Total Staff Expenditures per capita at Public Libraries, 1991-2002

Abstract

Total Staff expenditures rose over the period for all groups and stayed ahead of inflation as measured by the Consumer Price Index. Even taking into account the peculiarly low values of the First and Second Quartile figures for 1991, these expenditures for the period 1992-2002 were still ahead of inflation. The obviously peculiar values of this set of libraries is discussed below in the Anomalous Values section below.

There is a Rank Order Table for this variable in 2002 and a summary table of the ranks by state from 1992-2002.

Per capita Collection Expenditures Calculated Using UNDUPLICATED Population Served

First the smaller libraries:

And the larger ones:

Data

The chart summarizes these values:

Note: The value for the Town of Ulster Public Library, Kinston, New York (newkey = NY0259) for POPU_UND for 1992 was 8,478, 1993 was 1, and 1994 was 11,507. In this analysis, the value for 1993 was changed to 9,993, the mean of the 1992 and 1994 figures.

Anomalous Values

See a similar discussion at Paid FTE Librarians per 25,000 population.

The data from the first two quartiles causes three kinds of questions, with my speculations:

1. Why are the 1991 values of these quartiles so much lower than the other years?
   • The first year of data collection--in this case of unduplicated population--often results in peculiar values. It often takes a few years of collection for the differences in understanding of definitions to be cleared up and for the methods of collecting new data to be established and refined. Commonly, then, these kinds of swings occur in the early years and disappear in the later years as everyone learns how to collect data.
2. This number is a ratio that has expenditures for staff divided by the population and as such, it can have great swings, particularly in small libraries, as seen in other chats in this series of Web pages, such as that at Paid FTE Librarians per 25,000 population or Total FTE Staff per 25,000 population. Note the value of the ratio of FTE librarians for one library in Kansas changes dramatically as the result of a minor change in absolute terms in the small numbers reported for this library. Note the following table of one library's experience with staff expenditures per capita, the variable summarized here at a library in Illinois (newkey=IL0037). The right column is the variable in question. To give a sense of scale: the average value for this variable for all libraries in the US as calculated in the rank order table was $18.72 in 2002.
   

   Total Staff Expenditures per capita at Bedford Park, Illinois, P.L.D.

   	Staff Expenditures	Unduplicated Population Served	Total Staff	Expenditures per capita
   1991	$220,312	988	3.00	$223
   1992	228,004	988	10.68	231
   1993	272,972	988	10.68	276
   1994	288,859	988	10.68	292
   1995	308,762	566	10.68	546
   1996	324,264	566	10.68	573
   1997	305,829	566	11.22	541
   1998	337,052	566	12.32	595
   1999	405,489	566	12.82	716
   2000	413,572	566	12.97	731
   2001	444,686	566	12.62	786
   2002	429,486	574	12.63	748

   • What are we to make of these numbers? They certainly are unusual but they illustrate an aspect of how these kinds of ratios work--we will get values that move around dramatically. One of the advantages of grouping a number of like libraries together is that it damps such swings and allows us to see underlying patterns without having our eyes or minds distracted by the unusual. In the several cases cited in this section, we have swings that may be a function of unusually high or low ratios because of unusual situations.
   • Are these numbers wrong? Should they be changed?
     
     There are times, as cited just below the main table reporting on this variable where it is noted that the library in Kinston, New York had an obvious error in its reported population served in 1993. This error caused a spike in the plot of the group this library is in similar to what we see with the First Quartile in 1996 although there does not appear to be a similar error in that year. We can tell this 1993 error from the context and the solution I used to keep this library in the sample was reasonable and explicit. "Fixing" 1996 is different because there is not one obvious error. What about Bedford Park? Or all libraries with a ratio above some number we might agree on? There are about 120 cases in the dataset where this ratio was above $100--should we drop them?
     
     I think the numbers should be left as they are unless we have some authority to change them or something compellingly obvious. I have made this argument before in the Gerould Statistics (see Section 2.5) where I argued that authority, rather than some notion of consistency, should be used in such non-obvious cases. A fundamental problem is that libraries are financed, staffed, and run in many different ways in the several states--that's Federalism in operation. And, a second fundamental problem is that we have only the vaguest notions of those different ways libraries operate and what could be considered typical ways or, worse, good ways of the many possible. We are at the stage where we need to use the data to construct theories about the state of libraries, not use our unformed notions about these libraries to change data.
     
     However, if anyone wants to do something like that in analysis, then go ahead. The test of any analytical technique is--if it is disclosed--does it shed any light on the problem?
3. What about the large difference between the staff expenditures per capita at the First Quartile libraries and the 95th percentile libraries--is this difference believable?
   • I don't know but I suspect so. If you look at the staffing and expenditure patterns of the various groups, it is clear that the libraries in the various groups solve the problem of running these libraries in different ways, that is, with different mixes of resources. In addition, this method of grouping libraries masks other variables that may be more telling in helping us understand how libraries are provisioned in the various states.
     
     As said above, we do not know much about what is state of these libraries is through time so we are at the beginning of this investigation. Until we know more, we should take what the data say seriously but be aware that the next analysis may lead to another question we haven't thought of today or shine a light on something that is obscure.

October 25, 2004