on the offending party's ability to pay. I'm not naive, I recognize that juries, especially for punitive damages, take into consideration how deep the money well they are reaching into is, but that still doesn't account for things like lawyers costs.

Besides, if the argument is really 'how deep is the well' in terms of figuring out the damages then one needs liability insurance not just for one's existing net worth but also for one's future net worth. Courts can and do 'garnish' wages (e.g. take some fixed percentage of all money earned now and into the future) to pay off jury awards. So just covering how much we are worth today isn't enough, we would need to cover how much money we will make from now until we die.

Of course if the lawyers learn we have umbrella insurance they are likely to sue us for even more so they can drain the umbrella insurance and then move on to our actual assets and earnings. So maybe having umbrella insurance is a bad idea since it could just encourage lawsuits?

This all also begs a fundamental question – how likely is it that we will actually be sued? In deciding to cover an event with insurance one needs to take into account both how much the damage might be and how likely the event is to occur. Neither fact by itself is really very useful in determining appropriate coverage. Even in litigious America I have no clue as to how likely we are to be sued and what kind of legal fees and settlement amounts we are likely to be discussing.

Umbrella insurance, I suspect, certainly has a role to play in an insurance portfolio. It covers certain kinds of risks, mostly around "personal injury" we cause to others like liable, slander, physical harm, etc. that isn't covered by our existing home or auto liability insurance.

But still, the whole situation smells bad. Without some idea of what the size and frequency of the risk is (and I can't find a bloody clue in that area) how are we supposed to know how much umbrella insurance, if any, is prudent?

Color me confused.

References