Our loans held-for-investment are accounted for at amortized cost and our loans held-for-sale are accounted for at the lower of cost or fair value, unless we have elected the fair value option for either. The following tables summarize our investments in mortgages and loans as of March 31, 2020 and December 31, 2019 (dollars in thousands):

(1)	Calculated using LIBOR or other applicable index rates as of March 31, 2020 and December 31, 2019 for variable rate loans.

(2)	Represents the WAL of each respective group of loans as of the respective balance sheet date. The WAL of each individual loan is calculated using amounts and timing of future principal payments, as projected at origination or acquisition.

(3)

First mortgages include first mortgage loans and any contiguous mezzanine loan components because as a whole, the expected credit quality of these loans is more similar to that of a first mortgage loan. The application of this methodology resulted in mezzanine loans with carrying values of $968.2 million and $967.0 million being classified as first mortgages as of March 31, 2020 and December 31, 2019, respectively.

(4)	Subordinated mortgages include B-Notes and junior participation in first mortgages where we do not own the senior A-Note or senior participation. If we own both the A-Note and B-Note, we categorize the loan as a first mortgage loan.

(5)	During the three months ended March 31, 2020, $422.7 million of residential loans held-for-investment were reclassified into residential loans held-for-sale.

As of March 31, 2020, our variable rate loans held-for-investment were as follows (dollars in thousands):


	Carrying		Weighted-average
March 31, 2020	Value		Spread Above Index
Commercial loans	$	8,360,382	4.2	%
Infrastructure loans		1,397,404	3.8	%
Total variable rate loans held-for-investment	$	9,757,786	4.1	%

Credit Loss Allowances

As discussed in Note 2, we do not have a history of realized credit losses on our HFI loans and HTM securities, so we have subscribed to third party database services to provide us with industry losses for both commercial real estate and infrastructure loans. Using these losses as a benchmark, we determine expected credit losses for our loans and securities on a collective basis within our commercial real estate and infrastructure portfolios.

For our commercial loans, we utilize a loan loss model that is widely used among banks and commercial mortgage REITs and is marketed by a leading CMBS data analytics provider. It employs logistic regression to forecast expected losses at the loan level based on a commercial real estate loan securitization database that contains activity dating back to 1998. We provide specific loan-level inputs which include loan-to-stabilized-value (LTV) and debt service coverage ratio (DSCR) metrics, as well as principal balances, property type, location, coupon, origination year, term, subordination, expected repayment dates and future fundings. We also select from a group of independent five-year macroeconomic forecasts included in the model that are updated regularly based on current economic trends. We categorize the results by LTV range, which we consider the most significant indicator of credit quality for our commercial loans, as set forth in the credit quality indicator table below. A lower LTV ratio typically indicates a lower credit loss risk.

For our infrastructure loans, we utilize a database of historical infrastructure loan performance that is shared among a consortium of banks and other lenders and compiled by a major bond credit rating agency. The database is representative of industry-wide project finance activity dating back to 1983. We derive historical loss rates from the database filtered by industry, sub-industry, term and construction status for each of our infrastructure loans. Those historical loss rates reflect global economic cycles over a long period of time as well as average recovery rates. However, due to limited information in the first 20 years covered by the database, we have further applied a recessionary multiplier to those historical loss rates as of March 31, 2020 to reflect the current economic deterioration caused by the COVID-19 pandemic which seems to most closely resemble the magnitude of the economic distress of the 2008

financial crisis. We categorize the results between the power and oil and gas industries, which we consider the most significant indicator of credit quality for our infrastructure loans, as set forth in the credit quality indicator table below.

As discussed in Note 2, we use a discounted cash flow or collateral value approach, rather than the industry loan loss approach described above, to determine credit loss allowances for any credit deteriorated loans.

We regularly evaluate the extent and impact of any credit deterioration associated with the performance and/or value of the underlying collateral, as well as the financial and operating capability of the borrower. Specifically, the collateral’s operating results and any cash reserves are analyzed and used to assess (i) whether cash flow from operations is sufficient to cover the debt service requirements currently and into the future, (ii) the ability of the borrower to refinance the loan and/or (iii) the collateral’s liquidation value. We also evaluate the financial wherewithal of any loan guarantors as well as the borrower’s competency in managing and operating the collateral. In addition, we consider the overall economic environment, real estate or industry sector, and geographic sub-market in which the borrower operates. Such analyses are completed and reviewed by asset management and finance personnel who utilize various data sources, including (i) periodic financial data such as property operating statements, occupancy, tenant profile, rental rates, operating expenses, the borrower’s exit plan, and capitalization and discount rates, (ii) site inspections and (iii) current credit spreads and discussions with market participants.

The significant credit quality indicators for our loans measured at amortized cost, which excludes loans held-for-sale, were as follows as of March 31, 2020 (dollars in thousands):




	Term Loans												Revolving Loans		Total		Credit
	Amortized Cost Basis by Origination Year												Amortized Cost		Amortized		Loss
As of March 31, 2020	2020		2019		2018		2017		2016		Prior		Total		Cost Basis		Allowance
Commercial loans:
Credit quality indicator:
LTV < 60%	$	443,347	$	1,261,635	$	874,787	$	1,083,919	$	150,108	$	251,913	$	—	$	4,065,709	$	11,005
LTV 60% - 70%		135,804		928,481		1,448,921		474,098		53,385		169,668		—		3,210,357		25,356
LTV > 70%		159,833		789,538		326,551		127,046		—		60,712		—		1,463,680		14,840
Credit deteriorated		—		—		37,568		7,755		—		105,589		—		150,912		29,853
Defeased and other		—		—		—		—		—		23,303		—		23,303		—
Total commercial	$	738,984	$	2,979,654	$	2,687,827	$	1,692,818	$	203,493	$	611,185	$	—	$	8,913,961	$	81,054
Infrastructure loans:
Credit quality indicator:
Power	$	—	$	257,135	$	307,233	$	120,627	$	199,828	$	249,584	$	23,778	$	1,158,185	$	9,654
Oil and gas		—		154,429		84,790		—		—		—		—		239,219		6,479
Total infrastructure	$	—	$	411,564	$	392,023	$	120,627	$	199,828	$	249,584	$	23,778		1,397,404		16,133
Residential loans held-for-investment, fair value option																274,758		—
Loans held-for-sale																1,175,059		125
Total gross loans															$	11,761,182	$	97,312

As of March 31, 2020, certain first mortgage, mezzanine and unsecured promissory loans with an amortized cost basis of $101.4 million related to a residential conversion project and two subordinated mortgages on department stores with an amortized cost basis of $12.0 million were credit deteriorated and 90 days or greater past due, as were $7.8 million of residential loans. In accordance with our interest income recognition policy, these loans, along with a $37.6 million credit deteriorated loan still current in payment (also related to the residential conversion project), were placed on non-accrual status. We apply the cost recovery method of interest income recognition for all these credit deteriorated loans.

The following tables present the activity in our credit loss allowance for funded loans and unfunded commitments (amounts in thousands):


					Loans
	Loans Held-for-Investment				Held-for-Sale		Total
Three Months Ended March 31, 2020	Commercial		Infrastructure		Infrastructure		Funded Loans
Credit loss allowance at December 31, 2019	$	33,415	$	—	$	196	$	33,611
Cumulative effect of ASC 326 effective January 1, 2020		10,112		10,328		—		20,440
Credit loss provision, net		37,527		5,805		—		43,332
Charge-offs		—		—		(71)		(71)
Recoveries		—		—		—		—
Credit loss allowance at March 31, 2020	$	81,054	$	16,133	$	125	$	97,312


	Unfunded Commitments Credit Loss Allowance (1)
	Loans Held-for-Investment				HTM Preferred
Three Months Ended March 31, 2020	Commercial		Infrastructure		Interests (2)		Total
Credit loss allowance at December 31, 2019	$	—	$	—	$	—	$	—
Cumulative effect of ASC 326 effective January 1, 2020		8,348		2,205		—		10,553
Credit loss (reversal) provision, net		(195)		2,645		1,980		4,430
Credit loss allowance at March 31, 2020	$	8,153	$	4,850	$	1,980	$	14,983
Memo: Unfunded commitments as of March 31, 2020 (3)	$	2,447,330	$	180,829	$	17,261	$	2,645,420

(1)	Included in accounts payable, accrued expenses and other liabilities in our condensed consolidated balance sheet.

(2)	See Note 5 for further details.

(3)	Represents amounts expected to be funded (see Note 21).

Loan Portfolio Activity

The activity in our loan portfolio was as follows (amounts in thousands):


	Held-for-Investment Loans
Three Months Ended March 31, 2020	Commercial		Infrastructure		Residential		Held-for-Sale Loans		Total Loans
Balance at December 31, 2019	$	8,517,054	$	1,397,448	$	671,572	$	884,150	$	11,470,224
Cumulative effect of ASC 326 effective January 1, 2020		(10,112)		(10,328)		—		—		(20,440)
Acquisitions/originations/additional funding		1,089,096		62,929		100,720		646,160		1,898,905
Capitalized interest (1)		36,072		—		—		—		36,072
Basis of loans sold (2)		—		—		(604)		(789,259)		(789,863)
Loan maturities/principal repayments		(689,972)		(37,051)		(48,620)		(20,680)		(796,323)
Discount accretion/premium amortization		11,559		411		—		110		12,080
Changes in fair value		—		—		(25,619)		9,485		(16,134)
Unrealized foreign currency translation loss		(83,263)		—		—		(4,056)		(87,319)
Credit loss provision, net		(37,527)		(5,805)		—		—		(43,332)
Transfer to/from other asset classifications		—		(26,333)		(422,691)		449,024		—
Balance at March 31, 2020	$	8,832,907	$	1,381,271	$	274,758	$	1,174,934	$	11,663,870


					Loans
					Transferred
	Held-for-Investment Loans				As Secured
Three Months Ended March 31, 2019	Commercial		Infrastructure		Borrowings		Held-for-Sale Loans		Total Loans
Balance at December 31, 2018	$	7,075,577	$	1,456,779	$	74,346	$	1,187,552	$	9,794,254
Acquisitions/originations/additional funding		1,005,325		275,079		—		747,265		2,027,669
Capitalized interest (1)		22,137		—		—		—		22,137
Basis of loans sold (2)		(393,556)		—		—		(733,645)		(1,127,201)
Loan maturities/principal repayments		(251,530)		(281,898)		(74,692)		(22,845)		(630,965)
Discount accretion/premium amortization		7,043		137		346		—		7,526
Changes in fair value		—		—		—		11,266		11,266
Unrealized foreign currency translation gain		12,089		—		—		2,119		14,208
Credit loss reversal (provision), net		11		—		—		(774)		(763)
Loan foreclosures		(8,963)		—		—		—		(8,963)
Transfer to/from other asset classifications		46,495		—		—		(46,448)		47
Balance at March 31, 2019	$	7,514,628	$	1,450,097	$	—	$	1,144,490	$	10,109,215

(1) Represents accrued interest income on loans whose terms do not require current payment of interest.

(2) See Note 11 for additional disclosure on these transactions.