Printer Friendly

Improved estimates of future occupational replacement needs.

Many people and organizations need information about projected job openings by occupation-openings that result from employment growth or the need to replace workers who leave an occupation. For example, students and vocational counselors need such information to make informed decisions affecting career choices, planners of training programs need it to formulate rational education policies, and personnel specialists need it to focus their recruiting efforts. During the past several decades, information about employment growth has been provided biennially by the BLS employment projections program. While recognizing the importance of replacement needs in estimating j ob openings, BLS stopped developing estimates of such needs in the early 1980's because of concerns about the quality of the data and the methods of developing data appropriate for different users.

In 1990, BLS began an extensive project to review the methods used to develop estimates of replacement needs in the past and to determine whether improved estimates could be developed. This research summary presents an overview of the results of that project.(1)

Most descriptions of the labor market, such as those based on data from the monthly Current Population Survey (cps), are developed from information pertaining to a single point in time that provides a snapshot of current conditions. Individuals are classified as employed, unemployed, or not in the labor force. Employed persons are further identified by occupation. For any pair of snapshots, whether taken a month or a year apart, the number of individuals in each category generally does not change very much. The image thereby projected is one of great stability in the labor market. However, this is practically never the case. During any time period, there is a great deal of movement into, out of, and between occupations. Measuring this movement to develop estimates of separations from occupations requires longitudinal data about workers at two points in time, data that are much less common than snapshots of current conditions. The research in the BLS project focused on the development of procedures that, using available data, would provide the best measure of movements of workers out of occupations over time.

The research concluded that two distinct types of estimates of occupational separations should be developed to meet the needs of all users. The first type of estimate, total separations, would measure all individuals who leave their occupation. The second, net separations, would measure the net movements of experienced workers into and out of occupations. It was found that both measures of separations are best developed using data from the cps, but through different data elements. Total separations are best measured by identifying the experiences of individuals over a year's time, a finding that reinforces research conducted in the late 1970's and early 1980's. By contrast, net separations are best measured by following age cohorts of workers over a longer period of time, a methodology that results in a new approach to developing net occupational separations.

Concepts and definitions

Over the years, a variety of concepts have been used to calculate estimates of occupational replacement needs and job openings. These different concepts result in significantly different estimates of separations for the same occupation that often have confused users of the information. Accordingly, this section presents a brief summary of those concepts, in an effort to dispel whatever confusion might otherwise occur.(2)

Employment growth. If employment is measured at the beginning and end of a given period and is observed to increase, that increase is a measure of employment growth. A positive net change in employment, employment growth creates opportunities for workers to enter an occupation. It results from increased demand for goods and services in the economy and from changes in the occupational structure of industries and is the source of job openings identified by BLS projections.

Determining employment growth requires only information about employment at two points in time; no information about separations is required. However, employment growth also may be determined by using information about the labor market dynamics of an occupation. For example, employment growth over a given period can be calculated by subtracting the number of persons separating from an occupation from the number entering that occupation.

Total separations. Total separations identify the flow of individuals leaving an occupation, for any reason whatsoever, without regard to persons entering the occupation. Total separations are the largest measure of separations. During a given period, some individuals may leave an occupation for a variety of reasons and must be replaced. Some become employed in a different occupation-the result of a promotion, a desire to change careers, the loss of an existing job, the need for a different job while attending school or training or caring for a family, or some other reason. Others who leave an occupation stop working altogether, because they retire, desire more leisure time or time for an extended vacation, assume family responsibilities, return to school, move out of the geographic area, become ill, or have some other reason. If employment in an occupation is to grow or remain the same, those individuals who left the occupation must be replaced. Thus, total occupational separations are, in most cases, replacement needs and a source of job openings. However, if employment is declining, occupational separations exceed replacement needs by the decline in employment because some persons who leave the occupation are not replaced. (Individuals who change employers but remain employed in the same occupation are not included in counts of replacement needs because job changes by these individuals have no impact on the number of openings for persons desiring to enter an occupation.)

Net separations. For any occupation, net separations differ from total separations in that they summarize movements of workers into and out of the occupation over a specific period. Net separations provide an estimate of the number of openings for new entrants to replace workers who leave the occupation.

Employment data, by age, for two points in time are used to estimate net separations. For example, occupational employment, by age, is prepared for a base year and for a second year 5 years later. Then, employment figures for each age group in the base year are compared with employment figures for the group that is 5 years older in the second year. For example, employment in the age group 55-59 in the base year is compared with employment in the age group 60-64 in the second year for a given occupation. If employment has increased from the one group to the other, then the increase measures the net entrants into the occupation for the second group, and the net separations from the occupation for that group are zero. If, instead, employment has declined across the two age groups, the decline is recorded as the net separations from that occupation for the second age group. The total net separations from the occupation in question are then the sum of the net separations from that occupation for all age groups.

It is important to note that within any age group, individuals may have left the occupation and started working in another occupation, stopped working altogether, or left the geographic area, no longer to be included in employment data for the area. Similarly, individuals entering the occupation may have been working in another occupation, may not have been working at all, or may have come from another geographic area, to be added to the number of employees working in the area. Thus, the change measured over the period in question reveals only whether there were more or fewer entrants than separations over the period and reveals nothing about the magnitudes of the total entrants, total separations, or any of their components. That is, the change indicates whether the size of the original age group increased or decreased, but indicates nothing about the specific actions of individuals constituting the group. However, inferences can be made from the age distribution of the net separations that explain the movements recorded, as will be illustrated later.

Replacement needs. Total job openings consist of employment growth and replacement needs resulting from total separations. Similarly, net job openings consist of employment growth and replacements needs due to net separations. In developing estimates of replacement needs, the distinction between total and net separations and replacement needs pertaining to an occupation must not be overlooked. When employment in an occupation remains the same or increases over a given period, replacement needs equal separations. However, when employment declines, replacement needs are less than separations because some individuals leaving an occupation are not replaced.

During a period when employment in an occupation declines, total separations will be greater than they would be if employment increased, because more individuals lose their jobs; and net separations would be greater not only because more individuals leave, but also because fewer enter the occupation. Because a decline in employment represents individuals who left an occupation and were not replaced, replacement needs during a time of declining employment are determined by reducing observed separations by the decline in employment.

Total job openings. Total job openings equal growth plus replacement needs resulting from total separations and provide the broadest measure of openings in an occupation. Estimates of total job openings are useful for identifying differences in the demand for additional employees across occupations. For example, there are about as many waiters and waitresses employed as there are elementary school teachers, but total job openings for waiters and waitresses are much higher because annual replacement needs are triple those of elementary school teachers.

Net job openings. Net job openings equal growth plus replacement needs resulting from net separations. For some purposes, estimates of total job openings are not very helpful. Planners of training programs, for example, cannot use such estimates to identify the number of teachers to train annually, because some openings are filled by teachers who previously left the occupation and some are filled by workers employed in other occupations who qualify for teaching positions but who do not require additional training. To identify their training needs, planners must know the number of openings for new entrants to the occupation that are due to growth and replacement needs resulting from net separations from an occupation.

Measures of total separations

All individuals who leave an occupation-those who transfer to another occupation or who stop working for any reason-must be included in a measure of total separations. Producing such a measure requires longitudinal data that include information about individuals at two points in time. From the late 1970's to the early 1980's, using cps data, BLS developed a procedure for estimating the total number of job openings arising from workers who leave their occupation over a 1-year period.(3) Annual data are preferable to data with other periodicities because most data on training program completions are compiled on an annual basis. Thus, annual total separation data facilitate analyses of occupational supply and demand.

Complete descriptions of the methodology of developing estimates of total separations and discussions of the limitations of the estimates are presented in the Bureau's publication, Total and Net Occupational Separations.(4) Briefly, the methodology consists of creating a matched sample over a 1-year period from the CPS. Matched data are created for each of 12 months and then are combined, resulting in a sample of about 500,000 persons aged 15 and older in the initial year. For the project discussed in this report, matched data for 1986-87 dealing with changes in labor force status were merged with data on occupational transfers from a special study conducted as part of the January 1987 cps, the latest available survey of this nature when the project was conducted. Occupational transfer data from the January 1987 cps were used because matched CPS data overstate the number of workers who change occupations. The excessively large estimate of occupational transfers in matched CPS data occurs because individuals may respond differently to the same CPS question about their occupation, responses may be recorded differently by interviewers collecting the data, or recorded information may be interpreted and coded differently by clerks preparing files for computer processing. All these actions result in a different occupation being recorded in the second year when, in fact, no change of occupation occurred. The results of combining 1986-87 matched CPS data and occupational transfer data from the January 1987 CPS are termed merged data and provide a composite description of movements into, out of, and between occupations over a 1-year period. The procedure results in data that identify the numbers and types of separations and the characteristics of workers who change occupations, become unemployed, or leave the labor force.

Total separation data for occupations with fewer than 50,000 employees in 1986 were judged unreliable because of the limited number of observations in the sample. Data for the remaining occupations were examined individually, and if data identifying specific reasons for leaving the occupation appeared suspect, another detailed occupational group was selected to serve as a proxy and provide substitute data.

The CPS is conducted primarily to obtain current data on the labor force status of individuals, rather than data that measure changes over time. Therefore, there are significant limitations to the data that describe change. Because the CPS is a household survey that obtains data about persons living at a specific address, one limitation to the matched sample is that information can be developed only from the responses of individuals who do not change their residence. Movers tend to change their labor force status more than nonmovers; hence, the separation rates are biased downward because movers are not included. Separation rates also are biased downward because the CPS excludes individuals who die between surveys.

By contrast, response and coding errors bias the separation rates upward. For example, if employed persons were incorrectly classified as not in the labor force during the second survey, the matched data would indicate movement where none occurred. Although the net effect of the various biases on the movements is not known, their impact is offsetting and not concentrated by occupation.(5)

It must be emphasized that total separation rates developed from merged cps data are not measured rates based on longitudinal data about individuals, but rather, are a composite estimate of movements from occupations based on CPS data from two distinct sources. However, the rates are occupation specific and, in addition to their value in estimating job openings, are extremely valuable for describing the labor market. The 1986 edition of Occupational Projections and Training Data describes differences among occupations and discusses demographic factors affecting total separation rates.(6)

Measures of net separations

Because the classification system used in the CPS has remained constant since 1983,(7) a comprehensive measure of occupation-specific net separations can be developed by using changes in age groups over a 5-year period. When the size of a group increases, a measure of net entrants is recorded; when it declines, net separations are identified. Net changes in an age group capture the net effect of transfers into and out of occupations, immigration, and emigration, as well as of labor force entrants and separations, including deaths. A 5-year period was chosen so as to reduce the impact of cyclical variations that might accompany a shorter period. However, data for other periods can be developed. Data also can be developed by industry, by educational attainment, by sex, and by a variety of other demographic variables. Thus, this new "cohort" technique provides a powerful tool for analyzing labor market changes.

Employment data for appropriate age groups, by occupation, were developed for 1983-88, 1984-89, and 1985-90. Initially, approximately 850,000 records containing information on occupation, age, and many other characteristics for all employed persons in all months of 1983 were combined, and occupational employment by age group was tabulated. The process was repeated to obtain data for desired age groups in 1988. To increase the sample size and reduce cyclical fluctuations, data for the same age groups as recorded for 1983 were developed for 1984 and 1985, and data for the age groups used in 1988 were developed for 1989 and 1990. Data on employment by occupation, by age group, then were averaged and used to prepare the data presented in this report. To simplify the presentation, all references to 1985 data represent averages for 1983, 1984, and 1985, while references to 1990 data represent averages for 1988, 1989, and 1990.

Net leavers in most occupations occur only in the older age groups, generally above age 45. This pattern typically describes individuals leaving in large numbers to retire. A different pattern displayed in some occupations is the vast majority of all net separations taking place in the youngest age groups. In this case, large numbers of workers probably obtained employment in the occupation when they first entered the work force. When they were ready to begin full-time jobs, or when they qualified for higher paying jobs, they transferred to another occupation. In both patterns, the net separations quantify the number of persons who left the occupation permanently. Table 1, which illustrates how net separations for registered nurses and for waiters and waitresses were calculated, shows these different patterns.

In the table, employment data by age group for registered nurses and for waiters and waitresses in 1985 are compared with data in the same categories for a 5-year-older group in 1990. For example, the number of registered nurses aged 20-24 in 1985 are compared with the number of registered nurses aged 25-29 in 1990, and the difference is calculated. If the difference is positive, more individuals aged 20-24 in 1985 entered than left the occupation. Nothing is known about the number of persons transferring into the occupation, coming from outside the labor force, or coming from another country, and nothing is known about the number of persons transferring out of the occupation, leaving the labor force, or leaving the country. The difference between the two groups simply identifies the amount by which total entrants exceed total leavers. If, by contrast, the difference is negative, more individuals left than entered the occupation. Only a negative difference results in a measure of net separations. Positive differences are recorded as zero net separations for the age group. The separation rate for an age group is calculated by dividing net separations by 1985 employment in the age group. Net separations for all age groups were totaled and divided by total employment in 1985 to obtain the 5-year net separation rate for the occupation.

From the table, registered nurses experience net separations only in the older age groups. Those aged 20-24 in 1985 increased by 104,000 during 1985-90, the largest increase of any age group. Much of the increase probably reflects newly qualified graduates entering the occupation. Because few nurses leave the occupation after becoming qualified, there are no net separations in the younger age groups. The group aged 45-49 in 1985 is the first age group having a measure of net separations. Most net separations for nurses occur in the groups aged 55-59 and 60-64 in 1985, ages at which many nurses retire.

A much different pattern exists for waiters and waitresses, who experienced the largest number of net separations in the group aged 20-24 in 1985 and smaller, almost steadily declining numbers in all of the remaining age groups. Only the group aged 16-19 in 1985 experienced net entrants. The data suggest that many workers get jobs as waiters and waitresses when they first enter the labor market or as part-time jobs while attending school. After gaining experience, completing school, or qualifying for a full-time job, many transfer to other occupations. Indeed, few remain in the occupation long enough to reach retirement age. Thus, most net separations among waiters and waitresses result from occupational transfers, whereas net separations among registered nurses are due mostly to retirements. In each case, however, replacement needs exist.

Table 1 also presents information on the percentage of leavers in each age group for registered nurses and for waiters and waitresses. This measure is calculated by dividing net leavers in the age group by 1985 employment in the age group. Information about the percentage of leavers in each age group is especially valuable because it permits estimates of net leavers in the future, as is discussed later.

Registered nurses and waiters and waitresses are large occupations, so the cps sample for these occupations provides quite reliable employment data for each age group in them. For small occupations, however, such as actuaries, statisticians, and mathematical scientists not elsewhere classified (n.e.c.), shown in table 2, the sample is too small and the net separation data are unreliable. For example, statisticians have an irregular distribution of net separations among the age groups, and their net separation rate of about 25 percent is inconsistent with rates for other professional occupations.

To obtain a separation rate for each detailed CPS occupation, one of two procedures was used when an occupation was judged to be unreliable on the basis of its data. When a larger detailed occupation had characteristics similar to those of the occupation in question, the larger occupation was chosen as a proxy for it, and the separation and employment data for the proxy occupation were substituted for the unreliable data and were used to calculate separation rates. When there was no larger detailed occupation with characteristics similar to those of the occupation in question, separation and employment data for a summary occupation group were substituted for the unreliable data. In this case, the procedure for determining separation rates was not as straightforward. Note in table 2 that for the summary occupation group, "mathematical and computer scientists," no net separations are measured in the data until age 55. Yet, of the detailed occupations comprising the group, actuaries, statisticians, and mathematical scientists, n.e.c., exhibit net separations prior to that age. The summary occupation group does not register those separations because total net entrants in the other detailed occupations-computer systems analysts and operations and systems researchers-exceeded the total of net separations among actuaries, statisticians, and mathematical scientists, n.e.c. However, to exclude the measure of net separations from the summary occupation group would result in an understatement of separations from detailed occupations. To overcome this limitation, net separations in each age group for summary occupation groups were calculated by summing the net separations for each detailed occupation in that age group. Thus, the net separation data for each age group for the summary occupation group, "mathematical and computer scientists," in table 3 is the sum of the data measured for computer systems analysts, operations and systems researchers, actuaries, statisticians, and mathematical scientists, n.e.c. (Because unrounded data for detailed occupations are used, the totals shown may not be the sum of the data presented in table 2.)

The adjusted separation data for the summary group, "mathematical and computer scientists," were used to calculate separation rates that became the proxy separation rates for actuaries, statisticians, and mathematical scientists, n.e.c. Net separations for other summary occupations found throughout the occupational structure of the economy were developed in the same manner as for mathematical and computer scientists.

Projected separations

Thus far, all the information presented about separations in this research summary has described what has occurred in the past. However, the Bureau's employment projections program focuses on future opportunities, a purpose that requires projections of employment change and, in addition, projections of replacement needs due to total and net separations.

Total separations. Total separation rates for all detailed occupations were developed from merged cps data for the period 1986-87. As described in the previous section, total separation rates from proxy occupations were substituted for small occupations because the data appeared unreliable. If employment in the occupation in question was the same or increased from 1986 to 1987, the 1986-87 total separation rate also was the replacement rate and should be used to estimate replacement needs during a projection period. However, if employment declined, the replacement rate was calculated by subtracting the employment decline from the separations.

To estimate annual average total replacements in an occupation during a projection period, the 1986-87 replacement rate should be multiplied by the employment in the occupation at the midpoint of the projection period. Because labor market conditions affect the replacement rates, attempts to adjust the rates would be fraught with difficulties because not enough is known about the effect of cyclical factors and other labor market conditions on the rates.

Net separations. To develop a net separation rate for an occupation, employment figures for that occupation in a given age group in 1985 were compared with employment in the occupation in 1990 for a group that was 5 years older. As noted earlier, data for 1985 actually consist of the average of data for 1983, 1984, and 1985; data for 1990 consist of the average for 1988, 1989, and 1990. If employment for the group increased, no net separations occurred, and separations were recorded as zero. If employment declined, the number was recorded as net separations for that age group. The 5-year net separation rate for the age group was calculated by dividing the number of net separations by employment in 1985. (See table 1.) The 5-year net separation rates for 1985-90 for each age group could then be applied to employment in future years to obtain a projection of net separations.

Between 1985 and 1990, employment in most occupations increased or remained the same, and the 1985-90 net separation rates, by age, were used without adjustment to estimate replacement needs during the projection period. If employment declined, however, one of several adjustments to the age-specific separation rates was used to obtain a replacement rate that reduced the occupational separation rate by the rate of decline in employment. When the employment decline was less than the number of net separations among persons aged 16-49 in 1985, the number of net separations among persons aged 16-49 was reduced by the employment decline. The decline was distributed in proportion to the number of net separations in each age subgroup in the group aged 16-49. This was the most frequently used technique; it confines the adjustments to the ages most affected by adverse economic conditions, because older workers are more likely to remain employed until they retire. In the remaining cases, the net separations were reduced in a like fashion for persons aged 16-54 or persons aged 16-65, depending on the distribution of net separations in the occupation and the amount by which employment declined. The adjusted age-specific rates then were used to calculate future net replacement needs for persons employed in 1990.

Table 4 illustrates the method for calculating net leavers over the period 1990-2005 from the stock of persons employed as registered nurses in 1990. First, net leavers were calculated for 1990-95 by multiplying 1990 employment obtained from the cps for each age group by the replacement rate in 1985-90 for the same age group. Then, before net leavers in 1995-2000 were calculated, employment in 1995 for each age group was determined by identifying employment in 1990 for a 5-year-younger age group and subtracting any net leavers for the period 1990-95. For example, table 4 shows the 1995 employment figure for registered nurses aged 55-59 to be 98,000. This estimate was arrived at by identifying the 1990 employment figure for nurses aged 50-54 (114,000) and subtracting the 16,000 net leavers in 1990-95 from that age group. When employment for each age group for 1995 was developed, the resulting figure was multiplied by the replacement rate for that age group to estimate net leavers for 1995-2000. The process was repeated to obtain employment for each age group in 2000 and to estimate leavers in 2000-2005. Summing the number of net leavers for each of the 5-year groups provided an estimate of net leavers for the 15-year period 1990-2005. Dividing the net separations for 1990-2005 by 15 yielded annual average net separations; dividing the annual average net separations by 1990 employment yielded an annual average net separation rate.(8)

New entrants-that is, individuals who were younger than age 16 in 1990 but can be expected to join the group of employed persons after 1990-are not included in the estimate of separations for 1990-2005. If they were included, estimates of separations with net transfers in the younger age groups--such as those for waiters and waitresses--would be larger.

OES-survey-based separation rates

The preceding section described procedures for estimating annual average rates of total and net separation for detailed cps occupations for the 15-year period 1990-2005. However, the BLS projections program uses an industry-occupation matrix to estimate current and projected employment data that are based on the occupational classification system of the Occupational Employment Statistics (OES) survey. Current and projected occupational employment data based on the OES survey are used for calculating the employment change component of projected job openings. To obtain the replacement need components of projected total and net job openings, estimates of total and net separations based on the OES survey had to be developed. The procedure required total and net separation rates for all detailed occupations, including collapsed occupations, based on the OES survey.(9) The development of these rates was accomplished by identifying the cps occupation or occupations that were equivalent to the detailed OES-survey-based occupation and either using the CPS rate directly or calculating a weighted rate using OES or CPS employment figures as weights if the occupation consisted of more than one OES or CPS occupation. Table 5 presents data for selected occupations.

Footnotes

1) A more comprehensive report on the project is presented in Total and Net Occupational Separations: A Report on Recent Research (Bureau of Labor Statistics, August 1991), which can be obtained upon request from the Bureau of Labor Statistics, Office of Employment Projections, 600 E Street N.W., Washington, DC 20212. In addition to the project's research on national data that is discussed in this research summary, the report includes material prepared by Patrick Berkery of the New York State Department of Labor that discusses the applicability of national data in developing estimates of replacement needs for States.

2) In this research summary, data are presented on total and net separation rates only for selected occupations developed in the BLS project. Data covering all occupations examined in the project are presented in Total and Net Occupational Separations. The 1992 edition of Occupational Projections and Training Data, scheduled for publication in the spring, also will present comprehensive data, including 1990 data on total separations taken from a special supplement to the January 1991 cps that were not available when the BLS project was being conducted. Results of the project will be incorporated into some of the analyses presented in the 1992-93 edition of the Occupational Outlook Handbook, available in spring 1992. Some of the data on net separations from the BLS project are used in the article, "Occupational employment projections," pp. 64-94.

3) An early version of the procedure was developed in 1978 and produced 1977-78 data that were presented in Measuring Labor Force Movements: A New Aproach, Report 581 (Bureau of Labor Statistics, 1980). Modifications were incorporated into the procedure in 1982, and the revised procedure was used to develop 1980-81 data. (See Alan Eck, "New occupational separation data improve estimates of job replacement needs," Monthly Labor Review, March 1984, pp. 3-10.) The estimation procedure appears in Occupational Projections and Training Data, Bulletin 2202 (Bureau of Labor Statistics, 1982), appendix B, pp. 67-69. The methodology used to prepare 1986-87 data is identical to that used to prepare the 1980-81 data.

4) See footnote 1.

5) A more detailed discussion of the limitations of merged CPS data appears in Occupational Projections and Training Data, Bulletin 2202 (Bureau of Labor Statistics, December 1982), appendix B, pp. 73-75.

6) Occupational Projections and Training Data, Bulletin 2251 (Bureau of Labor Statistics, April 1986), pp. 17-23.

7) In 1983, the cps converted its occupational classification system to the one used in the 1980 decennial census. From 1972 through 1982, the cps used the occupational classification system used in the 1970 census.

8) An annual average net separation rate will vary depending on the projection period, because the age distribution changes, Also, separations are based on the number of employed persons in the base period. New entrants to the occupation during the projection period are not included in net separation estimates.

9) The term "collapsed occupations" refers to occupations based on the OES survey for which national data were not published, but were aggregated in a summary occupation. However, information on the detailed occupation may be presented in State publications, and therefore, an estimated replacement rate was developed. TABULAR DATA OMITTED
COPYRIGHT 1991 U.S. Bureau of Labor Statistics
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1991 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Eck, Alan
Publication:Monthly Labor Review
Date:Nov 1, 1991
Words:5435
Previous Article:Occupation employment projections.
Next Article:Profiles in safety and health: fabricated structural metal.
Topics:


Related Articles
Occupational employment based on 1972 and 1987 SIC.
Markets, news.
Occupation employment projections.
The future of jobs for college graduates.
Evaluating the 1990 projections of occupational employment.
The nature of occupational employment growth: 1983-93.
Occupational employment to 2005.
Occupational employment projections to 2006.
Korean Occupational Outlook Handbook: first edition.
Forecasting future demands: what we can and cannot know.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters