Cr4re.com



A Simple Method for Estimating Gross Equity Extracted from Housing Wealth

James Kennedy

August 2010

Abstract

In papers published in 2005 and 2008, Alan Greenspan and I discussed a system we developed that provided estimates of equity extracted from housing wealth. Gross equity extraction (GEE) was the broadest measure of equity extraction in our system. For reasons explained in the paper, the system has not been updated since the end of 2008. This paper offers a method for updating gross equity extraction, one of the most-often requested series, based on publicly available data.

The views presented in this paper are those of the author and do not necessarily represent those of the Federal Reserve or its staff

I. Overview

In papers published in 2005 and 2008, Alan Greenspan and I discussed a system we developed that provided estimates of, among other things, mortgage flows and equity extracted from housing wealth. Gross equity extraction (GEE) was the broadest measure of equity extraction in our system. We also presented estimates of net equity extraction (GEE minus transactions and closing costs associated with equity extraction); and free cash generated by equity extraction.

I continued to update all of the variables from our system through the end of 2008. However, it became increasingly apparent by the middle of that year that many of the assumptions and parameters in the system needed to be re-evaluated. One reason is that we were moving further away from the system’s most recent benchmarks, most of which were estimated on data through 2004. We updated only a few the benchmarks in our second paper. In addition, the turmoil in mortgage and housing markets undermined many of the key relationships in the system. It became apparent by the end of 2008 that the system needed either to be retired or rebuilt completely.

For a number of reasons (mainly time constraints), I deemed that rebuilding the system was not feasible. This paper offers a method for updating gross equity extraction, which is one of the two or three most requested series. Unfortunately, I cannot offer straightforward methods for estimating many of the other variables of interest from our system.

II. Estimate of Gross Equity Extraction

In our 2005 paper, we identified three types of equity extraction: (1) turnover extraction, which results from sales of existing homes;[1] (2) junior liens, both closed-end loans and lines or credit; and (3) cash out refinancing, in which equity the new mortgage exceeds the mortgage retired.

As we showed in our first paper, gross equity extraction equals:

(1) GEE = {(PE - RHS) – (PE – OE)} + {OR – RRE} + ΔHE - RU

Where PE is the price of existing homes, RHS first lien repayments resulting from home sales, OE first lien originations to finance the purchase of existing homes, OR refinance originations, R2 repayments of first liens associated with refinancing, ΔHE is the change in junior liens outstanding, and RU is all other types or repayments or liquidations (charge-offs and unscheduled payments). We showed in our first paper that (1) reduced to

(2) GEE = ΔMDO – ON + A

MDO denotes one-to-four mortgage debt outstanding, excluding construction loans; A denotes repayments of principle resulting from amortization; and ON is originations to purchase new homes. I discuss each of the variables on the right-hand side of (1) in turn.

Mortgage Debt Outstanding

From is from the Federal Reserve’s Flow of Funds Accounts of the United States (FFA). It includes both first liens and junior liens. ΔMDO is the not seasonally adjusted change in total home mortgage debt outstanding (table L.218) excluding construction loans.[2] These mortgages often are referred to as “one-to-four” mortgages because they are used to finance the purchase of structures with from one to four units, or condominiums and cooperatives in structures with five or more units. Line 5 of table 1 shows the change in MDO.

Originations to Purchase New Homes

(2) ON = VN * m * LTP

VN is the value of new homes purchased, m is the mortgaged share of purchases, and LTP is the average loan-to-price ratio of mortgaged purchases (lines 6 to 8 of the table). All of the data pertain to new homes. In our 2005 paper, we adjusted (2) in order to calibrate it to the estimates of total and home purchase originations from our mortgage system. The system did not employ and of the series on the right-hand side of (2) in its estimate of total originations.[3] The average adjustment factor during the 1991-2008 period was about 0.93; that is, the adjustment to new home originations from (2) typically was about -7 percent, or about -$3-1/4 billion. I decided not to incorporate the adjustment in this paper for two reasons. First, it is relatively small (averaging less than $3 billion per quarter from 1991 to 2008). Second, we have no idea how it has changed in recent years.

Various Census publications provide the source data for the value of new homes. The main input is the average value of new, single-family homes sold; we also account for owner- and contractor built homes, homes in structures with two to four units, and condominiums and coops.[4] Included are the types of homes eligible to be financed with a one-to-four mortgage.

Various issues of “The American Housing Survey” and the “Residential Finance Survey” provide the main source data for the mortgaged share of purchases. As shown in line 7, the share averaged about 89 percent and varied little over the 1991-2008 period.

The LTP ratio is a weighted average of the LTP for homes purchased with a conventional mortgage and those purchased with FHA and VA loans. The LTP for conventional purchases is from the Federal Home Finance Board’s “Monthly Interest Rate Survey.” The LTPs for FHA and VA are from those agencies.[5]

Amortization

The formula requires three arguments: the level of mortgage debt outstanding (the unpaid principal balance) and both the average term to maturity and the average interest rate on outstanding mortgages.[6] The average term ‘to maturity is a weighted average of eleven outstanding pools reported by Citicorp: 30-year FRM and ARM pools from Freddie Mac (FRE), Fannie Mae (FNM), and Ginnie Mae (GNMA), respectively, 15-year FRM pools from FRE, FNM, and GNMA, and two non-agency pools, one FRM and the other ARM.[7] The source data are monthly. The weights equal to each pool’s share in the total pool balance. The weighted average coupon rate (WAC) is from the Bureau of Economic Analysis, reported in the table “Mortgage Interest Paid, Owner- and Tenant-Occupied residential Housing.” Lines 12 to 14 show the total payment, interest, and principal.

Gross Equity Extraction

Line 15 shows gross equity extraction, calculated as in (2): line 5 minus line 9 plus line 14. Line 16 shows the estimate of GEE from the most recent update of the mortgage system, based on data through 2008:Q4. As shown in figure 1, the GEE series in lines 15 and 16 are quite similar. On average, the current estimate is $3.2 billion more than the previous estimate; the correlation between the two series is 0.996. Most of the difference between the two series is traceable to new home originations, specifically eliminating the adjustment factor mentioned above. Revisions to the published series on changes in MDO and the different average coupon rate and term variables (mentioned above) also explain part of the difference.

Other key variables from our mortgage system included net equity extraction (GEE net of closing and other transactions costs) and free cash generated by equity extraction (net equity extraction minus the portion used to pay off junior liens and unscheduled mortgage repayments). In addition, we showed estimates of free cash generated by the three types of equity extraction: home sales, cash-out refinancing, and home equity loans (or junior lines). Unfortunately, I am not aware of a straightforward method to estimate any of these series. Nevertheless, the correlation between GEE and net equity extraction from 1991 to 2004 was 0.996. Thus, GEE would be a reasonable proxy for net equity extraction in a regression equation. The correlation between GEE and free cash is smaller, 0.931, in part, because free cash is a much more complex calculation than net equity extraction. Finally, the correlation between GEE and the sum of free cash generated by refis and junior liens (sometimes referred to as “active mortgage equity withdrawal” ) was 0.937 over the 1991-2008 period.

Regarding the three types of equity extraction:

Equity extracted from sales of existing homes in our mortgage system was based on a number of parameters. The parameters in the system have virtually no knowledge of the kind of calamity we have seen in housing markets in recent years, which brings their validity into question.

Equity extracted using junior liens is available in table L.218, line 22 of the FFA (we used the first difference of this series). The FFA does not publish data closed-end and home equity lines of credit, but those series may be downloaded from the Fed’s data download program.[8]

An alternative estimate of the gross cash out from refis is available from Freddie Mac. [9] Freddie’s quarterly series begins in 1993. Although Freddie’s method is similar to ours, there are some important differences. For example, Freddie estimates cash outs from conventional mortgages, whereas ours purported to cover the entire market. In addition, we calibrated our estimate of the gross cash out from refis to the estimate of refinance originations from our mortgage system, which is different from Freddie’s refi originations series. Nevertheless, during the 1993-2004 period, the correlation between the two estimates of cash out refis is quite strong (0.98) considering the differences in purview and source data.[10]

References

Greenspan, Alan and Kennedy, James, “Estimates of Home Mortgage Originations, Repayments, and Debt on One-to-Four-Family Residences,” 2005-41

Greenspan, Alan and Kennedy, James, “Sources and Uses of Equity Extracted from Homes,” Oxford Review of Economic Policy, vol. 24, no. 1, pp. 120-144.

-----------------------

[1] Turnover extraction is net of cash down payments by homebuyers; it is equal to the mortgage taken out by the buyer minus the first lien retired by the seller.

[2] The Fed does not publish the series on construction for one-to-four homes; however, the may be downloaded from the “data download program” at the Fed’s website:

[3] See our 2005 paper for a detailed explanation of the sources we used to estimate both total and home purchase originations. The adjustment reflects discrepancies between the variables in (2) and those used in our mortgage system, which included HMDA home purchase originations and mortgage liquidation rates at depositories and the GSEs. It is not surprising that there is a discrepancy, given the disparate data sources we employed; indeed, we were surprised that the discrepancy cited above, as well as others in our system, were not larger.

[4] The sources from Census include “New Residential Sales,” “New Residential Construction,” and "The Survey of Market Absorption of Apartments.”

[5] Specifically, the weighted LTP is a weighted average of the LTPs for the three types of loans, with the weights based on the value of purchases.

[6] The table is available at h¡G}hÕVjCJOJ HYPERLINK "" . The calculation uses Excel’s “PMT” function.

[7] The average term and average couple rate series used in this paper are different from those we used in our 2005 and 2007 papers. The series we used in our previous papers had somewhat broader coverage than those used here. However, those series are calculated internally at the Fed and based, in part, on data we purchase from vendors. The series used here were chosen because they are publicly available and not much different from those we used before.

[8] See footnote 2 for the web address.

[9]

[10] The correlation from 1993-2008 is lower, at 0.93. I cite the 1993-2004 correlation because our estimate of cash our refis was benchmarked over that period. The correlation between our estimates are Freddie’s is weaker after 2004, probably because of an increase in measurement error of some of the variables in our estimate.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download