EL PROCEDIMIENTO ADMINISTRATIVO COMO GARANTÍA FORMAL Introducción

Allocative efficiency is an important topic in the public finance literature and for policy- makers and citizens given current budget constraints. The case of public libraries in West Virginia is of special interest. State and local governments have been facing severe fiscal problems as the coal industry, the main driver of local government revenues, has declined over the past decade. The state also allows library to receive funds from local governments through special laws that in some cases date back to the 1930s. Moreover, special library levies have been placed onto local ballots in recent years and passed, suggesting the median voter in some communities desire more library output. In order to better understand the institutional determinants of TE in WV, we regress the technical efficiency on the revenue from different sources and the legal base of the library system. Formally we have:

T Eict = αRevict+ βGovit+ γXict+ µc+ δt+ ict (4.1)

where T Eict is the technical efficiency of library system i, in county c, in year t; Rev is the

vector of revenue from local, state and federal governments, Gov is the vector of dummy variables for governance of each library system,X is a vector of control variables such as existence of branches, if the library is operating under increasing returns to scale, unem- ployment rate, personal income per capita, high-school graduation rate, and demographic controls such as gender, race and age; µc is county fixed effect, and δt year fixed effect.

Our primary variables of interests are those related to governance structure and the source of funding. Following Hoxby (1999) and Hall (2007) we expect local spending to be positively related to TE. We do not have strong priors with respect to governance structure, although

Amir B. Ferreira Neto Chapter 4. Technical Efficiency of Library Systems 71 the work of Ostrom et al. (1961) suggests that cities and multi-jurisdictional governments might be more efficient as they might be more “organic” than county governance. We also include a binary variable for whether or not the library district had a special law passed by the state legislature that gives it dedicated local funding from the property tax.7 _{Table 4.4}

provides summary statistics of the control variables used in the econometric analysis. As previously discussed, we are interested in three set of results. Primarily, we want to know if public libraries governed by local authorities are relatively more efficient than others. Then we want to test if governments are investing in efficient libraries, which in turn suggests that they favor voters and their demand for more types of public goods. Lastly, we want to know which county characteristics influence the technical efficiency of public libraries.

Our baseline empirical results are presented in Table 4.5. Column 1 has the results esti- mated using Ordinary Least Squares (OLS) with no year and county fixed effects. Columns 2 and 3 include year, and year and county fixed effects respectively. Column 4 and 5 estimate the model using Weighted Least Squares (WLS) with personal income per capita used as the weight. The WLS estimates also include either year or county and year fixed effects. Demographic control variables are included, but we only report those statistically significant. County governments are the excluded type of jurisdiction, thus city and multi-jurisdictional should be interpreted as in reference to county libraries. For demographic controls we exclude population with less than 15 years-old and percent of white population. Standard-errors are clustered by county.

Looking at our results, the first thing that is apparent is that the source of public funding, i.e. local, state or federal, is not correlated to the level of technical efficiency. The exception is a positive and statistically significant correlation for state funding when controlling for time and county fixed effects. Turning to our governance variables, we find across all five specifications that compared to library systems that are governed by county governments, city governance is negatively related to TE. Depending on the specification, a city governed

7_{Special laws are laws initiated locally and passed by the Legislature. These laws set aside part of the}

tax dollars from local government (Board of Education, County or City) to public libraries. Some of special laws have been in place since 1930s and the most recent is from 1994. However, since 2013 it is no longer a mandatory that Board of Educations to continue providing funding to public libraries. For more on these special laws, see West Virginia Library Commission (2013).

library is between 5 to 11 percentage points less technically efficient, ceteris paribus. We find no statistically significant results for multi-jurisdictional libraries. In addition, libraries op- erating under increasing returns to scale are on average 16 percentage points less technically efficient than those which are not. This result is consistent across the different specifications. Finally, we focus on demographic characteristics. In terms of high-school graduation rate, public libraries in counties with higher graduation rates are associated with lower technical efficient. However, when controlling for county unobserved characteristics these results are not different from zero. In addition, after controlling for county and year fixed effects, the results imply that libraries in counties with higher multi-racial population are also technically less efficient, while those with higher population between 45-64 years old are technically more efficient.

One concern about our results in Table 4.5 is that our dependent variable is bounded between 0 and 1 and therefore the normality assumption of OLS is not valid. To deal with this issue we employ the fractional logit method with and without county and year fixed effects. These results are presented in Table 4.6. Both columns show the same specifications as in Table 4.5, with Column 1 excluding county and year fixed effects and Column 2 including year fixed effects, and Columns both year and county fixed effects. The results corroborate our previous analysis showing a negative and statistically significant relationship between technical efficiency and city governance. The results for increasing returns to scale libraries, libraries in larger multi-racial counties, and with larger population within 45-64 year-old are also maintained.

Unlike in our OLS and WLS regressions, however, we find a negative association between technical efficiency and local funding, and positive association with state funding. Our em- pirical approach does not allow us to speak to causality on the funding source/technical efficiency relationship. It could be that having larger share of the budget from state gov- ernment lead to public libraries being more technically efficient. We cannot, however, rule out reverse causality. Our results could reflect that more technically efficient libraries are more likely to receive state funding, especially in light of matching requirements from local funding. Similarly we cannot rule out that technically inefficient public libraries are more likely to garner higher levels of local financial support vis-a-vis the receiving larger share

Amir B. Ferreira Neto Chapter 4. Technical Efficiency of Library Systems 73 from local governments leading to more inefficiency.

Additionally, having branches is associated with larger technical efficiency, as well as larger black population. However, latin population and population between 15 and 24 years- old are associated with lower technical efficiency. These results, however, are not statistically different from zero when controlling for county fixed effects.