Results of Analysis of Trends in Public Libraries, 1990-2002

Robert E. Molyneux

The latest

As of mid-November, I have completed an examination of trends for the 20 variables in the state rank order tables, a series of tables that rank the several states in order by these variables. These tables have been published annually since the mid-1990s and have proved popular. Those pages are linked to the 2002 and summary rank order tables.

I have also expanded beyond those 20 variables. I have added the variables collected since FY 1995 on electronic materials. In addition, I have gone beyond the formula of using per capita figures to calculate means by the quartiles and now totals for all public libraries in the US. Analytically, grouping libraries is useful for reasons discussed at length in these pages but totals are easier to grasp and explain in many ways.

Calculating totals for "all" libraries ends up being slightly more complicated to explain than one might think. I am currently calculating totals for two groups of libraries. The first are the 8,259 libraries in the "1990" dataset which are used for the bulk of the trend studies. These libraries submitted data each year to NCES from 1990-2002. The totals are listed by quartile and year to make them comparable to the other trend figures. In addition, I have a total for each year for those libraries. This dataset does not have imputed values and cannot be regarded as a national estimate of these totals. For a national estimate, I use PUSUM, the State Summary/State Characteristics data file. This set of data has included imputations for non-response and gives state totals for all public libraries by year. I have summed these totals for the third table in these totals pages. All these additions have meant that I have to relabel everything and that process is going on now. Sorry for any confusion as this process takes place. I am trying to restrict the use of the word "total" to those pages with the figures for totals. This practice is not followed with the rank order tables where the word is used rather more freely. Note that the links to these totals pages are in larger type so they are distinctive.

Abstract

This page contains links to this first analysis of trends at US public libraries from 1990-2002 using data from the longitudinal file PLDF3 created from the NCES annual compilations of library data. This analysis uses 8,259 public libraries, that is, all reporting data each year from 1990-2002. The results of analysis are reported in quartiles made up of libraries grouped by their sizes of population served. The rationale for this grouping is elsewhere.

The variables analyzed so far indicate that for the large subset of all public libraries in this study, the years from 1990-2002 were good ones. Expenditures and staff, including ALA-MLS librarians, increased as did expenditures per capita. Expenditure categories also kept ahead of inflation. Given that this period was good economically and that public libraries rely on tax revenues and gifts and that both go up during good times, this fact is not surprising and consistent with the experience in academic libraries. The data now run through fiscal year 2002 and but for a few worrisome indicators such as the 2002 drop in staff at the smallest libraries or the drop in ALA-MLS librarians employed at the largest libraries indicators examined are predominately positive.

Introduction

The 2002 data have been added to the longitudinal file so analysis of the data through 2002 will be showing up in these links as they are completed. I will update the documentation after the analysis is done, the 2002 data being enough like the 2001 data that little will be lost for now. I will update the conclusions here when the results are in.

One thing that is clear in the changes from 2001 to 2002 is that figures reported for 1990-2002 have gone up from those reported for the 1990-2001 data merely as a result of the fact that only libraries reporting in all years are included here. The libraries in dropping out between 2001 and 2002 had two effects on the set of libraries examined: 1) they were largely not as strong as the ones that stayed and 2) as the size of this set of libraries declines, the various groups get smaller, too. Thus, the 415 libraries in the 95th percentile in the 1990-2001 group becomes 413 in the 1990-2002 group so the two smallest libraries move out of that group. Each group gets smaller and the process of shrinking increases the average size of both the group the smallest libraries leaves and the one they join, all other things being equal.

I have decided to look at trends on the variables analyzed in the Public Library Rank Order Tables. These tables are included in the annual ED Tabs from NCES and in recent years have included state ranks on 20 variables. It appears they began in 1993 as a supplement to that year's ED Tabs and by 1995, they were published annually in its Appendix. The advantage of analyzing these variables is that they have proven to be useful to the audience and the reader will find these variables in boxes with links to the 2002 Rank Order Tables for that variable as well as a summary table of these ranks from 1992-2002.

There are three problems, however, with analyzing this set of variables. One is that the ranks are per capita and the per capita figures are based on "unduplicated" population and that figure does not exist until 1991 so we lose another year from the 1990 start data for time series analysis of these data. (Unduplicated population refers to the fact that in some jurisdictions two libraries may count the same person in their legal service areas. As a result, per capita comparisons are of uncertain value. To correct this problem, a second measure of population has been developed. So there are two measures of population in the legal service area, POPU_LSA which was first collected in 1987 and allowed duplication and POPU_UND, which began in 1991 and corrected for this population duplication.) As a result, for the per capita ratios that appear in the Public Library Rank Order Tables the calculations will be from 1991-2002 while other variables will be analyzed from 1990-2002. In addition, there are some variables that are important but are not included in these tables such as those measuring electronic materials. The next pass through the data will use unduplicated population and first analyze trends in the rank order tables for 1991-2002. This means that some work already done will be redone as necessary. The third problem is that there are anomalies in the data for the smallest libraries that are exacerbated by the formulas for figuring some ratios as discussed in a section on Anomalous Values in Librarians per 25,000 population. A result of this problem is that there are often large and not very believable jumps in the behavior of ratios calculated on data from these smallest libraries. There are techniques to deal with these kinds of anomalies and they will be used at a later stage to get a less cluttered picture.

Analysis of data often answers some questions but results in more questions being asked and this experience is no less true with these data. At this point, the data are being looked at a national level but closer analysis of disaggregates of the national data such as the experience of each state or other groups of libraries will shed more light on these libraries and their behavior. What is looked at here is averages and it will be important to find those libraries that are better than average and those that are worse than average. And why.

The data analyzed here are a subset of the PLDF3 dataset divided into five groups by population of each library's legal service area (POPU_LSA) as discussed in the "Analysis of 2001 data". These five groups are formed from organizing the libraries into four quartiles and a fifth group--a subset of the fourth quartile--those libraries in the 95th percentile. Given the high correlations shown by these variables, population, in effect, also groups libraries by size, by the number employed, and the dollars expended.

Presentation of the Results

The analysis is reported by topic on separate pages linked to above. So far, these pages give plots of the data for mean values for each of the five groups by year from either 1990-2002 or 1991-2002, depending on how the per capitas are calculated and tables reporting the data plotted. Each follows with conclusions that are summarized briefly below on each page.

The plots use five solid lines to summarize the mean values for the five groups through the period. The color is an indicator of the group with green, for instance, being the smaller libraries in all these plots and purple representing the values for the group of libraries in the 95th percentile. When the variable analyzed is in dollars, the solid line represents the nominal or reported dollars for that group's libraries and the dotted line in the same color as the solid line is the calculated 1990 "constant" dollar.

Plots are a useful tool in exploring data but they are a dubious analytical tool and one that should be used with caution. Consider the example of two plots for Total Expenditures for the New York State Library, one with the vertical axis with an origin at 90 and one with an origin at 0. They use the same data but the effect of the two is different. Similarly, if we had more year's data, the 12-year period on the horizontal axis would be shorter and the slopes would look more dramatic as my work with The Gerould Statistics and Association of Research Libraries' data has shown. Three graphs with data spanning up to 95 years are included elsewhere on this site. So, in all cases, we need the numbers themselves and use them and the graphs to arrive at conclusions that are correct and are not swayed by the construction of graphs.

In most of the plots, the vertical axis does not represent constant values throughout its range because of the characteristic skewing of the data and the result that distinguishing between the lines representing the two or three smallest groups becomes difficult. In order to see these lines, most vertical axes exaggerate lower values. Where values plotted are dollars--such as total expenditures--there is the question of the effect of inflation. These plots have the nominal dollar value--that reported each year--in a solid line with the calculated constant 1990 dollar value in a broken line next to it. The "constant dollar" is calculated using the Consumer Price Index.

Results

Visits to these libraries have risen over the period for each of the groups.

Total Visits to Libraries, 1990-2002

Three measures of total visits to public libraries in the US shows they rose over the period.

Reference transactions per capita rose. Curiously, the very largest libraries have the highest mean number of these transactions with the smallest libraries next highest.

Circulations per capita fell for the smallest libraries after a rise in the early part of the period.

Total Circulations, 1990-2002

Three measures of total circulations at public libraries in the US shows they rose over the period.

ILLs received showed dramatic increases for all groups.

Volumes held increased over the period for all groups.

Audio materials increased over the period for all groups.

Video materials increased over the period for all groups.

Volumes held increased over the period for all groups.

Mean Current Serials Subscriptions, by Quartile, 1990-2002

Number of Current Serials Subscriptions (SUBSCRIP) increased over the period for each of the groups.

The number of paid staff varies by size of size of the population served. Overall, the larger the population served, the fewer staff per person in the legal service area, and the smaller the population, the greater the staff per person. The first quartile libraries declined in this measure over the period while the other groups show an increase over the period.

Mean Paid FTE Staff, by Quartile, 1990-2002

The mean number of paid staff (TOTSTAFF) in each of the groups rose over the period. The largest libraries added the largest number of staff members while the smallest libraries added the fewest but the percent increases are roughly the same. The numbers rose each year for each group except for the first quartile--the smallest libraries--which had a peak in 2001 of 1.04 mean total staff before falling below that level in 2002. Still, the mean number of total staff at the first quartile libraries increased over the period by .15 while the mean staff at the 95th percentile libraries increased by 26.

Total paid staff varies by size of size of the population served. Overall, the larger the population served, the fewer staff per person in the legal service area, and the smaller the population, the greater the staff per person. The first quartile libraries declined in this measure over the period while the other groups show an increase over the period.

The number of librarians increased faster than the populations they served. The percent increase over the period was inversely related to the size of the library.

The number of librarians increased faster than the populations they served. The percent increase over the period was inversely related to the size of the library.

Paid FTE Librarians, with ALA-MLS, by Quartile, 1990-2002

The number of ALA-MLS librarians also varies by size of library. The smallest two size groups, on average, employ less than one full time librarian. The absolute number of librarians has increased for all groups over the time period.

The number of librarians increased faster than the populations they served. The percent increase over the period was inversely related to the size of the library.

Operating income increased over the period for all groups.

State income increased over the period for all groups.

State income increased over the period for all groups.

Other increased over the period for all groups.

Operating Expenditures (TOTOPEXP) for each library were divided by the population in its legal service area . For each of the five groups, this measure rose and also beat inflation.

Mean Operating Expenditures, by Quartile, 1990-2002

Operating Expenditures (TOTOPEXP) increased for each of the five groups over the period and if inflation is discounted using the Consumer Price Index, the expenditures in real terms still rose.

Total Operating Expenditures, 1990-2002

Total expenditures increased in real and nominal terms over the period for each of the groupos of libraries analyzed.

Collection expenditures rose over the period for all groups in nominal and inflation-adjusted dollars. The Fourth Quartile and 95th percentile libraries as a group fell in real terms from 2001-2002.

Staff expenditures rose over the period for all groups in nominal and inflation-adjusted dollars.

Expenditures for salaries rose over the period for all groups in nominal and inflation-adjusted dollars.

Expenditures for Electronic Materials per capita, 1995-2002

Expenditures per capita at the largest and smallest libraries increased while the Second and Third Quartile libraries fell.

Expenditures for Electronic Access per capita, 1995-2002

Expenditures per capita for access at the largest and smallest libraries increased while the Second and Third Quartile libraries fell.

Mean Number of Electronic Materials, by Quartile, 1995-2002

Expenditures per capita for access at the largest and smallest libraries increased while the Second and Third Quartile libraries fell.

Number of Electronic Materials per capita, 1995-2002

Expenditures per capita for access at the largest and smallest libraries increased while the Second and Third Quartile libraries fell.

November 19, 2004