STUMP » Articles » Illinois Idiocy: Scenario-Testing the Pension Obligation Bond Idea » 14 February 2018, 18:07

Last week, I set up a basic spreadsheet to project potential cash flows related to the proposed Illinois POB plan.

I used historical data for projection purposes.

This week, I’m using an economic scenario generator hosted by the Society of Actuaries to run a bunch of random scenarios.

And sorry, I’m letting the animated gifs go for this post. It’s just going to be a bunch of graphs.

THE BASIC SET UP FOR SCENARIOS AND PROJECTIONS

The ESG (economic scenario generator) I’m using can be found at the Society of Actuaries site at this link. To keep it simple, I’m taking the version sitting there as of right now, and starting the scenarios at 12/31/2016. I could have updated the info to 12/31/2017, but I don’t want to do that right now (I’m part of the volunteer group that tests the updates and I don’t want to do it as part of my blog. Y’all can wait for the official update with everybody else).

I used two of the fund return results from that ESG: the broad-based U.S. equity index and a long-term corporate bond index. I pulled the scenarios annually, and grabbed 10,000 scenarios. These scenarios are based on a model originally developed by a volunteer group at the American Academy of Actuaries, calibrated to historical scenarios. These are real-world scenarios for cash flow projections (as opposed to risk neutral scenarios for valuation).

I’m using the cash flow model on the liability side as from last week’s post, and based on Prof. Feng’s illustration. I will be comparing against his “baseline” flows. I’m also using his initial balance. I don’t feel like stressing the liability side — but anybody doing a serious analysis for official purposes should look not only on the asset side deviating, but that the liability flows deviate. I could do something there, but it would be cruder than what I’m doing on the asset side.

I wrote VBA code to run through the 10K scenarios in the spreadsheet model, which you can grab for yourself here. I did not make it user-friendly. If you know what you’re doing, open up the VBA editor, and you can run the macro for yourself. If you have questions, you can email me at marypat.campbell@gmail.com.

There is one input: what percentage of the asset allocation goes to the broad-based U.S. equity. That’s it.

RESULTS

I started out with a 60% equity/40% bond allocation, like I did last week.

And then I projected the funds through the scenarios, and graphed the resulting percentiles.

Here’s the graph:

Let me interpret this for you — seek out the black dotted line. That’s the “baseline” from the presentation – that if everything goes as planned, that’s what the funds would look like.

Note the expectation that funds would rise by the end. Yay.

But what’s this? My median projection (red dotted line) has much worse results. Some of the lower projections even take the fund into failure territory — i.e., it goes negative. The percentile that comes closest to the baseline projection is the 75th percentile — meaning only the top 25 percent of the scenarios do as well or better. That’s not terribly comforting.

I decided to try the two endpoints of allocations as well.

Here’s the curves with 0% equity… so it’s all in bonds:

That doesn’t look too good.

And here it is 100% in equity:

Notice that the median is in line with the baseline projection for the 100% equity. But look at the resulting variation in results.

FAILURE CURVE

One thing I want to capture is at what probability level failure occurs. And note: my definition of failure is the fund amount going negative. That’s bad.

You start getting non-zero probability of failure about 20 years into the projection. And for the 100% bond allocation, it escalates rapidly after that. The other two have more moderate failure rates, but it’s definitely a lot more than zero.

VARIABILITY IN RESULTS

Finally, here is a graph of the standard deviation of the fund balance at each year.

Again, the 100% equity allocation provides a median result closest to the projected baseline. But it also has the most variability in results — one can focus on the upside, but that standard deviation also captures the downside risk. I could do a one-sided variance measure, but I’m not right now.

MORE INVOLVED ANALYSIS NEEDS TO BE DONE

Note, I’m doing something extremely simplistic here. But Prof. Feng also did a very simplistic model. That’s why I could pick it up and do these things, after all.

There are many practical problems with issuing $100 billion in POBs when you’re a state in such a hole like Illinois. If they even get closer to trying to do this, I’m sure many bond people will come out of the woodwork to explain what price they may actually get. I don’t feel like stressing that aspect.

There is the liability-side issue of the pension contribution requirements being other than what I projected. That’s the complicated part. I would hope that the practical issues of issuing that much in bonds at a price that will likely turn out to be disadvantageous would head that off at the pass, but hey. It’s Illinois. I’m sure somebody has money to blow to continue this awful idea…for now.

But this is something I did very simple, just trying out the asset-side assumptions. I didn’t do anything fancy. The issue with this kind of move is that more leverage is injected into the system, meaning that while you may have improved survival rates, if it goes bad, it can make things much much worse.

The “extreme scenario” that Prof. Feng projected wasn’t extreme at all. It just used a short portion of Illinois’s actual experience with asset returns, and then kept returns level at something a little bit under expectations.

I just did something based on real-world-calibrated economic scenarios, and the result isn’t nearly so rosy. For something this big, you need to at least put a little more work into it to quantify the risks and rewards.