Adoption of Technological Change in the Public Sector: Evidence From US States

Abstract

Our paper explores the broad influences that stimulate technological change in governmental service delivery. Using panel data by state, we examine whether residents are able to apply for Unemployment Insurance (UI) benefits via technology services such as the internet or automated telephone, or are required to apply in person. The reduced form model tests for the relative influence of residential demand, political pressure, and bureaucratic influence. We find that it is the dispersion of the urban population across a state that provides the impetus for government adoption of new technology, in stark contrast to the importance of urban concentration found for the private sector. A unique additional influence we test is the ability of the entrenched bureaucracy to impede technology options. We find that governors of either political party in their term-limited term—when compromise with bureaucrats is less important—save 4% in administrative costs. We find that technology adoption can be delayed but not prevented by bureaucratic interests.

Keywords

city size and city systems economic analysis economic growth and development local public finance methods other policy and applications policy and applications regional econometric models spatial structure urban and regional economic development urban and regional spatial structure

Introduction

Urban locations are becoming well known for spurring technological innovation in the private sector (Henderson 2010; Mansfield 1961).¹ It therefore seems reasonable to explore whether we should expect the same forces to influence the public sector. More specifically, as the demand for new technology grows, and as the conditions for developing new technologies expand, then to the extent that the public sector reflects the demand by residents, it is natural to expect that the use of innovative technology would occur in the public sector as well. While the impact and speed of technological development on the business community has been widely studied (Bertinelli and Black 2004; Henderson 2003), our question in this paper is whether urbanization-influenced technological development has occurred in the provision of government services as well.

Our analysis will therefore explore how demand in the public sector for using new technologies might be different than demand in the private sector. Further, there are institutional differences that impact the public sector’s willingness to supply, or develop, new technologies that are distinct from private sector institutions. A key distinction between private demand and demand in the public sector is that public sector demand by constituents is filtered through the political process. Thus, the primary innovation in our work is to examine not simply the determinants of public sector demand, but to consider the processes by which that demand is communicated to politicians. Since urban concentrations have been the most important spur to technological progress in the private sector, we apply special attention to urban concentrations. We empirically find an important difference between the public and private sectors, which is that the dispersion of the urban population is more important in the public setting. We believe this result is consistent with the political environment, where in a legislative setting the broad support from legislators is needed to pass policy proposals.

In virtually all advanced representative democracies, decisions are undertaken by a combination of legislatures and members of the executive branch. One of the long-time research objectives has been to examine the extent to which this modern governmental form reflects the demand of residents. Without explicitly addressing this literature, we nonetheless consider its lessons—that some of the structure of governmental form may impact the public policy choices.

Thus a potential distinction between the public and private sectors is that there is a wedge between demand and implementation in the public sector, which is that public policy is implemented by a government bureaucracy. That is, politicians hire bureaucrats when tasks are complicated enough to require specific skills (Banks and Weingast 1992; Tummers and Rocco 2015). The problem for politicians is that bureaucrats have their own agenda that does not necessarily coincide with the desires of politicians or residents (Craig, Hoang and Kohlhase 2021; Ujhelyi 2014). Thus, the political environment provides a platform where the employees that may be replaced by new technologies have a forum to forestall the process (Alesina and Tabellini 2007; Epstein and O’Halloran 1994). We test for several different avenues by which the interests of the government bureaucrats may not be identical to either that of the politicians for whom they work, or the residents that ultimately receive their services.

Our paper focuses on a single governmental program, Unemployment Insurance (UI). A single program has the benefit of measurable activities, and specific technological changes. The administrative change in the UI program is the transition from a traditional in-person application process to using automated telephone and on-line technologies. In the traditional system, a newly unemployed worker traveled to a government office, and was assisted by a government employee in the application for UI benefits. The initial change in technology by state governments was to design systems where applicants could apply by automated telephone without use of government administrators; more recently, the technology has been expanded to include on-line applications.

A further benefit of utilizing UI to explore technological change is that every state has wide discretion over the operation of its UI program, although under a federal programmatic umbrella (Craig and Palumbo 1999). Each state finances basic UI benefits by setting its own tax rate as well as eligibility rules and benefit schedules. The specific choices in each state’s UI system allow us to estimate how variation in public sector attributes impacts the adoption of the new UI application technologies.

Our initial hypothesis—consistent with what has been found in the private sector—is that urban residents will express demand for new technologies earlier than rural residents, so that more urbanized states would be expected to have both greater advocacy for and greater utilization of the new UI application technologies.² To test for the political influence of urban residents, we therefore carefully consider the distribution of the urban population, and find stark differences compared to the private sector. Henderson (2003) for instance explores the importance of the distribution of urban residents as a driver for technological change in the private sector, and finds that the largest city in an economy (state) drives technological change, a process he terms “primacy.”³ In contrast, we find that a more dispersed urban population across a state leads to faster governmental innovation than does a larger share of the population in the largest city. This finding we believe is consistent with the political application of influence.⁴ The other surprising finding, also consistent with a distinct public sector influence, is that we find that internet access is more important than simply an urban population. Internet access suggests personal benefit is higher—irrespective of the location of the residential constituents, and thus is found to be a more direct influence on public sector technological adoption behavior.

The third political attribute distinct from the private sector, and one consistent with the term limit finding above, is the impact of the state bureaucracy. We explore bureaucratic influence by two different methods. In our first method, the reduced form demand equation shows that measures related to the ability of the bureaucracy to influence decisions have significant impacts. We use two measures of potential bureaucratic power: the extent of general public sector unionization and the size of the state capital city relative to the urban population. These variables generate an empirical test providing a rare opportunity to discern the relative bargaining power of public sector employees compared to the power of politicians, who presumably reflect the desires of the electorate.

The second method we employ to assess differences in technological adoption in the public compared to the private sectors is a separate analysis of administrative costs. The federal government pays the entirety of UI administrative costs. We nonetheless analyze how those costs vary by state with respect to the political environment. Because bureaucrats are a significant share of administrative costs, cost variation is a different window on the relative strength of the bureaucracy compared to the expression of residential demand through the political structure. Our budget investigation provides further evidence that the relative influence of the bureaucracy explains some of the variation in the rate of technological adoption.

The rest of our paper is organized as follows. Section II discusses the empirical specification of the reduced form demand equation, specified to include the influence of the urban population, the political structure, and bureaucratic influence. Section III describes the panel data and discusses the data definitions. Administrative data from the US Department of Labor (DOL) on UI program administration is combined with Census data and other sources to describe the political and demographic environment of each state. Section IV presents the empirical results which illustrate the several ways in which public sector adoption of technology significantly differs from the private sector. Section V summarizes and interprets the political elements that cause distinct differences in the determinants of public sector technological adoption.

Empirical Model

Our empirical model is a reduced form that captures the essential elements of governmental choice in the delivery of a specific service, determination of UI eligibility. Our objective is to discern the drivers of technological adoption, and then illustrate their relative importance in a public sector setting in a reduced form empirical panel-data model. In the political context of our model, we examine differences in the distribution of urban residents throughout the state, other components of residential demand, the influence of political structure, and measures of the effectiveness of bureaucratic influence. As we will empirically show below, there are very distinct differences that motivate the public sector compared to the private sector over technological change.

The labor departments from each state screen UI applicants to determine whether an unemployed person is eligible for government UI assistance. The basic purpose of UI is to assist the “involuntarily unemployed,” rather than people who choose to separate from their employment. The details of this determination are important, and vary substantially between each state.⁵ Stringency in determining voluntary unemployment is one dimension. Attachment to the workforce is another complex dimension where rules vary widely between state UI systems. For example, the definition of part-time workers and their eligibility varies. Other sources of eligibility variation include the length of time spent working, and time between unemployment spells.

Once it is determined whether an unemployed worker is eligible for UI, the second aspect of the UI program is to determine the level of income support payments. This amount is paid per week for a certain number of weeks, the maximum duration for standard UI is 26 weeks.⁶ The level of weekly benefits depends on wages and other forms of compensation. For example, tips, bonus pay, and overtime pay are treated differently across states. The difficulty in navigating the complexities of the program will affect a potential recipient’s willingness to use the new technologies to apply for UI, and may affect the ability of the local governments to develop the new technologies for service delivery.

The new technologies for UI service delivery allow an applicant to apply for UI either by automated phone or by using the internet, both of which we refer to as “on-line.” We generally aggregate the two forms because we believe resistance to change comes from the threat to the administrative UI workers. Clearly, the state government has to construct the infrastructure so that potential recipients can use the new technologies. The actual architecture and the extent to which it is “user-friendly” is determined by each state. Once the infrastructure is in place, the state government has the choice of whether to leave the traditional offices open, and in what locations. While we do not have the specific information to measure these two decisions, we nonetheless have the result from the policy choices. That is, we can measure the share of recipients that apply for UI either in the traditional manner by presenting themselves to an UI office in person, or by using a newer technology.

Our estimating equation examines the extent to which potential recipients use either the traditional in-person UI office, or the available on-line systems to apply for UI:

Y_{s, t} = β_{0} + {β_{1} {U r b a n}_{s, t} + β}_{2} O t h e r {R e s i d e n t i a l_D}_{s, t} + β_{3} {B u r e a u c r a c y}_{s, t} + β_{4} {P o l i t i c s}_{s, t} + μ_{s} + γ_{t} + ε_{s, t}

(1)

where

Y_{s, t}

is one of the three ways to measure outcomes that we examine for each state s at time t. The primary outcome is the intensity of use of the new technologies, as measured by the share of UI applicants that use the newer application technologies.⁷ The second and third outcomes are the fixed and variable costs per capita paid by the federal government for the total administrative costs. While institutionally costs would not be expected to influence state governments, the analysis of costs allows consideration of political choices in other programmatic contexts.

The decision to allow recipients to apply for UI benefits using new technologies reflects a reduced form, as state governments need to select whether to offer non-traditional options for UI application by developing the technological infrastructure, and potential recipients need to choose whether to use the newer forms of application. We separately explore the four sets of attributes that impact governmental choice over how technology gets used in a state.

As discussed above, urbanization might impact governmental choice as in the private sector. The distinction from the private sector that we will explore is that the demand by urban residents is expressed through elected representatives. The distribution of the urban population is used to integrate the reality of politics into the decision by state governments to technically innovate. Therefore, in addition to the size of the urban population, we measure the regional distribution of the urban population using two alternative measures.⁸ One is the share of the urban population in a state’s largest city; the second is a Herfindahl index of city population shares.⁹ We specify the first variable to capture the potential of the largest city to have out-sized influence on policy corresponding to the primacy measure from Henderson. The Herfindahl index, in contrast, takes into account the entire urban spectrum across all the cities of the state. The estimated impacts of both variables are found to be similar, and both yield the result consistent with political expectations that dispersion aids the adoption process.

We explore two other variables in the vector Urban that might be correlated with residents’ familiarity to new technologies. Broadband connectivity is empirically found to be an important correlate with new technology adoption. We find it is correlated with, but dominates, the straightforward urban measure.¹⁰ We also add patents per capita as a potential measure of the importance of research.¹¹ Since the model is reduced form, both measures may possibly measure costs as well as demand, and thus we interpret their coefficients as the net effect of the two processes.

The second set of attributes OtherResidential_D. includes socio-demographic factors which are used as control variables. For example, higher incomes might indicate a greater taste for and ability to use new technology (Audirac 2005), but might also indicate demand for personal attention such as offered by using in-person applications. Therefore, we include Gross State Product per capita as a measure for income, along with demographic variables, such as the share of the state population that is white, the share of the population working in the manufacturing sector, and the share of the population with a high school degree. Perhaps not surprisingly states with younger populations are found to be more likely to implement new technologies earlier than others.

The third vector in equation (1) is Bureaucracy, which consists of measures to explore the potential strength of resistance that might be offered by government workers. Unlike the private sector, the public administrators already employed within the government have considerable employment protections (Ujhelyi 2014). Because the new technologies ostensibly involve less labor than the traditional form of using a government worker to collate and present the information from a UI applicant, public sector workers may feel threatened. We utilize two measures in an attempt to control for how bureaucratic influence varies between states, with the hypothesis that stronger bureaucratic influence will slow technological change.

Public sector unionization is a potential indicator of bureaucratic strength. The state legislature is able to write laws that can profoundly impact the degree of public sector unionization. States where unionization is high may therefore be more interested in protecting employment of government workers.¹² Bureaucratic power can be thought of as resulting from a bargaining model inherent in the principal-agent allocation to bureaucracies that politicians face. In such a model, unions can be expected to be stronger and, therefore, more likely to impede rapid deployment of new labor-saving technologies. Thus, a more unionized public sector work force might be expected to delay the transition to newer service technologies, or even prevent them altogether.¹³

We construct as the second measure of bureaucratic strength the degree to which the state capital city is relatively large. Our hypothesis is that holding constant the overall level of urbanization, the state with more regulation and hence a greater demand for government employees is one where government employees are likely to be more influential.¹⁴ If administrators feel threatened by labor-saving technologies, therefore, we would expect that a larger share of the urban population that is located in a capital city would lead to slower adoption of the new technologies. We view the share of the urban population in the capital city as a separate measure from overall urbanization; the share variable solely reflects the concentration within the capital city as opposed to other areas within the state.

The final set of attributes we examine is the political party composition of the institutions of government, indicated by the variable Politics in equation (1). We test for the political party of the governor since Democrats have generally been found to be more supportive of government workers, and we test for the party composition of the state legislature. In some contexts, it has been found that preferences by the governor are more likely to be observable when the Governor is in the last term of office in states where term limits have been imposed (Besley and Case 1995; Smart and Sturm 2013). ¹⁵ We test for the possibility and find supportive evidence.

Unlike our success with the governor’s party, however, we were unable to empirically find any meaningful way in which the party composition of the legislature is useful. We tried party shares of each of the legislative houses. We also explored a designation for when a single party controls both houses, and additionally included whether the legislative parties are aligned with the governor. None of the legislative party composition variables, however, are fruitful.¹⁶

Data

The data for our paper is a panel of all 50 US states over the period 1996 through 2011. Table 1 reports the descriptive statistics and variable definitions for the variables describing the state unemployment insurance programs and other state characteristics. The variables in the reduced form model are meant to capture the forces of citizen demand including urbanization, governmental structure, and bureaucratic strength.

Table 1.

Descriptive Statistics for US State Unemployment Insurance Programs 1996-2011.

	1996		2011		1996-2011^d
Dependent variables	Mean	Std	Mean	Std	Mean	Std
Unemployment insurance variables
% Technology Claims (not in-person)^a	6.4	18.9	86.2	23.8	58.4	41.5
State level percentage of UI claims filed via automated telephone or internet % Internet Claims	0.0	0.0	50.2	26.7	20.0	27.0
State level percentage of UI claims filed via internet % Automated Telephone Claims	6.4	18.9	36.0	26.4	38.4	35.6
State level percentage of UI claims filed via automated telephone system Total Admin UI Cost per capita (2011 $)^b	12.8	6.3	13.8	9.2	12.1	5.9
State level amount of UI administrative costs per capita in 2011 dollars Fixed UI Admin Cost per capita (2011 $)^b	4.7	1.3	4.7	6.0	4.7	1.9
State level amount of fixed UI administrative costs per capita in 2011 dollars Variable UI Admin Cost per capita (2011 $)^b	5.5	2.1	6.1	1.9	5.7	2.3
State level amount of variable UI administrative costs per capita in 2011 dollars
Independent variables	Mean	Std	Mean	Std	Mean	Std
Residential demand variables % Urban	71.3	14.7	73.8	14.5	72.4	14.6
State level percentage of residents residing in urban areas % Total Urban Pop in Largest City	29.3	22.0	27.2	17.1	28.0	18.3
State level percentage of urban residents residing in the state’s largest city based on population size Herfindahl index of City size	0.1	0.1	0.14	0.14	0.1	0.1
State level measure of dispersion of the urban population.
Broadband Connections per capita^c	––	––	0.8	0.1	0.3	0.2
State level number of residential and business high speed broadband connections per capita
Patents per capita	0.0002	0.0001	0.0004	0.0003	0.0003	0.0002
State level number of patents per capita
	1996		2011		1996-2011^d
Independent variables	Mean	Std	Mean	Std	Mean	Std
Political and bureaucracy variables Governor Term limit	24.0	43.1	16.0	37.0	31.5	46.5
State level dummy variable indicating whether the Governor’s current term is term limited
Democrat governor	34.0	47.9	40.0	49.5	43.2	49.6
State level dummy variable indicating whether the Governor’s political party is Democrat
Democrat Control	40.0	49.5	54.0	50.3	42.0	49.4
State level dummy variable indicating whether the state’s legislature is controlled by Democrats
Split Control	24.0	43.1	0.16	37.0	21.7	41.2
State level dummy variable indicating whether the state’s legislature is evenly controlled by Democrats and Republicans
% Public sector workers unionized	34.4	18.0	33.7	18.8	33.4	17.7
State level percentage of public sector workers that are unionized
% Total Urban Pop in Capital City	13.9	15.3	12.8	12.0	13.1	12.9
State level percentage of urban residents residing in the state’s capital city
Socio-demographic variables
Unemployment rate (%)	5.1	1.1	8.1	2.0	5.4	2.0
State level percentage of the labor force that is unemployed
GSP per capita (2011 $)^b	40,935	6,852	52,130	12,014	46,348	9445
Gross State Product per capita in 2011 dollars
% of workers in manufacturing	13.8	5.2	8.9	3.4	11.2	4.6
State level percentage of workers that are in the manufacturing sector
Poverty rate (%)	12.8	4.0	14.2	3.1	12.3	3.3
State level percentage of households with income below the poverty line
% Young	26.1	2.0	25.1	2.2	25.2	1.9
State level percentage of residents that are under the age of 18
% Old	12.8	2.0	12.8	2.0	12.7	1.8
State level percentage of residents that are over the age of 64
% white	79.3	14.1	77.8	16.2	76.4	15.2
State level percentage of residents that are white
% College	15.4	2.9	21.1	3.8	18.3	3.9
State level percentage of residents that have a college degree
% HS and over	83.2	4.3	89.1	2.9	86.5	4.1
State level percentage of residents that have at least a high school degree

^aTechnology claims include by automated telephone and by the internet. There is a small other category for appeals and other non-standard processes.

^bAll dollar values are in real 2011 dollars.

^cData on broadband connectivity does not start until 2000.

^dThe data reported here are pooled across states and time, from 1996 to 2011, totaling 800 observations. The individual year means include 50 observations. For the tables estimating shares of new technology usage, the data employed cover 2002-2011.

A key data element is the technology used by UI applicants, available from the US Department of Labor (DOL). There are three possible technologies to make unemployment insurance claims: in-person, automated telephone, and internet. Generally, the phone application process was started prior to the time when applicants could apply by internet. Both of the newer technologies pose a potential threat to public sector labor interests, so we present results for the use of any new technology, as well as the individual technologies. The key findings are presented through examination of the share of UI applications made through an on-line structure compared to the traditional in-person office using DOL data.

The other administrative output data we briefly explore are the administrative cost data, also available from the DOL. The cost data are only for total costs; however, below we decompose the total into fixed and variable costs. Dollar values are deflated using the CPI in 2011 dollars.

We expect that urban populations disproportionately desire the newer technology for UI applications, so two different variables are constructed that capture how the dispersion of the urban population influences the political process. Our first measure of dispersion develops a Herfindahl index based on city size (see footnote 11). The alternative measure we also try, and which is about equally successful, is the share of the state’s population in the largest city (analogous to Henderson’s primacy variable).¹⁷

The remaining independent variables comprise the vector OtherResidential_D. They are demographic characteristics of the states’ residents, including variables taken from the Census Bureau such as the share of the population over age 65, under 18, white, with a high school or higher degree, with a college or higher degree, the unemployment rate, and the poverty rate. Also, we include the share of the population that works in the manufacturing sector, which is taken from the DOL, and Gross State Product per capita from the Bureau of Economic Analysis.

We further include two variables potentially related to technology demand: the number of patents granted in each state per year from the US Patent and Trademark Office, and broadband penetration in each state using data from the Federal Communication Commission (FCC).¹⁸ The FCC data reports connections at the state level starting in 2000. Both variables are multi-faceted, our reduced form specification captures the net from all effects. For example, patents reflect legal and industrial characteristics as well as innovation. Broadband penetration can result from either supply or demand factors.

We construct two variables linking the political environment to bureaucratic desire to protect their interests as indicated by Bureaucracy in equation (1). The first is the relative share of the urban population that resides in the state capital city. Greater government regulation is generally associated with not only a larger public sector labor force, but may also indicate a greater belief in the abilities of the public labor force.¹⁹ If so, it may be that greater employment in the capital city of a state is associated with greater influence of the bureaucracy. To the extent governmental importance is correlated with bureaucratic power, the inclusion of the relative dominance of the capital cities by their share of the urban population in the state will be correlated with the ability of the bureaucracy to alter decisions to protect their (perceived) interests.

A second variable related to the relative strength of the bureaucracy is the extent to which the public sector is unionized. These data are collected by Hirsch, Macpherson and Vroman (2001).²⁰ While not specific to UI workers, it is possible that the extent of public sector unionization is an indicator of the extent to which the interests of government workers are weighted in the decisions by the state government.

Balancing the desires of the bureaucracy are the politicians. We obtain political variables from The Book of the States (annual), including a dummy variable if the governor of the state is a Democrat, whether that governor is in their last term due to term limits, and as well the party composition of each house of the legislature. We find that the party that controls both houses of the legislature is more important than when the parties split control of the two legislative houses, although there is little overall contribution from these variables.²¹

Estimation Results on Technology Adoption in Government

The literature on technology in the private sector is clear that urban areas both stimulate the creation of new technologies as well as embrace its use (Bertinelli and Black 2004; Capello et al. 2017; Henderson 2003). Our primary test is therefore to examine how urban areas impact the usage of new technologies to apply for UI in the public sector. The distribution of the urban population within a state is important because of the impacts on political choices. Thus, we examine not only the share of the urban population in each state, but also whether the urban population is narrowly or widely distributed. While concentration seems important to private outcomes, we show below that it is the broader distribution of population that speeds technological adoption in the public setting.

The other contribution of our work is to document the real impediment to technical change imposed by government workers––those who feel threatened by technical change. We find that consistent with their incentives, stronger bureaucracies impede technological innovation. Despite the impediments, however, the technical change nonetheless occurs.

Sample Issues

A necessary precursor to discussing our results about technical change is to discuss the sample. Specifically, we explore the sample selection issues that are raised by considering which states offer the new UI application technologies. Unfortunately, however, a complete analysis is not possible with the available data. The data on the method for UI application is only available starting in 1996, by which time 22 states had already initiated a new technology. All states had some form of new application technology by 2002. Despite the limitations, we attempt to employ dichotomous methods to examine the states’ decisions to initiate a new technology for the period 1996-2001. Since the analysis only captures the late adopters, we find the results are not informative, but nonetheless are reported in the Appendix. Similarly, we test our continuous model results for potential sample selection bias using all the data from 1996-2011 (Heckman 1979).²² We find no evidence of selection bias, and so relegate the analysis to the Appendix as well.²³ Therefore our final data set uses data covering the period 2002-2011 when all states had adopted the new systems, and we focus on the intensity of new-technology usage.

Usage of Automated Telephone and Internet Applications

Table 2 presents the core statistical analysis of the share of UI applications that are made on-line from 2002 to 2011. Our data start in 2002 because all states offered at least one of the new technology application methods by then.²⁴ The specification is in natural logarithms, so the estimated coefficients can be interpreted as elasticities.²⁵ We present the results in three tables, Tables 2–4. Table 2 compares any new technology to the traditional administrative method of in-person application. We find that the aggregate results (where we group automated phone and internet applications) are the sharpest, suggesting that the form of the technology is less important than the threat to the traditional administrative structure.²⁶ Nevertheless, for completeness we present results for the disaggregated new technologies as well, the ability to apply for UI by internet in Table 3, and by automated telephone in Table 4.

Table 2.

Analysis of Technology Usage (Internet and Automated Phone) for UI Claims.

	Period: 2002-2011 Dependent variable: ln % UI Technology Claims
Independent variables	(1)	(2)	(3)	(4)	(5)	(6)
ln % Urban	5.38 (4.09)	7.48* (4.06)	6.50 (4.32)	2.88 (4.40)	4.98 (4.37)	3.89 (4.68)
ln % Urban Pop in largest City	−3.19*** (0.89)		−1.58 (2.45)	−3.20*** (0.91)		−1.77 (2.48)
ln City Herfindahl		−4.07*** (1.22)	−2.19 (3.47)		−4.04*** (1.24)	−1.95 (3.50)
ln % Urban Pop in Capital City	−2.77* (1.51)	−2.73* (1.52)	−2.79* (1.52)	−2.08 (1.47)	−2.04 (1.49)	−2.10 (1.48)
ln % Public sector workers unionized	−0.68** (0.34)	−0.62* (0.35)	−0.65* (0.36)	−0.58* (0.34)	−0.52 (0.34)	−0.55 (0.35)
ln Broadband Connectivity per capita				0.51* (0.27)	0.50* (0.27)	0.51* (0.27)
Gov Term limit	−0.05 (0.13)	−0.05 (0.13)	−0.05 (0.13)	−0.04 (0.13)	−0.04 (0.13)	−0.04 (0.13)
Dem Gov	−0.13 (0.09)	−0.12 (0.09)	−0.13 (0.09)	−0.16* (0.09)	−0.15 (0.09)	−0.15 (0.09)
(Gov Term Limit)*(Dem Gov)	−0.34** (0.17)	−0.34** (0.17)	−0.34** (0.17)	−0.38** (0.17)	−0.38** (0.17)	−0.38** (0.17)
ln Patents per capita	−0.67*** (0.25)	−0.69*** (0.25)	−0.68*** (0.25)	−0.59** (0.24)	−0.61** (0.24)	−0.60** (0.24)
ln % Manuf workers	−4.03*** (0.98)	−3.98*** (0.99)	−4.02*** (0.99)	−4.01*** (0.97)	−3.95*** (0.98)	−4.00*** (0.98)
ln % Unemp	−0.15 (0.27)	−0.14 (0.27)	−0.14 (0.27)	−0.08 (0.27)	−0.07 (0.27)	−0.07 (0.27)
ln real GSP per capita	−1.19 (0.95)	−0.97 (0.95)	−1.07 (0.92)	−1.27 (0.97)	−1.05 (0.97)	−1.17 (0.94)
ln % Poverty	−0.27 (0.39)	−0.31 (0.39)	−0.29 (0.39)	−0.28 (0.38)	−0.32 (0.38)	−0.30 (0.38)
ln % Young	2.12** (1.06)	2.14** (1.06)	2.14** (1.06)	2.13** (1.05)	2.14** (1.06)	2.14** (1.05)
ln % Old	0.16 (0.70)	0.13 (0.71)	0.14 (0.70)	0.22 (0.71)	0.19 (0.72)	0.20 (0.71)
ln % white	0.20 (1.07)	0.02 (1.09)	0.09 (1.06)	−0.20 (1.09)	−0.37 (1.10)	−0.30 (1.07)
ln % College	0.51 (0.75)	0.48 (0.75)	0.50 (0.75)	0.49 (0.76)	0.46 (0.76)	0.48 (0.76)
ln % HS and Over	−0.34 (3.21)	−0.05 (3.20)	−0.22 (3.20)	−0.26 (3.22)	0.05 (3.22)	−0.15 (3.22)
Observations	500	500	500	496	496	496
R-squared	0.74	0.74	0.74	0.75	0.75	0.75

Notes: Robust standard errors in parentheses. Significance levels: *: 10%; **: 5%; ***: 1%. This table is estimated using data for 50 US states and the period 2002-2011. For columns 4-6, Hawaii is omitted for 2002-05 due to missing broadband data. The dependent variable is the sum of the shares of UI internet claims and UI automated telephone claims. All states adopted both or either of the new technologies (internet and automated telephone) by 2002. State and year fixed effects are included in the regressions.

Table 3.

Analysis of Technology Usage (Only Internet) for UI Claims.

	Period: 2002-2011 Dependent variable: ln % UI Internet Claims
Independent variables	(1)	(2)	(3)	(4)	(5)	(6)
ln % Urban	0.82 (6.44)	0.74 (6.67)	0.27 (7.15)	0.32 (6.45)	0.21 (6.71)	−0.22 (7.24)
ln % Urban Pop in largest City	0.02 (1.25)		−0.60 (3.43)	0.05 (1.25)		−0.56 (3.43)
ln City Herfindahl		0.13 (1.63)	0.85 (4.48)		0.16 (1.63)	0.83 (4.49)
ln % Urban Pop in Capital City	0.05 (2.17)	0.08 (2.14)	0.04 (2.18)	0.70 (2.16)	0.73 (2.14)	0.69 (2.16)
ln % Public sector workers unionized	−0.49 (0.34)	−0.49 (0.34)	−0.50 (0.35)	−0.42 (0.34)	−0.43 (0.34)	−0.43 (0.34)
ln Broadband Connectivity per capita				0.44 (0.31)	0.44 (0.31)	0.44 (0.31)
Gov Term limit	0.04 (0.18)	0.04 (0.18)	0.04 (0.18)	0.04 (0.17)	0.04 (0.17)	0.05 (0.17)
Dem Gov	0.09 (0.13)	0.09 (0.13)	0.09 (0.13)	0.08 (0.13)	0.08 (0.13)	0.07 (0.13)
(Gov Term Limit)*(Dem Gov)	−0.36 (0.22)	−0.36 (0.22)	−0.36 (0.22)	−0.35 (0.21)	−0.35 (0.21)	−0.35 (0.22)
ln Patents per capita	−0.97** (0.42)	−0.97** (0.43)	−0.97** (0.42)	−0.90** (0.41)	−0.90** (0.41)	−0.90** (0.41)
ln % Manuf workers	0.57 (1.67)	0.60 (1.68)	0.59 (1.68)	0.57 (1.64)	0.60 (1.65)	0.59 (1.65)
ln % Unemp	−0.02 (0.40)	−0.02 (0.40)	−0.02 (0.40)	0.08 (0.41)	0.08 (0.40)	0.08 (0.41)
ln real GSP per capita	−1.45 (1.17)	−1.46 (1.16)	−1.51 (1.19)	−1.48 (1.20)	−1.49 (1.19)	−1.54 (1.22)
ln % Poverty	−1.33*** (0.49)	−1.33*** (0.49)	−1.33*** (0.49)	−1.29*** (0.48)	−1.29*** (0.48)	−1.29*** (0.48)
ln % Young	2.05 (1.43)	2.04 (1.43)	2.05 (1.44)	2.01 (1.43)	2.00 (1.43)	2.01 (1.44)
ln % Old	−0.86 (1.10)	−0.86 (1.10)	−0.85 (1.10)	−0.84 (1.09)	−0.84 (1.09)	−0.83 (1.09)
ln % white	0.36 (1.37)	0.38 (1.38)	0.42 (1.38)	−0.10 (1.40)	−0.08 (1.39)	−0.04 (1.39)
ln % College	0.84 (0.93)	0.84 (0.93)	0.85 (0.93)	0.85 (0.93)	0.86 (0.92)	0.86 (0.93)
ln % HS and Over	−2.47 (4.49)	−2.44 (4.49)	−2.48 (4.49)	−2.66 (4.48)	−2.64 (4.47)	−2.67 (4.48)
Observations	437	437	437	437	437	437
R-squared	0.78	0.78	0.78	0.78	0.78	0.78

Notes: Robust standard errors in parentheses. Significance levels: *: 10%; **: 5%; ***: 1%. This table is estimated using data for 50 US states and the period 2002-2011. For columns 4-6, Hawaii is omitted for 2002-05 due to missing broadband data. The dependent variable is the share of UI internet claims. As discussed in the main text, not all states have adopted internet systems for all years, see Appendix Table A2a for a comparable selection model. State and year fixed effects are included in the regressions.

Table 4.

Analysis of Technology Usage (Only Automated Telephone) for UI Claims.

	Period: 2002-2011 Dependent variable: ln % UI Automated Telephone Claims
Independent variables	(1)	(2)	(3)	(4)	(5)	(6)
ln % Urban	2.45 (4.27)	3.37 (4.26)	0.63 (4.68)	1.07 (4.53)	1.97 (4.53)	−0.82 (5.01)
ln % Urban Pop in largest City	−1.94 (1.32)		−4.57* (2.76)	−1.92 (1.29)		−4.64* (2.75)
ln City Herfindahl		−1.60 (1.98)	3.50 (4.12)		−1.56 (1.96)	3.62 (4.15)
ln % Urban Pop in Capital City	−3.23 (1.97)	−3.18 (2.00)	−3.16 (1.95)	−2.90 (1.88)	−2.86 (1.92)	−2.82 (1.86)
ln % Public sector workers unionized	−0.75** (0.38)	−0.71* (0.38)	−0.80** (0.39)	−0.70* (0.37)	−0.66* (0.37)	−0.76** (0.38)
ln Broadband Connectivity per capita				0.25 (0.25)	0.25 (0.25)	0.26 (0.25)
Gov Term limit	−0.06 (0.13)	−0.05 (0.13)	−0.06 (0.13)	−0.04 (0.13)	−0.04 (0.13)	−0.05 (0.13)
Dem Gov	−0.04 (0.09)	−0.04 (0.10)	−0.05 (0.10)	−0.05 (0.10)	−0.05 (0.10)	−0.06 (0.10)
(Gov Term Limit)*(Dem Gov)	−0.29* (0.17)	−0.29* (0.17)	−0.29* (0.17)	−0.32* (0.18)	−0.32* (0.18)	−0.31* (0.18)
ln Patents per capita	0.60* (0.31)	0.59* (0.31)	0.60** (0.30)	0.65** (0.32)	0.64** (0.32)	0.65** (0.31)
ln % Manuf workers	−5.00*** (1.16)	−4.99*** (1.17)	−5.00*** (1.15)	−4.96*** (1.16)	−4.94*** (1.18)	−4.95*** (1.15)
ln % Unemp	−0.59* (0.33)	−0.60* (0.33)	−0.60* (0.33)	−0.54 (0.33)	−0.55* (0.33)	−0.55* (0.33)
ln real GSP per capita	−1.09 (1.02)	−1.09 (1.00)	−1.26 (0.98)	−1.13 (1.02)	−1.13 (1.00)	−1.31 (0.99)
ln % Poverty	−0.47 (0.35)	−0.49 (0.34)	−0.43 (0.35)	−0.46 (0.35)	−0.48 (0.34)	−0.42 (0.35)
ln % Young	1.27 (1.32)	1.23 (1.32)	1.24 (1.32)	1.27 (1.32)	1.23 (1.33)	1.24 (1.33)
ln % Old	−0.91 (1.00)	−0.90 (1.00)	−0.88 (0.99)	−0.92 (1.03)	−0.91 (1.03)	−0.89 (1.02)
ln % white	−0.31 (1.17)	−0.24 (1.16)	−0.15 (1.14)	−0.53 (1.20)	−0.45 (1.18)	−0.37 (1.16)
ln % College	0.62 (0.89)	0.61 (0.90)	0.64 (0.89)	0.55 (0.94)	0.55 (0.94)	0.58 (0.93)
ln % HS and Over	1.34 (3.64)	1.43 (3.68)	1.20 (3.64)	1.63 (3.72)	1.72 (3.76)	1.47 (3.72)
Observations	490	490	490	486	486	486
R-squared	0.85	0.85	0.85	0.85	0.85	0.85

Notes: Robust standard errors in parentheses. Significance levels: *: 10%; **: 5%; ***: 1%. This table is estimated using data for 50 US states and the period 2002-2011. For columns 4-6, Hawaii is omitted for 2002-05 due to missing broadband data. The dependent variable is the share of UI automated telephone claims. As discussed in the main text, not all states have adopted automated telephone systems for all years, see Appendix Table A2b for a comparable selection model. State and year fixed effects are included in the regressions.

The first three columns of Table 2 weakly suggest that urban citizens demand, as expected from private sector analyses, more non-traditional technologies. The last three columns, however, suggest that residents who purchase broadband connections are the driving force for demand by urban residents. Because we have data on the distribution of the urban population, we use the significant correlation between urban residents and broadband connectivity to further explore the distribution of the urban population for its political implications.²⁷

The results on the regional distribution of the urban population clearly show important differences between the public and private sector demand for new technologies. We explore two alternative methods of testing for how the political environment communicates citizen demand. In columns 1, 3, 4, and 6 of Table 2, we include the share of the urban population in the largest city of the state. When included by itself (columns 1 and 4), the urban population share in the largest city is shown to reduce the usage of the new technologies. That urban dispersion is central is a surprising result to the extent we believe large cities are the hub of technological innovation. The concentration of interests can be detrimental in the political environment, however, because a larger urban share in the largest city suggests that the given urban population is concentrated in a smaller geographic area. The political process may dictate that the number of legislators that are interested is a factor in fostering political action. As opposed to the primacy result in Henderson (2003), the process by which urban interests are communicated politically is more important when it comes to public sector technological development.

An alternative measure of the dispersion of urban interests is the Herfindahl index of city size. A larger value of the Herfindahl is associated with an urban population concentrated in fewer cities, and thus is positively correlated with the share of the urban population in the largest city (Henderson 2003). The estimated coefficients in Table 2, like those for the share of the population in the largest city, demonstrate that a more concentrated urban population results in reduced usage of the new technologies. Thus both measures of urban concentration consistently are found to show that a more dispersed urban population yields a more pervasive use of new technologies.²⁸

Table 2 also offers insight about the potential impacts from the influence of the bureaucracy. The share of the public sector workforce that is unionized is found to have a negative influence on technology adoption, and is often statistically significant.²⁹ Since the variable does not explicitly refer to UI workers, it is interpreted in the general context of how the interests of the bureaucracy are weighted within the government structure.³⁰ The finding is consistent with the results we previously discussed, that in aggregate, new technologies may be perceived as a threat to the traditional bureaucrats that staff UI offices.

Supporting the unionization result is the variable that measures the percent of the state’s urban population in each state’s capital city. The Table 2 results illustrate that in the specifications without broadband connections, there is a negative and significant coefficient indicating that states with relatively larger capital cities adopt new technologies more slowly. This result qualitatively holds when broadband connections replace the urban share but is smaller in magnitude and not significant at conventional levels.³¹

Our political structure results generally show little impact involving political parties, but we find that Democratic governors are very supportive of bureaucratic interests. Although relatively small and statistically significant in only one specification, we find states with Democratic governors have reduced progress at adopting new technologies. The direct interest of governors, however, is shown by when a Democrat governor is in the final term due to term limits; we find a significant additional reduction in usage of the new technologies. The negative and significant effect holds for all specifications in columns 1-6 in Table 2. Besley and Case (1995) first found empirical evidence that term-limited governors are able to operate closer to their own preferences in their final term. The source of the term-limit effect here is unclear, and could be to reward unions, bureaucrats, or some other motivation. The term limit effect, therefore, seems to indicate internal factors within the government push on adopting the new technologies, and Democratic governors can only resist these forces when they face relatively less electoral scrutiny.

In addition to using the urban population to represent demand by residents, we also find that citizen demand is reflected in some of the residential characteristics. We find the share of the workforce in manufacturing is associated with lower usage of new technologies. The traditional orientation of UI programs in many states has been toward manufacturing, which have a common reputation of being subject to economic cycles. In the U.S., manufacturing has also been associated with older, traditional workers. To the extent this generalization is true, the negative coefficient on the usage of new technologies in states with greater manufacturing would also be consistent with a demand interpretation. Similarly, we find that states with a larger share of young people are associated with a greater usage of the new technologies. Finally, a surprising result is that more patents per capita are found to be associated with reduced usage of new technologies. We conjecture that the patent variable may be more closely associated with legal aspects of specific industries than with the overall demand for technology.³²

Impacts of UI Technology on Administrative Costs

In Table 5, we present the estimates of Equation (1) using administrative costs as the left-hand side variable.³³ Our hypothesis is that the various factors we have modeled to capture demand by citizens including urbanization, the political process, and bureaucratic influence might also impact administrative costs. To the extent total administrative costs are correlated with the wage bill the bureaucracy will therefore be concerned with total administrative costs, although perhaps in a different direction than taxpayers. On the other hand, costs also reflect the level of effort, and may therefore reflect elements of the overall administrative process. Thus, we conduct the analysis of administrative costs in an attempt to provide an additional channel to illustrate distinctions between public and private motivations for technological adoptions.³⁴

Table 5.

Analysis of Technology Usage on UI Administrative Costs per Capita.

		Period: 2002-2011
	Dependent variables:	ln total UI admin cost pc	ln total UI admin cost pc	ln fixed cost pc	ln variable cost pc
Independent variables		(1)	(2)	(3)	(4)
ln % UI Technology Claims		−0.05 (0.04)	−0.05 (0.03)	−0.06 (0.06)	−0.02 (0.02)
ln % Urban		−0.64 (1.29)	−0.27 (1.26)	−9.44 (8.09)	−1.12 (0.77)
ln % Urban Pop in largest City		0.29 (0.84)	0.32 (0.83)	−3.14 (3.01)	0.99 (0.62)
ln City Herfindahl		−0.39 (0.93)	−0.45 (0.91)	2.46 (3.98)	−0.46 (0.55)
ln % Urban Pop in Capital City		−0.05 (0.25)	0.07 (0.24)	2.06** (0.90)	−0.98*** (0.29)
ln % Public sector workers unionized		−0.01 (0.05)	−0.01 (0.05)	0.12 (0.22)	−0.03 (0.05)
ln Broadband Connectivity per capita		0.01 (0.03)	0.02 (0.03)	0.58 (0.38)	−0.04 (0.04)
Gov Term limit		−0.01 (0.02)	−0.01 (0.02)	−0.11 (0.13)	−0.02 (0.02)
Dem Gov		−0.02 (0.02)	−0.02 (0.02)	−0.05 (0.09)	0.01 (0.02)
(Gov Term Limit)*(Dem Gov)		−0.03 (0.02)	−0.03 (0.02)	0.11 (0.17)	0.01 (0.03)
(ln % UI Technology Claims)*(Gov Term limit)		0.05 (0.03)	0.05 (0.03)	0.02 (0.07)	0.03 (0.03)
(ln % UI Technology Claims)*(Dem Gov)		0.04 (0.03)	0.04 (0.03)	0.08* (0.05)	0.02 (0.02)
(ln % UI Technology Claims)(Gov Term Limit)(Dem Gov)		−0.07* (0.03)	−0.06** (0.03)	−0.06 (0.08)	−0.03 (0.03)
ln Patents per capita		0.01 (0.04)	0.00 (0.04)	−0.06 (0.28)	−0.02 (0.05)
ln % Manuf workers		−0.54*** (0.20)	−0.38* (0.20)	−0.70 (1.38)	−0.12 (0.19)
ln % Unemp			0.24*** (0.06)	−0.20 (0.52)	0.58*** (0.09)
ln real GSP per capita		0.33 (0.33)	0.55* (0.33)	0.86 (2.53)	0.35 (0.23)
ln % Poverty		−0.05 (0.06)	−0.10 (0.06)	−0.03 (0.39)	−0.16*** (0.06)
ln % Young		0.22 (0.40)	0.35 (0.40)	0.47 (2.18)	−0.48** (0.24)
ln % Old		−0.11 (0.18)	−0.11 (0.16)	−2.84* (1.60)	−0.23 (0.30)
ln % white		0.86** (0.42)	0.55 (0.41)	0.43 (3.07)	0.87*** (0.26)
ln % College		0.01 (0.12)	−0.00 (0.11)	−0.61 (0.52)	0.35*** (0.11)
ln % HS and Over		−0.36 (0.56)	−0.25 (0.53)	−3.57 (2.50)	−0.41 (0.55)
Observations		496	496	496	496
R-squared		0.94	0.94	0.26	0.93

Notes: Robust standard errors in parentheses. Significance levels: *: 10%; **: 5%; ***: 1%. This table describes the determination of fixed and variable costs. The coefficient on recipients is multiplied by the number of recipients to create variable costs, while the residuals from estimating the regression on the number of UI recipients, state and year effects, and a constant are used to construct fixed costs. Estimations are based on data for 50 US states and the period 2002-2011. Four observations on Hawaii are omitted for 2002-05 due to missing broadband data. State and year fixed effects are included in the regressions.

In addition to the specification used for the main results in Table 2, we add four new variables to capture the impact of the new technologies on costs. Specifically, we add the share of total UI claims that use the new technologies, and we interact the share of new technology claims with the three political influence variables from the second panel of Table 2 as new variables in Table 5; these variables indicate whether the governor is a Democrat, whether the governor is term limited, and the interaction of Democratic governor and term limits. The motivation for doing so is that the different political actors may have difference preference weights on cost savings––both to the government directly or to potential UI recipients.

To develop a variable cost measure, we run a regression of total administrative costs on the number of UI recipients plus fixed effects. We use the coefficient on recipients to calculate total variable costs, and subtract variable from total costs so that the remainder are fixed costs.³⁵ The rationale for separating costs is that creating the new technology is likely to entail significant fixed costs. The advantage of the higher fixed costs, however, is the potential for the new technologies to generate lower variable costs.

We present the estimation results in Table 5 by omitting the unemployment rate (Column 1) and including it (Column 2). The similarity of the other coefficients in the two tables suggests the addition of unemployment is orthogonal, and the coefficient represents the net of scale effects and congestion effects.

The most interesting aspects of Table 5 are illustrated when the political interaction terms are disentangled, as presented in Table 6. The marginal elasticities of Table 6 are grouped into three sections, which we term the Technology Claims Effect, the Term Limit Effect and the Democrat Governor Effect. The Technology Claims Effect shows that the usage of the new technologies in particular circumstances is associated with cost savings. Specifically, the estimated elasticity is a statistically significant −0.02 for states with a Democrat governor in the term that is term limited. The finding suggests that a 10% increase in technology claims may be associated with a 0.2% decrease in total administrative costs only when a Democrat governor is in the last term that is term limited.³⁶ Other than the effect for Democrat governors that are unable to run for office again, however, we do not find that the new technologies are associated with systematic cost savings.

Table 6.

Marginal Elasticities of Technology Claims and Political Structure on UI Administrative Costs.

	Period: 2002-2011
Dependent variables:	ln total UI admin cost pc (1)	ln total UI admin cost pc (2)	ln fixed cost pc (3)	ln variable cost pc (4)
Technology Claims effect (elasticity of change in technology claims evaluated at)
For democrat governor in term that is term limited	−0.020* (0.012)	−0.021* (0.012)	−0.021 (0.059)	−0.004 (0.015)
For democrat governor in term that is not term limited	−0.006 (0.011)	−0.004 (0.018)	0.022 (0.046)	0.000 (0.013)
For Republican governor in term that is term limited	0.001 (0.018)	−0.001 (0.017)	−0.045 (0.060)	0.009 (0.022)
For Republican governor in term that is not term limited	−0.050 (0.035)	−0.048 (0.033)	−0.063 (0.055)	−0.021 (0.025)
Term limit effect (evaluated at mean technology claims)
For democrat governor	−0.038 (0.020)	−0.032* (0.019)	0.024 (0.095)	−0.006 (0.017)
For Republican governor	−0.040** (0.019)	−0.038** (0.018)	−0.119 (0.112)	−0.037 (0.018)
Democrat governor effect (evaluated at mean technology claims)
For term that is term limited	−0.042 (0.026)	−0.036 (0.026)	0.053 (0.183)	0.032 (0.024)
For term that is not term limited	−0.045* (0.023)	−0.042* (0.022)	−0.090 (0.083)	0.001 (0.019)
Observation	496	496	496	496

Notes: Robust standard errors in parentheses. Significance levels: *: 10%; **: 5%; ***: 1%. This table presents the estimated marginal elasticities based on the interaction terms in the second panel of results in Table 5 appropriate for each column. For the interactions, all continuous variables are evaluated at their means.

The second set of results in Table 6 further suggests that the term limit effect is important, and that bureaucratic resistance to new technologies is substantive. Specifically, we see in the Term Limit Effect that for governors of either political party, when the governor is unable to run for office again that there are UI administrative cost savings. This term limit effect results in about 4% lower overall UI administrative costs for governors of either political party.³⁷

The final aspect of the political environment is shown in the Democrat Governor Effect. The coefficients illustrate the change in costs holding constant the share of UI claims with the new technologies. We find that for Democrat governors that UI administrative costs are about 4% lower than for Republican governors, irrespective of their office term.³⁸

The importance of the governor’s last term suggested in the results above is consistent with a tension between the governor’s office and the bureaucracy, since the bureaucracy is in the institution within the government that might alter actions toward new technologies. To support our interpretation, we further examine the Table 5 results in columns 3 and 4. We previously found that the relative size of the capital city resulted in significantly slower adoption of technology. Thus it might not be surprising that column 3 shows that a relatively larger capital city results in significantly higher fixed administrative costs, although column 4 shows there are also lower variable costs. Such a cost structure could be consistent with a more serious effort where bureaucratic interests are stronger. Despite the more serious design effort, however, we do not find that overall administrative costs are lower. Therefore, the pattern of the results suggests that bureaucrats may get some compensation for their reduced overall role, in the form of a more comprehensive responsibility in the design of the new technologies. An interesting aspect of the possible administrative bargain is that there may be a long run pay-off in the form of lower operating costs of the newer systems.

Robustness

To ascertain the robustness of the results, we perform a series of sensitivity analyses along several dimensions. Most of the sensitivity analyses are reported in the appendices and are only briefly discussed here. In addition to our concern about selectivity discussed earlier, we explore the empirical importance of the legislature, the functional form of the regression, and sensitivity to a couple of variables. In virtually all cases, our conclusions support our earlier findings from estimating Equation (1).

Given the importance of the political environment, we explore whether the political party of the legislature impacts technology adoption. On the one hand, the legislature interacts with the governor to pass policies, so it is natural to expect it might. On the other hand, the executive branch is where execution of policy occurs and is where politicians and the bureaucracy interact. Further, the state level is much less ideological politically, so the legislature and executive are less likely to oppose each other irrespective of party dominance. We find the political party shares in the legislature are completely insignificant in Equation (1) leaving all else unchanged. Moreover, as shown in Appendix Table A3, the primary analysis is unchanged when we use dummy variables to indicate whether the state legislature is controlled by Democrats, or is split between Democrats and Republicans.³⁹

Appendix Table A4 re-estimates the Table 5 cost results excluding the broadband variable, and Appendix Table A5 presents the marginal political contribution calculations analogous to Table 6 without any change in the results. Appendix Table A6 re-estimates the technology usage table (Table 2) excluding the share of the population with a college degree since that group is included in the variable of high school and above. Appendix Table 7 presents estimates using a spatial lag approach where the weight matrix is based on contiguous states, and finds the basic results are preserved.⁴⁰ Appendix Tables A8, A8a, and A8b are comparable to Tables 2–4; we find that changing the functional form to linear from double-log does not change the qualitative results. The final tests we present in Tables A9, A9a, and A9b omit all of the socio-demographic variables. The only substantive difference is that the relative size of the capital city is insignificant. In summary, the robustness tests indicate that our main results are preserved with only minor differences.

Summary and Conclusion

Our paper has examined the forces that impact the adoption of new technologies in the government sector. The examination is interesting because, as we demonstrate, attributes of political choice are quite different from choices made in the private sector. These attributes reflect differences in the choice process. Governments make decisions with a combination of legislatures and the members of the executive branch. These policy decisions and then implemented by a hired bureaucracy. The Table 2 results powerfully suggest that technological change comes to the public sector in a process different from the private sector––sensitive to both the distribution of high demanders and also the political influence of the bureaucracy. For example, we find it is not only citizen demand that impacts governmental choices, but how those demands are translated politically. Factors that distinguish government from private sector adoption of new technologies are found to be the distribution of interest groups (e.g. the urban population), the political orientation and incentives facing politicians, and the policy strength of the bureaucracy. We use the technology of UI applications as our empirical example. We apply panel estimation across US states over time to the choice between using traditional in-person bureaucratic offices compared to using automated telephone and internet applications.

We demonstrate the importance of the geographic distribution of the population by looking carefully at the distribution of the urban population within a state. We find that for a given urban population, a wider distribution is associated with more rapid adoption of the new administrative technologies.⁴¹ We believe the result is consistent with the premise that a wider range of interested politicians is more effective than narrow but stronger interests (Brueckner, Craig and Lee 2021).

Our empirical work finds that, with one exception, little influence on administrative technology can be attributed to particular political parties. We do find, however, that the interaction between the bureaucracy and politicians is important; we believe because the new technologies are perceived as a threat to traditional bureaucratic employment. We find public sector unionization slows technology adoption. We find further support by looking at the relative population size of the capital city. Our hypothesis is that a larger state capital city, holding constant the overall urban population, indicates larger importance of the government sector in the state economy. Our empirical finding is consistent with the interpretation that a larger influence of government workers slows the adoption of the new technologies.

The alternative path we examine is to show how the three sources of influence (residential demand including urbanization, politicians, bureaucracy) on the speed of technology adoption impact UI administrative costs. We find that costs vary with technological familiarity as illustrated by residential demographic characteristics. More important, however, is that our cost analysis reinforces some of the political forces we find in the analysis of technology adoption. In particular, we find that for governors that are ineligible to run for office again due to term limits, UI administrative costs are lower. We attribute the result to reduced bureaucratic resistance, as reinforced by a significant capital city effect.

Our results suggest that the adoption of new technologies in the government sector, which consists of at least 1/3 of the economy in most countries, is different than in the private sector. Given the additional complexities on choices within the government sector, in some senses it is surprising that the new technologies have evolved at all. Our analysis of both administrative usage and costs finds that, despite the myriad sources of potential political and bureaucratic barriers, the new technologies have been adopted. There appear to be few governmental administrative cost savings; we speculate the main savings are achieved by reduced queuing and transport time by UI recipients. Such cost savings are not achieved equally by all states. Areas with greater bureaucratic support and more concentrated residential support have received reduced cost savings, and have received the stream of cost savings later, than states where the bureaucracy is not as influential and where political support is more broadly based.

Supplemental Material

Supplemental Material - Adoption of Technological Change in the Public Sector: Evidence From US States

Supplemental Material for Adoption of Technological Change in the Public Sector: Evidence From US States by Steven G. Craig, Edward C. Hoang and Janet E. Kohlhase in International Regional Science Review

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Edward C. Hoang

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Alesina

Alberto

Tabellini

Guido

. 2007. “Bureaucrats or Politicians? Part I: A Single Policy Task.” American Economic Review 97: 169-179.

Audirac

Ivonne

. 2005. “Information Technology and Urban Form: Challenges to Smart Growth.” International Regional Science Review 28: 119-145.

Baumol

William J

. 1967. “Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis.” American Economic Review 57: 415-426.

Banks

Jeffrey

Weingast

Barry

. 1992. “The Political Control of Bureaucracies Under Asymmetric Information.” American Journal of Political Science 36: 509-524.

Bertinelli

Lusito

Black

Duncan

. 2004.“Urbanization and Growth.” Journal of Urban Economics 6: 80-96.

Besley

Timothy

Case

Anne

. 1995. “Does Electoral Accountability Effect Economic Policy Choices: Evidence from Gubernatorial Term Limits.” The Quarterly Journal of Economics 110: 769–798.

Brueckner

Jan K.

Craig

Steven G.

Lee

Kangoh

. 2021. “Regionalism Meets Samuelson: Local Production of a National Public Good,” Economica 88: 1-31.

Capello

Roberta

Caragliu

Andrea

Fratesi

Ugo

. 2017. “Modelling Regional Growth Between Competitiveness Austerity Measures: The MASST3 Model.” International Regional Science Review 40: 38-74.

Craig

Steven G.

Hoang

Edward C.

Kohlhase

Janet E

. 2021. Can Bureaucrats Manipulate Policy Outcomes. University of Houston working paper.

10.

Craig

Steven G.

Palumbo

Michael

. 1999. “Policy Variation in State Government Provision of Unemployment Insurance.” Journal of Regional Science 39: 245-274.

11.

Ebenstein

Avraham

Kevin

Stange

. 2010. “Does Inconvenience Explain Low Take-up? Evidence From Unemployment Insurance.” Journal of Policy Analysis and Management 29: 111-136.

12.

Epstein

David

O’Halloran

Sharyn

. 1994. “Administrative Procedures, Information, and Agency Discretion.” American Journal of Political Science 38: 697-722.

13.

Faggio

Giulia

Overman

Henry

. 2014. "The Effect of Public Sector Employment on Local Labour Markets." Journal of Urban Economics 79: 91-107.

14.

Hansen

Niles

. 1993. “The strategic role of producer services in regional development.” International Regional Science Review 16: 187-195.

15.

Heckman

James J

. 1979. “Sample Selection Bias as a Specification Error.” Econometrica 47: 153-161.

16.

Henderson

J. Vernon

. 2003. “The Urbanization Process and Economic Growth: The So-What Question.” Journal of Economic Growth 8: 47-71.

17.

Henderson

J. Vernon

. 2010. “Cities and Development.” Journal of Regional Science 50: 514-540.

18.

Hirsch

Barry T.

Macpherson

David A.

Vroman

Wayne G

. 2001. “Estimates of Union Density by State.” Monthly Labor Review 124: 51-55.

19.

Mansfield

Edwin.

1961. “Technical change and the rate of imitation.” Econometrica, 29: 741-766.

20.

Niskanen

William. A

. 1971. Bureaucracy and Representative Government. Chicago, IL: Routledge.

21.

Salmeron

Roman

Garcia

C. B.

Garcia

. 2018. “Variance Inflation Cactor and Condition Number in Multiple Linear Regression.” Journal of Statistical Computation and Simulation 88: 2365-2384.

22.

Smart

Michael

Sturm

Daniel

. 2013. “Term Limits and Electoral Accountability.” Journal of Public Economics 107: 93-102.

23.

Tolbert

Caroline J.

Mossberger

Karen

. 2006. “The Effects of E‐government on Trust and Confidence in Government.” Public Administration Review 66: 354-369.

24.

Tummers

Lars

Rocco

Philip

. 2015. “Serving clients when the server crashes: how frontline workers cope with E‐government challenges.” Public Administration Review 75: 817-827.

25.

Ujhelyi

Gergely

. 2014. “Civil Service Rules and Policy Choices: Evidence from US State Governments.” American Economic Journal: Economic Policy 6: 338-380.

26.

Uppal

Yogesh

Glazer

Amihai

. 2015. “Legislative Turnover, Fiscal Policy, and Economic Growth: Evidence from U.S. State Legislatures.” Economic Inquiry 53: 91–107.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.58 MB