Interesting. So unless you are setup as a trading company, you could not account for your capital losses? I.e You would be paying too much tax
Your Harmoney investment will inevitably lead to paying too much tax. All Harmoney "notes" have a default % attached to them. Depending on the riskiness of the notes you invest in, you can expect to have some of your loan notes remaining unpaid in an average year. So over the long run, your Harmoney investment is income generative and capital depletive!
I have looked on the Harmoney website to see how it fits into the financial arrangement scheme - but could find nothing. I sent their investor helpline an email a few days ago - but no response.
It will be interesting to see what they provide for end of tax year. Especially in relation to right offs - bacause to the extent you have earned interest on those loans, that should come off your capital loss.
Have built a bit of VBA to run a modest Monte Carlo simulation to generate a range of defaults and gross returns.
The simulations maintained a close proximity to Harmoney's advertised default rates and timing yet still generated a wide variance in default values. It's a cautionary view for those "testing the water" with a small sum of money - it is essentially gambling on being fortunate.
The results also generated a simulated return 1% below the return Harmoney suggest should be achieved. I could have a logical or modelling method error somewhere but haven't found anything yet so am slightly suspicious of their methodology.
Interesting figures. I think Harmoney have said that over time they may fine tune their default rate figures. Agree - Harmoney investments are a gamble with potentially high interest rates along with the potential for significant capital depletion.
As far as the tax accounting for the capital depletion, Harmoney FAQ states underForecast Annual return "This is the forecast return of the notes in your portfolio on an annual basis. It is calculated by taking forecast annual net income and dividing it by the outstanding principal. Forecast annual net income is interest less defaults and service fees. Defaults are calculated as the percentage of outstanding principal that is forecast to be written off as a bad debt over the next 12 months. "
The emphasis is mine. So it appears that Harmoney's net income calculation includes forecast bad debts. However whether taxable income allows a deduction even for actual bad debts, I do not know. I am more sure returned Taxable income would not allow for forecast bad debts or doubtful debts. It seems you need to refer to the law on financial arrangements, which is voluminous.
That's a very interesting simulation. I have a few questions for you.
I assume you ran this simulation across your own portfolio.
Is the wide variance in default values for grades E-F due to the small sample size?
Do you think the wide variance on default values would drop if an investor increased their investment perhaps to over 400 notes?
Why your default model slightly difference to Harmoney's? Did you use some software to read the default model chart from their site?
The XIRR yield is 18.1% from your simulations. What is the 'Net Yield' and 'Net Yield loss deductions' and why is it lower than the XIRR yield?
Very nice graphics, what software did you use to produce these?
Originally Posted by Halebop
Have built a bit of VBA to run a modest Monte Carlo simulation to generate a range of defaults and gross returns.
The simulations maintained a close proximity to Harmoney's advertised default rates and timing yet still generated a wide variance in default values. It's a cautionary view for those "testing the water" with a small sum of money - it is essentially gambling on being fortunate.
The results also generated a simulated return 1% below the return Harmoney suggest should be achieved. I could have a logical or modelling method error somewhere but haven't found anything yet so am slightly suspicious of their methodology.
If "bad debts" are not allowed to be taken into account for tax purposes then you will end up with a tax rate greater than 33% on the Harmoney definition of "annual net income".
As a matter of interest, for those with an investment with Harmoney, on average, what percent of your investment is in the Harmoney Investor Account (earning zero interest) compared to the proportion of your outlay invested in interest-earning "notes"?
It is probably much less than this in actuality, but if half your money with harmoney is sitting in the 0% interest account at any one time, then your xirr yield on your notes will need to be halved to get the yield on your total Harmoney outlay.
I assume you ran this simulation across your own portfolio.
Yes, my wife's portfolio as of a week or so ago.
Originally Posted by newtrader
Is the wide variance in default values for grades E-F due to the small sample size?
The entire portfolio is small so there is wide variation across the whole thing, it's just that a 100% movement on defaults for A1 notes will barely register because the base default rate is so low. But yes, to control this would need either a bigger sample or more simulations than the 300 per note I ran.
Keep in mind though that by running 300 simulations the sample is 193x300 or 57,900 data points - any single simulation is random but in aggregate this won't be too bad.
Originally Posted by newtrader
Do you think the wide variance on default values would drop if an investor increased their investment perhaps to over 400 notes?
You would certainly expect to see the core results narrow (particularly between the 25th and 75th percentiles). I've run the basic model dozens of times and the median point for default value has varied between 5.2% and 5.6%. My instinct is that a sample size of 400 would not get you close enough to certainty but it's better than 193! Remember though these are multiple simulations - in real life you only get to push run once.
Originally Posted by newtrader
Why your default model slightly difference to Harmoney's? Did you use some software to read the default model chart from their site?
I just copied their default by credit rating stats to a local table and replicated as best I was able their default time-line chart (this was an image on the Harmoney site, not a table).
The difference could just be natural statistical variation - I ran 300 random simulations using the default rates as business rules for comparison with randomly generated numbers. You would not expect this to be a perfect match. However, my aggregate result in both default rate (how many notes defaulted) and default timing (which month they defaulted) remained very close to Harmoney's model. So I was a little surprised (but not panicked) in the 1% difference in projected returns. It could just be down to differences in calculating the yield - I used the Excel XIRR method, there are alternative approaches.
Originally Posted by newtrader
The XIRR yield is 18.1% from your simulations. What is the 'Net Yield' and 'Net Yield loss deductions' and why is it lower than the XIRR yield?
This aligns to the current tax debate on the thread. I wasn't sure if some/all loan losses would be deductible. So the net yield just assumes 28% tax on the return before loan losses and the loss deduction version allows these to be expensed before 28% tax is calculated. The difference is not far off the projected British experience - they are allowing loan loss deductions in a couple of months and expect net investor returns to benefit by about +25%.
Originally Posted by newtrader
Very nice graphics, what software did you use to produce these?
Just Excel. The general design approach (type of charts, monochromatic, stripped bare/little framing etc) is current best practice thinking on data visualisation.
Just Excel. The general design approach (type of charts, monochromatic, stripped bare/little framing etc) is current best practice thinking on data visualisation.
Would you consider sharing your model (with any sensitive data removed)? I haven't done that sort of modelling in Excel, so interested to see your approach...
Would you consider sharing your model (with any sensitive data removed)? I haven't done that sort of modelling in Excel, so interested to see your approach...
Happy to share although I'm not pretending the model itself is best practice.
If you don't mind give me a couple of weeks to improve the speed of the code; I built the VBA one feature at a time and now it runs quite slow. This will become a big problem when there are many hundreds or thousands of notes. It just needs a little bit of thought to improve...
Bookmarks