Cross-posted from the LF Economics Blog:
The most recent bout of irrational exuberance provides an opportunity to examine the financial fundamentals underpinning the housing market. Housing prices are escalating as the mortgage debt accelerator rises; the former lagging the latter by two quarters.
In contrast, rents are experiencing low growth. According to CoreLogic’s asking rental price series, nationwide annual growth in rents turned negative in 2016 while the ABS’ established rent series shows rents have slowed to the lowest pace since the early 1990s recession, the worst economic downturn in post-WW2 history.
Additionally, mortgage interest rates are now rising, due in part to banks’ funding costs increasing and to stem APRA from imposing strict macroprudential controls. APRA’s ‘wet lettuce leaf’ approach to slowing the growth of Australia’s gigantic private debt bubble, centred primarily in the household sector, by capping interest-only mortgage issuance, will do little to keep household debt from rising.
While there is plenty of coverage of the latest movements in housing prices, a black hole exists in terms of a more thorough analysis of the financials of housing. Using the latest data, the following table attempts to summarise the financials of investment property. Some interesting aspects emerge.
Data and assumptions:
- Median house and unit prices and yields are based on CoreLogic’s latest monthly release (2017M04);
- Investment properties tend to be less expensive than owner-occupied dwellings, so a 10% price markdown has been incorporated;
- Investment yields tend to be higher than for owner-occupied dwellings (RBA paper 2014). This likely mirrors the more generous tax treatment for homes than investment properties and lower quality of the latter. Hence, the yield has been marked up by 100bps to reflects these differences. The estimated annual rents approximate those from the CoreLogic rental report;
- Running expenses as a proportion of annual gross rental income is based on ATO data collected from property schedules;
- The mortgage interest rate is the housing investor discounted rate from the RBA. This is a generously low rate, given the aggregate imputed interest rate on the household debt stock is 5.73% as of 2016Q4 and mortgage interest rates are currently rising;
- Mortgages are P&I, amortising, 25 years, with monthly payments;
- Dwellings are purchased with a standard 20% cash deposit, hence 80% LVR; and
- The mean income tax rate is assumed to be 30%, reducing to 15% with the application of the 50% capital gains tax (CGT) discount.
Even with low mortgage interest rates and 20% deposits, there is no city where investors can be positively geared at the median price. This is not surprising, though, as ATO tax statistics reveal that two thirds of property investors are negatively-geared (NG). Also, the data reveals the entire investor dwelling stock is negatively-geared, on aggregate.
Net rental income losses are predictably greatest in absolute terms in Sydney and Melbourne where prices are highest and yields lowest. This explains why government and the FIRE (financial, insurance and real estate) sector bitterly resist reforms to the negative gearing tax concession given it defrays part of those losses.
Yields are paltry relative to terms deposits delivering superior yields up to 3.3%, and given there are no fees or expenses, gross and net yields are identical without the default, liquidity and transaction risks inherent with real estate. Given the compressed yields and large net rental income losses, it becomes clear the only incentive to invest is the potential for capital price growth. This is actually a misnomer, given it is the rise in land price, not the structural or capital component, that delivers the gains.
The ‘loss cover growth rate’ indicates how much the median property has to rise in nominal terms on an annual basis to cover the net rental income loss after the NG deduction has been applied. The loss cover has been applied to the total price, not the amount covered by debt, which accounts for the interest forgone as well as interest paid (even though the applicable interest rates for the two components are different). Also, principal payments are not incorporated into the loss cover growth rate as it is calculated before the payment of principal, with owners able to recoup payments upon sale.
Where the net rental income losses are small relative to the median price (i.e. Hobart and Darwin), the loss cover growth rate is lowest. In Sydney, Melbourne, Brisbane and Canberra (the latter two have unusually high expenses relative to rent), the rate is higher to cover the largest income losses.
After the application of the discounted CGT, the loss cover growth rate inevitably rises to cover the liability. Finally, this adjusted rate can be compared to the annual change in nominal prices. The considerable prices rises in Sydney, Melbourne and Canberra easily cover the adjusted rate, while it is a mixed story elsewhere. Financially, houses have performed better than units.
That all cities are experiencing net rental income losses is significant in that the justification for present prices (given current interest rates, rents and expenses) depend upon the assumption that prices have to continually rise in order to cover the income losses. To justify the price paid, it is not good enough to simply make a capital gain (at least on paper); one must make a certain minimum capital gain in order to cover the interest losses – let alone other incurred fees and opportunity costs.
Eliminating the CGT discount would heighten the adjusted loss rate, which like NG reforms, is opposed by the government and FIRE sector. Removal of this tax concession, however, would result in lower price growth and hence smaller net rental income losses over a period.
The data explains Australia’s addiction to interest-only loans which has reached endemic proportions, both in terms of stock and flows (approvals). The two primary reasons why IO loans are attractive to investors is that 1) payment of interest but not principal means a larger loan can be secured against a given household income; and 2) interest, not principal, is tax deductible.
APRA has recently tightened ADIs’ ability to issue IO loans, as they assist with inflating residential land prices. In other words, they facilitate what amounts to Ponzi lending. On this basis, regulators ought to ensure existing IO loans be converted to P&I and ban future issuance altogether. An understanding of Hyman Minsky’s definition of an asset bubble would assist to confirm that IO loans have no place in the mortgage market.
The proportion of IO loan approvals relative to total approvals reached a stratospheric 46% in 2015, before APRA’s intervention caused a decline. In contrast to the Australian situation, Ireland, which experienced a monstrous housing bubble, IO loan approvals peaked at 15% of total approvals in 2007. In the US, IO loan approvals reached 23% in 2005.
The size of the deposit needed for the median investment property to be neutrally-geared upon purchase (assuming P&I mortgages) is 74% of the median nationwide house price, equating to an LVR of 26%. It is a little less imbalanced for units.
The following table presents prices derived from the simple capitalisation of rents. The results demonstrate Sydney and Melbourne are significantly overvalued, while houses in the rest of the cities, including all units, are fairly valued or undervalued. The latter results likely stem from using an unusually low mortgage interest rate.
The valuation can also be derived by comparing the difference between the mortgage IR to the gross yield (the Speculative Index). In effect, the analysis equates negative gearing (on gross yield assuming 100% LVR) with overvaluation.
This analysis may be considered too simple, as what needs to be capitalised is the growing (future) stream of rents, or rents plus the equivalent growth in capital prices. This becomes increasingly circular, however, as overvaluation based upon current market prices experiencing extreme growth can cover the income losses generated, thereby justifying the apparent overvaluation.
If net yields are used (gross yields after running expenses but before debt payments) to account for the unavoidable costs property investors incur, the overvaluation increases considerably. Interestingly, the estimated overvaluation for Sydney and Melbourne is similar to the growth in real housing prices between the trough in 1996 and peaks reached today.
Accordingly, one can take the view that prices, especially house prices in Sydney and Melbourne, are significantly undervalued because the extraordinary price growth has delivered such large returns (on paper). A more complicated approach to investigating the financials of housing is via the user cost model, which factors in current and assumed future movements in prices, property taxes, interest rates, running expenses and risk premiums.
This is what the RBA attempted in 2014 with its analysis of the housing market. The authors claimed that in 2015, nationwide property prices were actually undervalued by 30%, with that undervaluation continuing through to 2016. They were honest enough to note that similar user cost estimations in the US provided anomalous results.
In the US during the 2000s (and before the GFC), economists who denied prices represented a housing bubble (let alone their largest housing bubble on record) favoured the user housing cost model to demonstrate prices were in fact fairly valued or even undervalued. In the model, price rises are considered gains to owners, and a larger price rise – increasing the spread between the risk free rate and the returns from housing – actually increases the undervaluation. In other words, the user cost of housing fell further below the user cost of renting.
This led to the strange outcome that the most bubble-inflated cities like San Francisco and Miami were considered the most undervalued, whereas the less inflated cities in the Mid-West were fairly valued. These studies are available in the journal databases and can also be found online; they are worth reading through as good examples of how housing economists got their research wrong.
The user cost model would indicate that in 2016 – depending on the assumptions of future movements in variables, of course – housing prices were significantly undervalued. The following graph of long-term housing prices demonstrates a remarkable escalation from 1996 onward; readers can decide for themselves whether prices are undervalued as of 2016.
So what does the data indicate about the Australian housing market? The results suggest that even using a generously low mortgage interest rate, property investment is entirely focused upon capital (land) price speculation, not income generation. Using a more realistic mortgage IR in the range of 5.0% to 5.5% would significantly amplify the imbalance between prices and incomes.
Clearly there is something wrong when, at the median and aggregate, an asset class cannot produce positive net incomes to pay down the debt used to purchase said assets (Minsky’s definition of a bubble). These income losses, however, are easily overwhelmed given current and future expected price growth. From one perspective, housing is undervalued and will continue to become more so as price growth continues to rise, as the latest data demonstrates.
The present course is sustainable for as long as price growth maintains its current trajectory. Herein lies the problem: what is not sustainable is the rise in debts used to purchase housing. Eventually debt accumulation hits an endogenous limit where it can rise no further. Price growth stalls, and begins to decline. If this occurs for an extended period, the loss cover growth rate persistently falls behind what is needed to cover the net rental income losses and the debt-fuelled Ponzi scheme implodes.
The BIS database indicates Australia has the second-highest household debt to GDP ratio, the third-highest debt service ratio (DSR) and various OECD reports show we have a household debt to income ratio that ranks Australia around 4th to 7th highest. Given the BIS already has internationally-compatible household debt and household income series in their database, the organisation will eventually create a global debt to income ratio database.
Unfortunately, as one can analyse the housing market using ten different models, twenty different variables and thirty different parameter specifications, the results will be various and contradictory. Recent journal papers examining whether our housing market represents a bubble demonstrates this: “the national market is a bubble”, “not a bubble”, “Sydney is, but not the rest”, “some cities for some periods were a bubble but are currently not” and on it goes.
For as long as prices and debt continues growing, housing is undervalued from a mainstream neoclassical perspective. From a heterodox Minskyian or Post-Keynesian view which treats private debt and banking seriously, however, the housing market is in the midst of a gargantuan asset bubble.