Box A: An updated assessment of dairy sector vulnerabilities
This page contains an updated assessment of dairy sector vulnerabilities from the November 2015 Financial Stability Report.
The risk of a substantial rise in non-performing loans (NPLs) in the dairy sector has increased over the past two years, with the existing vulnerability of elevated debt levels being amplified by a large proportion of farmers making operating losses. Using a sample of farm unit records from DairyBase, this box gauges the potential scale of NPLs under hypothetical stress scenarios for the dairy payout and farm land values.1
Banks are currently working with customers under financial stress, and lending to existing customers on the basis of expected profitability under a status quo (SQ), or medium-term, payout that is significantly higher than realised in the 2014-15 and 2015-16 seasons. Demand for working capital has increased substantially, with about half of dairy farms expected to suffer a second consecutive year of operating losses in the 2015-16 season.2 As a result, debt levels have increased by around 10 percent over the past year. Significant demand for working capital is likely to continue while dairy incomes remain below the estimated average break-even payout of about $5.30 per kilogram of milk solids (kgMS).
Loans are likely to be classified as non-performing (with further lending curtailed) for farms where future periods of positive cash flow become unlikely and equity levels are eroded. Farms that have higher debt per kgMS tend to have both higher loan-to-value ratios (LVRs) and breakeven payouts (figure A1). For example, the 20 percent of debt with the highest debt per kgMS has an average LVR of 68 percent and breakeven payout of $5.80. Consequently, the risk of a loan becoming nonperforming will increase particularly rapidly for highly indebted farms during years with a low payout, especially if farm values decline or the SQ payout is revised downwards.
Figure A1: Average break-even payout and loan-to-value ratio by debt per kgMS
Note: Break-even payout is defined as in figure 2.3. Each bucket contains 20 percent of total dairy debt.
Table A1 shows three stress scenarios which provide a metric for assessing the resilience of the sector to lower payout and farm price outcomes. These are not a central forecast for outcomes over the next few years.
Table A1: Farm prices and effective payouts in stress scenarios
|Change in land price (%)|
|Effective payout ($/kgMS)|
Source: RBNZ assumptions.
- Under the base scenario, the effective milk payout is $4.15 in 2015-16 (the current DairyNZ forecast), recovers to $5.50 in 2016-17, and then increases by a further $0.50 per kgMS in the remaining seasons. Farm prices fall by 10 percent in 2015-16.
- Under the severe scenario, the milk payout is $4.00 in 2015-16 and 2016-17, and recovers very gradually thereafter (by $0.50 per kgMS per season). Farm prices are assumed to fall by around 40 percent by 2018-19, consistent with the persistently low milk prices under this scenario. This severe decline is the same as assumed in the joint APRA/RBNZ stress tests in 2014.
- The medium scenario is essentially the midpoint of the base and stress scenarios.
Farm balance sheets are updated after each season to reflect increased working capital required to cover negative cash flow, or any pay-down in debt, and the assumed change in farm value.3 A loan is modelled as non-performing when (i) cash flow in the current season is negative, (ii) the farm has an LVR greater than 90 percent, and (iii) the farm would still make negative cash flow under the SQ payout. The SQ payout is assumed to be $6.25 in 2015-16, and would gradually fall over time if the payout remains very low.4
Under the base scenario, NPLs are estimated to increase to 7.8 percent of sectoral debt (figure A2). Around half of these NPLs materialise in the 2015-16 season, reflecting the one-off decline in land values and the relatively quick recovery in the payout in subsequent years. The rise in NPLs is estimated to be much sharper and more prolonged under the medium and severe scenarios, reaching as high as 44 percent of debt (owed by 25 percent of farms) under the severe scenario. This partly reflects the more marked decline in farm values, which pushes a large number of farmers above an LVR of 90 percent. Another key driver is the muted payout recovery, which results in a sustained increase in working capital borrowing and a decline in the SQ payout.
Figure A2: Modelled NPLs under stress scenarios (% of original exposures)
Source: DairyNZ, RBNZ assumptions (see table A1).
The peak of NPLs under the base stress scenario is higher than the previous peak in early 2011 (see figure 5.7). This is consistent with the fact that the scenario features a more marked and prolonged decline in farm income than was experienced in the post-GFC period. Watchlist loans peaked at 18.2 percent of sectoral exposures, providing some indication of how large NPLs may have become if the post-GFC dairy situation had deteriorated further. This suggests that the modelled NPLs are well within the plausible range of estimates, given the scenario assumptions.
The proportion of NPLs that will eventually result in loan defaults is highly uncertain. Under the assumption that all NPLs result in defaults, the stress testing model can be used to estimate an upper limit for banking system losses.5 Loss rates for the banking system under the three scenarios are estimated to range from 2 to 14 percent of all dairy lending. These losses amount to around 2 to 18 percent of total before-tax profits, and a similar proportion of capital, of the five largest dairy lenders over a typical four-year period, suggesting that they are manageable for the system as a whole. The Reserve Bank has requested that the five largest dairy lenders undertake stress tests of their dairy portfolios, providing an institutional level view of potential losses under similar scenarios. Results are expected to be returned before the end of the year, and will be reported on in due course.
1 A forthcoming Bulletin article will provide a detailed overview of the data and expand upon the analysis contained in this box.
2 This estimate is constructed by updating farm unit records from the 2013-14 season in line with (i) DairyNZ forecasts for effective milk revenue, which differ from the headline payout due to factors like retrospective payments for previous production ($5.70 in 2014-15 and $4.15 in 2015-16), (ii) DairyNZ forecasts for cost containment (average farm working expenses and drawings fall by 90 cents per kgMS over the two seasons), and (iii) an assumption that interest rates on term debt fall by 50 basis points from their 2013-14 levels. Fonterra’s interest free loan is assumed to reduce working capital borrowing from banks by 30 cents per kgMS.
3 All scenarios allow for significant cost containment in 2015-16 in line with DairyNZ forecasts (see footnote 2 for more detail), and costs are assumed to vary positively with the assumed payout in later years. The model also assumes banks recognise only two-thirds of the change in market value of farms in any given year, and the remainder at the end of the scenario horizon. This reflects that valuations tend to lag market prices during periods of stress. The model is similar to Hargreaves, D and G Williamson (2011), ‘Stress testing New Zealand banks’ dairy portfolios’, Reserve Bank of New Zealand Bulletin, 74(2), June.
4 The SQ payout is modelled as a moving average of the five previous payouts and forecasts for the next two seasons, which is broadly in line with banks’ actual modelling.
5 The model assumes that banks face significant costs of disposing of foreclosed assets due to transaction costs, a fire-sale discount, and delays in selling the farm. These assumptions imply banks make losses whenever they foreclose on a farm with an LVR above 75 percent.