3. Goodwill

Our goodwill balance was $37.6 million and $134.5 million as of September 30, 2016 and December 31, 2015, respectively. We apply an accounting standard under which goodwill has an indefinite life and is not amortized. Goodwill is tested for impairments at least annually, or more frequently whenever an event or change in circumstances occurs that would more likely than not reduce the fair value of a reporting unit below its carrying amount. We test goodwill for impairment at the reporting unit level, which is at the project level and, the lowest level below the operating segments for which discrete financial information is available.

In the third quarter of 2016, we performed an event-driven goodwill impairment test. While declining power prices have been observed over the past two years, we identified a significant decrease in the long-term outlook for power prices in the regions where our reporting units operate in the third quarter of 2016. Because the estimated future cash flows of our reporting units are sensitive to fluctuations in forward power prices and these prices are the most impactful input in calculating a reporting unit’s fair value, we determined that it was appropriate to perform an event-driven impairment test. For two of our reporting units (Morris and Nipigon) we performed a qualitative assessment and concluded that it was likely that the fair values significantly exceed the carrying values. These reporting units have aggregate goodwill of $6.9 million and have PPAs with significant remaining time before their expiration and are not significantly impacted by the decrease in the long-term outlook for power prices.

The other five of the reporting units tested (Curtis Palmer, Mamquam, North Bay, Kapuskasing and Moresby Lake) failed step 1 of our quantitative two‑step test. Because five reporting units failed step 1 of the two-step goodwill impairment test, we identified a triggering event and initiated a test of the recoverability of their long-lived assets. The asset group for testing the long-lived assets for impairment is the same as the reporting unit for goodwill impairment testing purposes. In order to test the recoverability of the assets in the asset groups, we compared the carrying amount of the assets to estimated undiscounted future cash flows expected to be generated by the asset group. The carrying value of each asset group includes its recorded property, plant equipment, intangible assets related to PPAs and goodwill. Of the five asset groups tested, the North Bay and Kapuskasing asset groups (Canada segment) failed the recoverability test and we recorded property, plant and equipment impairment charges aggregating $5.9 million for the periods ended September 30, 2016. For these asset groups, we estimated their fair value utilizing an income approach based on market participant assumptions. These assumptions include estimated cash flows under the remaining period of their respective PPAs.

Subsequent to recording long-lived asset impairments, we performed the step 2 goodwill impairment test and recorded a $50.2 million full impairment at the Mamquam reporting unit, a $15.4 million partial impairment at the Curtis Palmer reporting unit, a $6.5 million full impairment at the North Bay reporting unit, a $6.7 million full impairment at the Kapuskasing reporting unit and no impairment at the Moresby Lake reporting unit for a total goodwill impairment charge of $78.8 million for the periods ended September 30, 2016. At the time of their acquisition in November 2011, the fair value of the assets acquired and liabilities assumed for the Mamquam and Curtis Palmer reporting units were valued assuming a merchant basis for the period subsequent to the expiration of the projects’ original PPAs. The forecasted energy revenue on a merchant basis, in the respective markets in which those plants operate, was higher than the energy prices currently forecasted to be in effect subsequent to the expiration of the reporting unit’s PPA. Power prices, in the respective markets in which those plants operate, have declined from 2011 and from the dates of our previous impairment assessments due to several factors including decreased demand, lower oil prices and lower natural gas prices resulting from an abundance of shale gas. Our forecasts for discounted cash flows also reflect a higher level of uncertainty for re‑contracting at prices than were previously forecasted in 2011. The decline in forward power prices for British Columbia since our last goodwill impairment performed as of November 30, 2015, in particular, had a significant impact on the estimated discounted cash flows of our Mamquam reporting unit and was the primary driver for its recorded goodwill impairment. British Columbia’s peak demand outlook has declined primarily attributable to a reduction in forecasted liquefaction build and need in the region and the associated loss of power demand. The resulting drop in the peak demand reduces the amount of needed capacity and therefore the capacity prices also were reduced. Furthermore, the PPA at the Curtis Palmer reporting unit expires at the earlier of December 2027 or the provision of 10,000 GWh of generation. Based on Curtis Palmer’s cumulative generation through the date of the goodwill impairment test, we anticipate the PPA expiring two years before December 2027. As a result, the discounted cash flow model for Curtis Palmer utilizes forward power prices for that two-year period that are substantially lower than the prices under the current PPA.

The long-lived asset and goodwill impairment charges were recorded in the third quarter of 2016 and not earlier in the fiscal year because we did not identify any triggering events that would have required an event-driven impairment assessment. While declining power prices have been observed over the past two years, the significant decrease in the long-term outlook for power prices in the regions where our reporting units operate identified in the third quarter of 2016 had the most significant impact to the key inputs to our long-term forecasted cash flow models. Additionally, the PPAs at our North Bay and Kapuskasing reporting units expire on December 31, 2017. As these projects approach the expiration date, the remaining estimated contracted future cash flows decrease.

We determine the fair value of our reporting units using an income approach with discounted cash flow (“DCF”) models, as we believe forecasted cash flows are the best indicator of such fair value. A number of significant assumptions and estimates are involved in the application of the DCF model to forecast operating cash flows, including assumptions about discount rates, projected merchant power prices, generation, fuel costs and capital expenditure requirements. The undiscounted and discounted cash flows utilized in our long‑lived asset recovery and step 1 and 2 goodwill impairment tests for our reporting units are generally based on approved reporting unit operating plans for years with contracted PPAs and historical relationships for estimates at the expiration of PPAs. All cash flow forecasts from DCF models utilized estimated plant output for determining assumptions around future generation and industry data forward power and fuel curves to estimate future power and fuel prices. We used historical experience to determine estimated future capital investment requirements. The discount rate applied to the DCF models represents the weighted average cost of capital (“WACC”) consistent with the risk inherent in future cash flows of the particular reporting unit and is based upon an assumed capital structure, cost of long‑term debt and cost of equity consistent with comparable independent power producers. The betas used in calculating the WACC rate were obtained from reputable third-party sources. We utilized the assistance of valuation experts to perform step 1 and step 2 of the quantitative impairment test for several of our reporting units. The fair value that could be realized in an actual transaction may differ from that used to evaluate the impairment of goodwill.

The valuation of long-lived assets and goodwill for the impairment analyses is considered a level 3 fair value measurement, which means that the valuation of the assets and liabilities reflect management’s own judgments regarding the assumptions market participants would use in determining the fair value of the assets and liabilities. Fair value determinations require considerable judgment and are sensitive to changes in these underlying assumptions and factors. As a result, there can be no assurance that the estimates and assumptions made for purposes of a goodwill impairment test will prove to be accurate predictions of the future. Examples of events or circumstances that could reasonably be expected to negatively affect the underlying key assumptions and ultimately impact the estimated fair value of our reporting units may include macroeconomic factors that significantly differ from our assumptions in timing or degree, increased input costs such as higher fuel prices and maintenance costs, or lower power prices than incorporated in our long-term forecasts.

The following table is a rollforward of goodwill for the nine months ended September 30, 2016:


		December 31,				Translation		September 30,
Reporting unit	Segment	2015		Impairment		adjustment		2016

Curtis Palmer	East U.S.	$	44.5	$	(15.4)	$	—	$	29.1
Morris	East U.S.		3.3		—		—		3.3
Kapuskasing	Canada		8.8		(6.7)		(2.1)		—
Mamquam	Canada		64.4		(50.2)		(14.2)		—
Moresby Lake	Canada		1.6		—		—		1.6
Nipigon	Canada		3.6		—		—		3.6
North Bay	Canada		8.3		(6.5)		(1.8)		—
		$	134.5	$	(78.8)	$	(18.1)	$	37.6