All privately owned housing units started or authorized in permit-issuing places are treated as nonfarm housing even though a negligible number are located on farm properties. All publicly owned housing units started are treated as nonfarm housing.

STARTS BY TYPE OF STRUCTURE

A total of 14 different sets of starts rates that change from month to month are employed to calculate number of housing units started by type of structure in permit places. Eight sets of starts rates are used for one-family houses: separate sets of rates for metropolitan and nonmetropolitan areas within each of the four regions. For structures with 5 units or more, separate sets of starts rates are used for each of the four regions. Single sets of starts rates are used for all regions for structures with 2 units and for structures with 3-4 units.

Starts by type of structure in nonpermit areas are calculated directly in the estimating procedure described above. Less than 5 percent of the units started in nonpermit areas in 1976 were in buildings with 2 housing units or more.

BUILDING PERMITS

Data on housing units authorized by local building permits relate to the time of issuance rather than to the actual start of construction. They do, however, provide some indication of activity in residential building in advance of the start of actual construction. Although construction is started on most residential buildings in the same month in which the permit is issued, several months may pass before start of construction.

The 14,000 areas with local building permit systems for which figures are currently given in this report account for a major portion of residential building in the United States. For the country as a whole, approximately 81 percent of the private housing units were constructed in permit-issuing places in 1976.

The monthly series and 1973-1976 annual data reflect the 14,000 places in the United States which were identified in 1972 as having local building permit systems. Annual data for both the 13,000 and 14,000 places are shown for 1971 and 1972. Annual data for 1968 through 1970 reflect the 13,000 places identified in 1967 as having local building permit systems.

Basically, the procedure followed in arriving at the monthly building permit authorization totals involves the cumulating of monthly data from all permit-issuing places that authorized 50 housing units or more (20 or more in some states) in a recent year along with estimates for the less active places based on a stratified probability sample of these places.

A more detailed discussion of the uses and limitations of building permit data is provided in the Census Bureau's monthly report on Building Permits (C40).

MOBILE HOME SHIPMENTS

Statistics on manufacturers' shipments of mobile homes are provided by the Manufactured Housing Institute and include estimates for firms not associated with the MHI.

The total of new housing starts and manufacturers' shipments of mobile homes is shown because an addition to the housing supply is made by mobile homes as well as by the construction of new housing units. Some of the mobile homes shipped by manufacturers, however, as well as some of the new housing units started, are used as seasonal homes and second homes and do not add to the supply of housing units occupied as usual places of residence. Furthermore, some are used for nonresidential purposes. The number of mobile homes or new housing units used in these ways is not now known.

SAMPLING VARIABILITY

The various estimates of private residential housing units started and private housing units authorized by local building permits, which are shown in this publication, are based on samples and may differ from statistics which would have been obtained from a complete census using the same schedules and procedures.

An estimate based on a sample survey is subject to both sampling error and nonsampling error. The accuracy of a survey result is determined by the joint effect of these errors. Sampling error reflects the fact that only a particular sample was surveyed rather than the entire population. Estimates of size of the sampling errors are provided by the standard error of the estimates. Nonsampling errors can be attributed to many sources: inability to obtain information about all cases in the sample; definitional difficulties; differences in interpretation of questions; inability or unwillingness of respondents to provide correct information; and errors made in processing data. No explicit measures of the effects of nonsampling errors are available; however, it is believed that most of the important response and operational errors were detected in the course of reviewing the data for reasonableness and consistency. As derived for this report, the estimated standard errors include part of the effect of nonsampling errors but do not measure any systematic biases in the data.

The particular samples selected for the Housing Starts and Building Permits surveys are each one of a large number of similar probability samples that, by chance, might have been selected under the same specifications. Estimates derived from the different samples would differ from each other. The standard error or sampling error of a survey estimate is a measure of the variation among the estimates from all possible samples and thus is a measure of the precision with which an estimate from a particular sample approximates the average results of all possible samples.

Estimates of the standard errors have been computed from the sample data for selected statistics in this report. They are presented in the form of relative standard errors, the standard error divided by the estimated value to which they refer.

The sample estimate and an estimate of its standard error ermit us to construct interval estimates with prescribed onfidence that the interval includes the average result of all ossible samples (for a given sampling rate). For example, uppose table 1 of this report showed an estimated 80,000 nits in 1-unit structures were started in a particular month. 'urther, suppose that the relative standard error of this stimate is 4 percent. Multiplying 80,000 by .04, we obtain 1,200 as the standard error. This means that we are onfident, with two chances out of three of being correct, hat the average estimate from all possible samples of >ne-unit structures started during the particular month is between 76,800 and 83,200 units. Doubling the interval gives is limits of 73,600 and 86,400, increasing our confidence of being correct to 19 out of 20 times that the interval contains he average value over all possible samples. We would have almost certain confidence by tripling the interval. The average estimate of one-family units started during the specified month may or may not be contained in any one of these computed intervals; but for a particular sample, one can say that the average estimate from all possible samples is included in the constructed interval with a specified confidence; that is, two chances out of three, etc.

Relative standard errors shown in tables 1 and 3 were computed on new private housing units started, since statistics on public housing units started are not based on samples. The relative standard errors in tables 1, 3 and 6 may be applied to the corresponding seasonally adjusted estimates in tables 5, 3 and 6 respectively.

SEASONAL ADJUSTMENTS

Seasonally adjusted estimates of housing starts are the actual number of housing units started in a month adjusted to remove the normal seasonal movement. This adjustment takes into account month-to-month variations resulting from normal or average changes in weather conditions, from differences in the lengths of the months and from differences in the composition (trading-day variation) of the months. The purpose of this seasonal adjustment is to bring out underlying cyclical trends.

The seasonally adjusted annual rate is the seasonally adjusted monthly rate multiplied by 12. The seasonally adjusted annual rate for a particular month, such as July, can be interpreted to mean that if the only changes which occur in housing starts from July through June of the following year were the normal seasonal changes described by the seasonal indexes, then the total starts in that 12-month interval would equal the seasonally adjusted annual rate for July. The seasonally adjusted annual rate has the advantage of facilitating comparisons with previous annual start figures as well as with the seasonally adjusted annual rates for prior months. The seasonally adjusted annual rate is, of course, neither a forecast nor a projection; rather it is a description of the rate at which housing starts are proceeding in the particular month for which it is calculated.

It should be noted that the seasonal indexes adjust for average monthly weather conditions among other normal

seasonal influences but not for year-to-year variations in weather or in other seasonal influences within the same month. Attempts to adjust the monthly housing starts figures for unusual weather conditions have thus far been unsucessful. A summary of research to date on this subject was given in an article which appeared in the August 1968 issue of Construction Review.

The seasonal factors for Private Housing Starts and Building Permits were developed using the X-11 version of the Census Method II. A description of the X-11 version appears in Bureau of the Census Technical Paper No. 15, "The X-11 Variant of the Census Method II Seasonal Adjustment Program."

Housing Starts

Seasonal adjustments for Private Housing Starts for the months covered in this report are based on data for the period August 1963-December 1976. Seasonal indexes were developed for total private housing starts (including farm) by region and type of structure within each region, e.g., single family units in the Northeast. The seasonally adjusted total is the sum of 12 seasonally adjusted components-3 structure types in each of the 4 regions. The seasonal indexes for Private Housing Starts shown in table A-1 include tradingday factors which were estimated internally by the regression routine.

Building Permits

Seasonal adjustments for Building Permits for the months covered in this report are based on data for January 1960-March 1976. Seasonal factors were developed separately for permits issued by geographic region and type of

structure.

Building permits data contain significant trading-day variation and are therefore adjusted for this variation in order to reduce irregular fluctuations in the series. The daily pattern obtained empirically from the data closely approximates a 5-day week in which Monday through Friday are assigned equal weight and Saturday and Sunday receive zero. weight, and thus the trading-day adjustment is based on this pattern. There is no holiday adjustment in the assignment of daily weights. The seasonal indexes shown in table A-2 are the product of seasonal and trading-day factors.

Mobile Home Shipments

Seasonal adjustment of mobile home shipments for the period beginning January 1973 is based on shipments from January 1959 through December 1974 and preliminary shipments from January 1975 through October 1975. Statistics on manufacturers' shipments of mobile homes are provided by the Manufactured Housing Institute and include estimates for firms not associated with MHI. The seasonal indexes shown in table A-2 include trading-day adjustment factors.