Forecasts of the expected number of MLS public librarian positions:
2007-08 to 2016-17

1.Backdrop 1992-2005

The demand for library services grew noticeably during the period. While population grew at a 1.3 percent annual rate (final reports should use resident population data), library visits grew more than twice as fast, 2.9 percent at an annual rate, according to data from annual censuses conducted until recently under the aegis of the National Center for Education Statistics (NCES).[1] In order to assess the sustainability of such usage growth, it will be necessary to evaluate how visits are affected by change in population mix, particularly age distribution.

It is noticeable, though expected, that not all may result in the same increment to the demand for librarians. Thus, circulation grew at a 2.2 percent rate and reference transactions even slower at a 1.8 percent rate as compared with the growth of visits. Interlibrary loan activities (incoming and outgoing) grew at double-digit rates. Data are needed, for at least one cross-section year, on the time librarians spend on each of these functions in order to determine whether changes in the mix of these demands affect the demand for librarians and/or; other staff input. In the interim, it is instructive to note that the rate of growth, library staff (1 ¾ the percent) was most in line with that of reference transactions. Staff growth during the period was about the same for all main components (librarians and other staff).

2.Demand projection methodology and data

The foregoing facts suggests that one approach to forecasting demand would be to explain reference transactions by visits, which in turn could be explained by population growth. Then, projected demand could be measured and compared with projections of potential supply of librarians (derived primarily from projections of librarian degrees awarded, retirements and stock of librarians not working as such) to see if demand and supply match. The econometric methods to accomplish this are simple and straightforward time series and cross-section analysis. If demand and supply do not match, various scenarios could be stimulated under which the gap could be eliminated.

The data to conduct much of the analysis come from the census of libraries NCES has conducted annually. The FY 2005 census data, latest available date is the 18th in this annual data collection. It is a cooperative endeavor with participation from agencies at the Federal and state and local level.

The data set used in this analysis was compiled from the electronic versions of the NCES data.[2] It covers the 14 year period 1992 to 2005. Data were available for several earlier years but gaps in them dictated starting the series in 1992. The fourteen years offered enough observations for statistical purposes and provided coverage of at least a decade of behavior, considered a minimum for this type of analysis. The data set also permitted the analysis of several cross-section data sets for comparison with time series estimates.

3.Time series results

To test the relationships just posted, two equations were estimated for the 1992-2005 period:

1.Relationship between population and visits to public libraries

V = a + b P where
V = natural log of visits
b = the elasticity of visits with respect to population
a = natural log of constant term
P = natural log of population, 18 and over

The results are:

_V = -27.34 + 2.54 P R2 = .996; DW = 1.52; n = 14

(31.0) (54.5)

The equation is quite robust – high R2, “t” ratios and Durban-Watson (DW) statistic, the test for autocorrelation. The DW indicates there is a very low probability that the parameter estimates – a and b – are biased. Nonetheless an alternative equation was estimated, one designed by the econometrician Koyck to get a more accurate estimate when autocorrelation is present.

The results are:

Vt = -a + b Pt + cVt-1

_Vt = -20.64 + 1.90 Pt + 0.262Vt-1 R2 = 996; DW – 1.62; n = 13

(2.9) (2.8) (1.0)

The coefficients are used to derive an unbiased estimate of “b” using the formula (b) where 1-c

“c” is the coefficient of the variable Vt-1. Thus:

1.90 = 2.57,

1-.262

not much different from the 2.54 estimate of b in the previous equation.

2.Relationships between visits and reference transactions

R = a + b V where

R = natural log of reference transactions

The results are:

_R = 9.95 + .46 V; R2 = .716; DW = .43; n = 14

(6.1) (5.8)

The parameter estimates of this equation are contaminated by autocorrelation as indicated by the DW statistic. Thus, the Koyck form of equation was estimated. The results are:

Rt = 6.42 + .60Vt + .068 Rt-1 R2 = .975; DW = 2.30; n = 13

(8.9) (9.0) (1.8)

Transforming the parameter estimates yield __.069_ = .172 = the elasticity of reference transactions with respect (1-.598) to visits.

The effect of population growth on reference transactions can be calculated directly by

R = evep = 2.57 x .172 = 0.44 percent, ie,

for every one percent increase in population 18 and over, reference transactions rise by 0.44 percent.

Latest Census Bureau population projections are for growth of 0.85 percent annually for the population 17 and over from 2010 to 2020. Multiplying that by 0.44 yields a growth in reference transactions of 0.37 percent per year.

As noted earlier, the employment of librarians (and other staff) grew at an annual rate of about 1.75 percent from 1992 to 2005, about the same as reference transactions. Thus, this model projects the growth rate will drop to 0.37 percent. That is exactly the same projection for librarians employed in all types of libraries published by the Bureau of Labor Statistics (BLS) in its Occupational Outlook Handbook for the 2006-16 period.

The BLS figure is a “net” projection, consistent with NCES data on public librarian employment and the model developed above. Neither are projections of the “demand” for librarians. Such demand includes both new demand and replacement needs. Rather, the rates of “demand” growth just presented are akin to job openings. The BLS handbook projection and the NCES data on librarians do not reflect the jobs that must be filled just to replace retirees. Survey data from this project should facilitate the estimation of replacement rates, as it did in the 1983 study.

The filling of that demand does not occur smoothly. That can be seen from the regression of public librarian employment on reference transactions:

L = a + by R, where L is the natural log of employed librarians.

_L = 9.78 + 1.048 R R2 = .808; DW = .41; n = 14

(2.7) (7.5)

The equation is adversely affected by autocorrelation, stemming from its likely mis-specification. Assuming reference transactions drive library jobs, such jobs grew more slowly than would have been expected in the mid to late 1990’s by this equation. This suggests that the number of librarian positions and the pace at which they are filled lags the growth of librarian “work“, measured by reference transactions.

Griffiths and King, A Strong Future for Public Library Use and Employment (ALA 2011)

[1] These censuses are now under the aegis of the Institute for Museum and Library services (IMLS).

[2] Thanks to Songphan Cheomprayong for developing this data set and calculating various analytical statistical measures based on them.