Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Associate Editor: Patrick Chau Acknowledging the global relevance of controlling corruption, our study contributes to the existing body of
literature by exploring the relationship between E-Government (EGOV) and Corruption Control at the EGOV
Keywords: subindices level. Toward this, we draw upon theoretical insights from Resource-Based View (RBV), Media Lit
E-Government eracy Theory, Agency Theory, Information Quality, and Transaction Cost Theory to derive a set of testable
Corruption Control
hypotheses. Using a panel dataset comprising 102 countries, we employ a Linear Mixed Effects (LME) estimation
Linear Mixed Effects (LME)
method to thoroughly investigate and uncover the relative effectiveness of EGOV subindices in controlling
Human Capital (HC)
Telecommunications Infrastructure corruption. Accordingly, we draw implications for theory, praxis and future research.
Online Services (OS)
* Corresponding author at: Office 308, IIM Jammu, Jammu, Jammu and Kashmir, India 180016.
E-mail addresses: prakrit@iimj.ac.in (P. Silal), ashutosh.jha@spjimr.org (A. Jha), ds@iimcal.ac.in (D. Saha).
https://doi.org/10.1016/j.im.2022.103735
Received 23 July 2020; Received in revised form 24 November 2022; Accepted 28 November 2022
Available online 30 November 2022
0378-7206/© 2022 Elsevier B.V. All rights reserved.
P. Silal et al. Information & Management 60 (2023) 103735
2
P. Silal et al. Information & Management 60 (2023) 103735
(governments, public service agents, and citizens) involved in the public Table 1
service delivery process [5,30–32]. The more the information asym Summary of Literature Review.
metry, the higher the discretionary power enjoyed by that stakeholder Ref. Authors, Year Nature of Analysis Major Findings
with information superiority [33]. Such power inequality arises pri Work Techniques / and/or Limitations
marily due to the inability of other stakeholders with relatively less in Level of Analysis
/ Data used
formation to hold the superior stakeholder accountable for the latter’s
actions [33]. Addressing the Corruption Control issue from this angle, [4] Mistry, 2012 Conceptual Case Study E-Government
multiple studies have already argued for the efficacy of EGOV in / Empirical Method, controls Corruption
Country-level by reducing
reducing the government-agent-citizen information asymmetry [14,15], analysis, single discretionary
thereby reducing the discretionary powers of public agents [4] and case of India powers through IT-
increasing the ability of citizens to hold their government and/or used enabled
representative agents accountable [15]. These scholarly works on transparency and
information access.
lowering agent discretionary powers [4,5] can be broadly classified
[17] Kim, 2014 Empirical Multiple E-Government has a
under two categories (Table 1): (i) reducing the information asymmetry Regression significant positive
[4,17,32] and (ii) infusing greater transparency in the public service Analysis, Cross- effect on Corruption
delivery process [5,8,15,16,32]. country level, Control.
Notably, as illustrated in Table 1, most of the empirical studies that 154 countries Relationship tested
(2005 and 2007 only for cross-
quantitatively validate the effectiveness of EGOV as an anticorruption data) section data.
tool fall short of explaining the country heterogeneities mentioned in the [32] Srivastava Empirical SUR Panel Data E-Government
Introduction section mainly due to their methodological constraints. et al., 2016 Analysis, Cross- controls permeated
This limitation is especially evident in the studies that adopt a cross- country level, 63 corruption via its
countries (2004 effect on corruption
sectional analytical approach, as discussed earlier. The failure to
to 2007) in legal and media
tackle the same infuses biases in the estimated effects of EGOV on institutions. The
Corruption Control, thereby casting serious doubt on the validity of their study does not
findings. Naturally, accounting for such heterogeneities is imperative for account for country
obtaining consistent and unbiased estimates. Moreover, barring a few heterogeneity in
terms of initial
exceptions [5,6,30], there is an evident lack of studies to assess the
corruption levels or
EGOV-COR relationship’s stability over a long period of time. But this type corruption variation
of study seems extremely useful, where EGOV is found to affect Cor across time.
ruption Control significantly. Though Elbahnasawy (2014) considered [16] Shim and Eom, Empirical Cross-sectional E-Government was
2008 Analysis, Cross- significantly and
for the longitudinal stability of the EGOV-COR relationship using panel
country level, negatively
regression analysis with data till 2009 [5], his findings have not been 127 countries associated with
tested further with post-2009 data. In summary, there exists only a (2003 -E- Corruption.
handful of studies that account for both country heterogeneities as well as Government Relationship tested
longitudinal stability while exploring the EGOV-COR relationship data; 2004 only for cross-
Corruption sectional data.
internationally.
Perception Index
Another distinct observation from the existing stream of EGOV-COR Data)
literature is the presence of nonuniformity across countries and regions in [8] Krishnan et al., Empirical SEM Cross- E-Government has a
the nature and strength of the EGOV-COR relationship. For instance, 2013 sectional significant negative
analysis effect on Corruption
Andersen (2009) found that EGOV-COR relationships may be slightly
(averaged (mediator).
weaker in OECD countries than in non-OECD countries [14]. Nam dataset), Cross- Relationship tested
(2018) too corroborated the presence of nonuniformity in the country level, only for cross-
EGOV-COR relationship with his analysis involving data pertaining to 105 countries sectional data.
179 countries in 2016 [15]. On the other hand, Bac (2001) observed that (2004 to 2008
Data)
the effect of transparency (whether EGOV-induced or not) in lowering
[5] Elbahnasawy, Empirical Panel Data E-Government has a
corruption might cease to exist or reverse the trend after an upper level 2014 Analysis, Cross- significant negative
of transparency is breached, due to “connections” effect overpowering country level, effect on
the “detection” effect [34]. In the context of developing countries, 160 countries Corruption. Similar
(1995–2009 analysis has not
contrary to popular belief, Schuppan (2009) argued that how EGOV
Data) been done for post-
could lead to more corruption due to the potential abuse of ICT solutions 2009 data. Also, the
by bureaucratic elites for their vested interests [35]. Although these study does not
findings clearly point toward the presence of nonuniformity in the account for country
EGOV-COR relationship across countries, few studies have taken it into heterogeneities in
terms of individual
consideration while analyzing the impact of EGOV on Corruption Con
corruption levels
trol. As mentioned earlier, most EGOV-COR studies conducted at the and corruption
composite index level [14] mask the individual effects of the subindices, variation across
viz. HCI, TII, and OSI, and hence fail to account for variations in their time.
individual effects on Corruption Control. While our study does not fully [15] Nam, 2018 Empirical Cross-sectional E-Government has a
Analysis, Cross- significant positive
investigate the underlying reasons for nonuniformity, we apprehend country level, effect on Corruption
that the possibility of the wide variation in the individual effects of HCI, 179 countries Control.
TII, and OSI on Corruption Control might be a cause for the observed (2016 Data) Relationship tested
nonuniformities in the EGOV-COR relationship [5]. It is worth only for cross-
sectional data.
mentioning here that Elbahnasawy (2014), using global data till 2009,
[14] Andersen, Empirical Panel Data E-Government has a
already demonstrated visible differences in the individual effects of the 2009 Analysis, Cross- significant positive
three subindices on Corruption Control – TII and OSI were found to be (continued on next page)
significant, while HCI did not have any significant effect on Corruption
3
P. Silal et al. Information & Management 60 (2023) 103735
4
P. Silal et al. Information & Management 60 (2023) 103735
government-agent-citizen information asymmetries and subsequently argue that citizens’ access to quality information increases substantially,
reducing the agent discretionary powers [5,32]. Accordingly, the EGOV which results in lower information asymmetries, enhanced trans
development of a country provides the government with core capabil parency, and subsequent reduction in corruption levels. We accordingly
ities that enable them to have greater control and monitoring capacity hypothesize that as follows:
over their agents [32]. EGOV may also provide unhindered public access H3: Increase in a country’s Telecommunications Infrastructure is posi
to information, which enhances the transparency of government func tively associated with an increase in its Corruption Control over time.
tioning and substantially empowers citizens to hold the government and
their agents accountable for their actions [5]. EGOV, thus, discourages 2.2.4. Relating online services and corruption control
corruption enablers and positively contributes to Corruption Control by Online Services (OS) capture the digitalization of the public service
enabling direct government-citizen connection and facilitating greater delivery infrastructure of a country and hence represent an ICT-enabled
control over agents. We, therefore, hypothesize that EGOV has a positive capability [15,42]. OS, therefore, seeks to streamline the public service
association with Corruption Control over time, as stated below. delivery process, which presently entails high transaction costs for cit
H1: Increase in a country’s E-Government development levels is positively izens while availing these services. Toward this, we draw from Trans
associated with an increase in its Corruption Control over time. action Cost Theory to argue that OS reduces the cost of availing public
services by lowering uncertainty and reducing asset specificity [63]. Un
2.2.2. Relating human capital and corruption control certainty, in this context, refers to the unreliability in quality, timeline,
Human Capital (HC), as conceptualized in the RBV stream of liter and experience of service fulfillment [64]. Online mechanisms help
ature, refers to the expertise and knowledge of employees in a firm [42]. citizens track the service delivery process end-to-end, thereby ensuring
These capabilities of employees represent a source of strategic advan transparency and reducing uncertainty [62]. Accordingly, a higher level
tage for a firm as another firm cannot easily imitate them. At the country of OS infrastructure in a country can enable a transparent, disin
level, HC is captured by the literacy levels and skill sets of citizens [52]. termediated, and accountable public service delivery system, which
Increasing mass education and literacy improves cognitive capacities, clamps down on any resource leakages and wipes out opportunities for
such as conceptualization ability [53] of citizens that help them analyze extorting bribes and other favors [15,32]. Asset specificity, in our context,
information better and improve its effectiveness in reducing information refers to the degree to which a public service agent (aka service pro
asymmetry and enhancing transparency. Accordingly, we draw from the vider) is responsible for fulfilling the public service needs of citizens (aka
Theory of Media Literacy to explicate the link between HC and Cor service consumers) [63]. The higher the agent’s degree of involvement,
ruption Control. The Theory of Media Literacy, which is conceptually the higher the asset specificity. Greater asset specificity implies a greater
rooted in the communications literature, concerns the ability of the dependence of citizens on the whims and fancies of the agent, which
citizen to “access, analyze, and communicate messages in a variety of may enhance their discretionary power resulting in a higher prevalence
forms” [54]. Media Literacy assumes high contemporary relevance of petty bribes and corrupt practices. On this note, government process
given the crucial need for citizens to make sense of the continuous influx re-engineering, enabled by the digitalization drive, restructures and au
of information enabled by EGOV [55]. Accordingly, improved cognitive tomates the public service delivery process (i.e., less human agent
capacities, due to HC improvements, equip citizens with the requisite intervention), which, in turn, may reduce the asset-specificity of services
Media Literacy to effectively process information present across media and minimize the scope for corrupt practices among agents, bureau
channels [56] while also enabling them to understand the underlying cratic elites, and politicians alike [32,63]. Accordingly, we employ the
ideologies of media messages [57]. As the effectiveness of the infor Transaction Cost Theory to argue that as countries attain higher levels of
mation provided is only as good as the person using it, with improved OS, they achieve higher effectiveness in controlling corruption through
HC levels of a country, citizens could make better sense of publicly reduction of service uncertainty and asset specificity. Therefore, we
available information and hold governments accountable for their ac hypothesize as follows:
tions [56]. Drawing from our earlier arguments, the push from citizens H4: Increase in a country’s Online Services is positively associated with
for greater government accountability could better help in curbing an increase in its Corruption Control over time.
corruption. Accordingly, we hypothesize that as follows:
H2: Increase in a country’s Human Capital is positively associated with 3. Research methodology
an increase in its Corruption Control over time.
Our research model, derived from Fig. 1, outlines the hypothesized
2.2.3. Relating telecommunications infrastructure and corruption control relationships as shown in Fig. 2. Our independent variables comprise
Telecommunications Infrastructure (TI) of a country plays an inte EGOV, HC, TI, and OS, while Corruption Control represents our outcome
gral role in the EGOV development of the country [5]. Borrowing from variable. Further, we have used a set of five control variables (to be
RBV, TI represents the capability of countries to deploy ICT infrastruc explained later) to account for other known determinants of Corruption
ture to ensure the widespread availability and seamless use of ICT devices Control. While we also checked for the possible moderating effects of
and allied services, including mobile, broadband, and the Internet [42]. these control variables (shown in Appendix 2), we opted for the most
TI, therefore, empowers citizens to enjoy unhindered and timely access parsimonious model without any moderating variables (Random In
to public information and seeks to reduce the information asymmetry tercepts and Slope model presented later in Table 13) having a condi
existing among government, public agents, and citizens [58]. On this tional R-squared value equal to 0.9848. In the following subsections, we
note, the extent to which TI can reduce the information asymmetry is first elaborate on data and measures, and then we discuss the research
contingent on the quality of information it is able to disseminate. Prior method employed.
literature defines information quality as the output quality of informa
tion systems, assessed in terms of their business relevance and useful 3.1. Data and measures
ness, ease of understanding, and ability to satisfy user requirements
[59]. Toward this, the Information Quality literature recognizes accuracy We use archival secondary data sourced from reputed institutions to
(concurrence with reality), consistency (absence of conflict between operationalize our set of constructs.
datasets), completeness (presence of all relevant data), and currency
(up-to-date) as the four dimensions of information quality [59]. 3.1.1. Outcome variable
Accordingly, the extant literature highlights the positive association of The outcome variable Corruption Control is operationalized using
information quality with reduced information asymmetry [60,61] and the measure “Corruption Perceptions Index” (CPI), published annually
enhanced transparency [62]. As the TI level in a country increase, we by Transparency International [65]. CPI is a well-known Corruption
5
P. Silal et al. Information & Management 60 (2023) 103735
Control measure and has been widely used across various disciplines footsteps of most researchers, we too make use of five parameters,
[5]. CPI captures citizens’ perceptions of the prevalence of corruption in namely Government Effectiveness (GE), Rule of Law (RL), Economic
national institutions spanning judiciary, legislature, police, and other Prosperity (EP), Anti-Press Freedom (APF), and Trade Openness (TO) as
law-enforcing agencies, military, public service, and media, among control variables. They are widely theorized in the existing literature as
others. In doing so, it provides a near proxy measure for capturing both predictors of corruption [5,30,33,68–72]. GE is collected from the World
bureaucratic and political (or grand) corruption. CPI values, obtained Governance Indicators [73] of the World Bank. In this, each country is
through expert assessments and opinion surveys, range from 0 to 100, scored as per the Kaufmann Index scale [74], which ranges from − 2.5 to
with higher values corresponding to lower corruption, i.e., higher Cor +2.5, where “− 2.5′′ signifies the lowest level of GE while “+2.5′′ sig
ruption Control in our context. In other words, the level of Corruption nifies the highest level of GE. GE captures the quality of policymaking
Control is directly calibrated to the CPI values. and public service delivery and the “credibility of government’s
commitment to such policies” [17]. Higher levels of GE can contribute
3.1.2. Independent variables toward an effective, efficient, and transparent government [75], thereby
The independent variables are operationalized using data from the discouraging any corrupt practices [76]. In line with prior studies [17,
United Nations E-Government Survey [24]. EGOV is operationalized 75,76], the higher the GE, the lower the corruption. RL, which concerns
using the composite index E-Government Development Index (EGDI). law and order, is also collected from the same World Governance In
HC is operationalized using the Human Capital Index (HCI), which is a dicators [73]. This variable “reflects perceptions of the extent to which
country-level measure comprising: (i) adult literacy percentage, (ii) agents have confidence in and abide by the rules of society, and particularly
gross enrolment ratio, (iii) expected years of schooling, and (iv) mean the quality of contract enforcement, property rights, the police, and the
years of schooling [23]. HCI ranges from 0 to 1, with a higher value courts, as well as the likelihood of crime and violence” [73]. The RL score is
corresponding to higher HC. TI is operationalized using the Telecom also based on the Kauffman Index scale, where “− 2.5′′ signifies weak law
munications Infrastructure Index (TII), which is assessed using five in and order, while “+2.5′′ signifies good law and order. Good law and
dicators: (i) individuals using the Internet (% population), (ii) mobile order, i.e., a high RL score, is expected to lower corruption, as demon
cellular subscriptions (per 100 people), (iii) fixed telephone sub strated in many past studies [69,72]. EP is captured in the form of GDP
scriptions (per 100 people), (iv) wireless broadband subscriptions (per per capita obtained from the World Development Indicators of the
100 people), and (v) fixed (wired) broadband subscriptions (per 100 World Bank [77]. EP can be considered as a proxy for national average
people) [5]. TII too ranges from 0 to 1, where a higher value represents living standards, where a higher national average prosperity is expected
better TI in the country. Finally, OS is operationalized using the OSI, to result in reduced competition for public resources resulting in lower
which assesses the deployment of ICT tools by governments for corruption [68]. EP has been shown to be a prominent determinant of
rendering public services [15]. The United Nations employs a rigorous corruption in the extant literature [33,71]. APF captures the extent to
process to ensure the reliability and validity of OSI measure [24]. The which the national media institutions are free from political influence
index is based on the evaluation of national websites, including national [78]. Greater media transparency or press freedom is expected to foster
service portals, e-service portals, and e-participation portals. It involves greater openness and transparency [70,72]. However, it may or may not
a primary research team who are adequately trained in all the UN offi lead to higher CPI because it is possible that citizens could be over
cial languages to ensure consistency in their assessments and also to whelmed by too many reports on corrupt practices and hence may
evaluate the websites with the mindset of a common person. The pri develop a negative perception regarding corruption in the country. The
mary research team, with the help of translators, calculates the scores data on press freedom published annually by Freedom House [78] is
using a uniform predefined guideline which encompass: (i) identifying used to measure APF, which scores countries on a scale of 0 to 100,
and locating the relevant national and ministerial websites, based on the where “0′′ represents the highest level of press freedom while “100′′
information supplied by respective member states and/or through represents the lowest level of press freedom. TO captures the extent to
search engines and (ii) closely evaluating each of these websites based which a country practices an open trade policy [72]. Higher levels of TO
on the Web Content Accessibility Guidelines of the World Wide Web are expected to lower corruption as the country’s economy would then
Consortium. The second quality assurance team follows who, besides be subject to market forces, and more transparent market-based pricing
ensuring objectivity and accuracy of the derived measures, carries out would limit opportunities for price manipulation. TO has been oper
final adjustments in the OSI scores [66]. OSI values range from 0 to 1, ationalized by two measures, namely Imports of Goods and Services, and
with a higher value representing better OS facilities in a country. Exports of Goods and Services [67], both of which are obtained from the
World Bank World Development Indicators [77]. TO is calculated as the
3.1.3. Control variables following: TO = Imports of Goods and Services (as% of GDP) + Exports
Regarding control variables, unfortunately, there is no universally of Goods and Services (as% of GDP).
agreed upon set as the determinant of corruption [67]. So, following the
6
P. Silal et al. Information & Management 60 (2023) 103735
Table 2
Descriptive Statistics of Outcome Variable and Main Explanatory Variables.
EGDI HCI TII OSI CPI
Year Mean SD Mean SD Mean SD Mean SD Year Mean SD
2008 0.542 0.174 0.843 0.151 0.294 0.237 0.490 0.192 2009 0.471 0.221
2010 0.519 0.166 0.855 0.142 0.310 0.220 0.395 0.196 2011 0.468 0.223
2012 0.589 0.186 0.782 0.166 0.414 0.245 0.572 0.202 2013 0.487 0.198
2014 0.580 0.197 0.728 0.167 0.468 0.244 0.543 0.234 2015 0.500 0.201
2016 0.604 0.186 0.719 0.167 0.468 0.220 0.624 0.220 2017 0.493 0.193
2018 0.658 0.175 0.737 0.158 0.512 0.211 0.726 0.203 2019 0.494 0.191
7
P. Silal et al. Information & Management 60 (2023) 103735
( ) ( )
U1i ∼ N 0, τ 2
and εit ∼ N 0, σ 2
(1) and GE.
As mentioned earlier, the coefficients were estimated using the Corruption Controlit |EGOV, time, U1i = β1 + β2 EGOVi(t− 1) + β3 timet + U1i
REML method, which appropriately accounts for the loss of degrees of + εit
freedom, while fitting the fixed effects to provide unbiased estimates of
the random effects. The REML estimates were calculated under a normal
where
distribution assumption for random intercepts and random slopes. This
( ) ( )
assumption is driven “more by mathematical convenience than by U1i ∼ N 0, τ2 and εit ∼ N 0, σ 2 (2)
empirical reality” [88]. To check the robustness of our findings to other
alternate distribution specifications, we relied on prior guidance [19,88] Corruption Controlit |HC, TI, OS, time, U1i
and repeated the analysis for all our models using the full maximum
= β1 + β2 HCi(t− + β3 TIi(t− + β4 OSi(t− + β5 timet + U1i + εit
likelihood estimation (MLE), which relaxes this normal distribution
1) 1) 1)
‘***’p<0.001; ‘**’p<0.01; ‘*’p<0.05; ‘⸸’p<0.1. On running this model using REML, we again obtain the same
8
P. Silal et al. Information & Management 60 (2023) 103735
Table 5
Random Intercepts Model with covariates for EGOV-CPI.
Eq. (2) No. of obs.: 612 Group Variable: Country No. of groups: 102
Log-likelihood: 936.1174 Akaike Information Criterion (AIC): − 1885.977
Coef. Std. Err. t-stat Sig. 95% Confidence Interval
Table 6
Random Intercepts Model for HC/TI/OS-CPI.
Eq. (3) No. of obs.: 612 Group Variable: Country No. of groups: 102
Log-restricted likelihood: 956.5217 Akaike Information Criterion (AIC): − 1935.454
Coef. Std. Err. t-stat Sig. 95% Confidence Interval
Table 7
Random Intercepts Model for HC/TI/OS-CPI with control variables.
Eq. (4) No. of obs.: 612 Group Variable: Country No. of groups: 102
Log-likelihood: 1080.725 Akaike Information Criterion (AIC): − 2212.476
Coef. Std. Err. t-stat Sig. 95% Confidence Interval
average slope of 0.0057 as obtained in the Unconditional Random In 3.2.4. Random intercepts and slope model with time-varying covariates
tercepts Model. Most countries enter the observation time period at the We now build on Eq. (5) and include our explanatory variables and
Corruption Control level of 0.465±0.023. Further, the negative corre control variables to obtain Eqs. (6), (7), and (8). Specifically, Eq. (6)
lation (− 0.67) observed between intercepts and time suggests that explores the effect of EGOV on Corruption Control. Subsequently, Eq. (7)
countries with higher Corruption Control levels have a slower temporal seeks to unwrap the EGOV “black box” as we assess the effect of the
growth rate, while countries that start at a lower Corruption Control individual EGOV subindices on Corruption Control. Finally, Eq. (8) as
level move up faster as they have more opportunity and scope for sesses the influence of HC, TI, and OS on Corruption Control, while
growth. The relevant results are summarized in Table 8. controlling for APF, RL, EP, TO, and GE. These augmented sets of models
Our final model specification augments the Unconditional Random are presented below. First, we present Eq. (6), subsequently followed by
Intercepts and Slope model by including the theorized predictor vari Eqs. (7) and (8).
ables while retaining the country-specific intercepts and country-
Corruption Controlit |EGOV, time, U1i , U2i
specific slopes. Through this model, we examine the EGOV-COR rela
tionship at the composite and subindices level after allowing countries = β1 + β2 EGOVi(t− 1) + β3 timet + U1i + U2i timet + εit
to vary in their initial Corruption Control levels and its growth over
time. where
9
P. Silal et al. Information & Management 60 (2023) 103735
Table 8 Table 9
Summary of Unconditional Random Intercepts and Slope Model. Random Intercepts and Slope Model with covariates for EGOV-CPI.
Eq. (5) No. of obs.: 612 Group Variable: No. of groups: Eq. (6) No. of obs.: 612 Group Variable: No. of groups:
Country 102 Country 102
Log-likelihood: 1026.312 Akaike Information Criterion (AIC): Log-restricted likelihood: 1027.243 Akaike Information Criterion (AIC):
− 2058.249 − 2063.702
Coef. Std. t-stat Sig. 95% Confidence Coef. Std. t-stat Sig. 95% Confidence
Err. Interval Err. Interval
Time 0.0057 0.0014 4.179 *** 0.0030 0.0084 EGOV 0.0755 0.0263 2.87 ** 0.0214 0.1319
intercept 0.4654 0.0228 20.398 *** 0.4205 0.5103 Time 0.0039 0.0015 2.641 ** 0.0009 0.0069
Random- Estimate 95% Confidence intercept 0.4277 0.0254 16.824 *** 0.3771 0.4788
effect Interval Random- Estimate 95% Confidence
Parameters effect Interval
nation: Parameters
unstructured nation:
Sd (time) 0.0126 0.011 0.015 unstructured
Sd (intercept) 0.2294 0.200 0.264 Sd (time) 0.01235 0.010 0.015
Corr (time, − 0.67 − 0.777 − 0.532 Sd (intercept) 0.21876 0.190 0.253
intercept) Corr (time, − 0.69 − 0.789 − 0.550
Sd (residual) 0.0238 0.022 0.025 intercept)
Sd (residual) 0.02401 0.022 0.026
‘***’p<0.001; ‘**’p<0.01; ‘*’p<0.05; ‘⸸’p<0.1.
‘***’p<0.001; ‘**’p<0.01; ‘*’p<0.05; ‘⸸’p<0.1.
( ) (( ) ( ))
U1i 0 τ11 τ12 ( )
∼ MVN , and εit ∼ N 0, σ 2
(6) analyzed above. Our analysis reveals interesting observations, as
U2i 0 τ21 τ22
captured in Table 12 and Fig. 3. TI is still found to have a consistent
Subsequently, Eq. (7) investigates the effects of HC, TI, and OS on significant and positive effect on Corruption Control, with the effect size
Corruption Control. peaking at a one-year lag. Meanwhile, HC takes a two-year lag to have a
Corruption Controlit |HC, TI, OS, time, U1i , U2i significant positive effect on Corruption Control. In particular, HC’s effect
on Corruption Control peaked at the two-year lag point, after which its
= β1 + β2 HCi(t− 1) + β3 TIi(t− 1) + β4 OSi(t− 1) + β5 timet + U1i + U2i timet + εit effect, while remaining positive, lost significance at the four-year lag
point. Further, OS more or less fails to have any significant effect on
where Corruption Control except at the two-year lag point, where the effect
( ) (( ) ( )) was negative. Also, given the observed significance in the effects of some
U1i 0 τ11 τ12 ( )
∼ MVN , and εit ∼ N 0, σ 2 (7) of the control variables (namely, RL, EP, and GE) on Corruption Control,
U2i 0 τ21 τ22
we ran some additional models to assess the possible moderating in
Finally, Eq. (8) considers the control variables, namely APF, RL, EP, fluence of the control variables (Appendix 2). As depicted in Appendix 2,
TO, and GE. we found a significant moderating influence of RL, EP, and GE in the
EGOV-COR relationship at the subindices level, which provides oppor
Corruption Controlit |HC, TI, OS, APF, RL, EP, TO, GE, time, U1i , U2i
tunities for future works.
= β1 + β2 HCi(t− 1) + β3 TIi(t− 1) + β4 OSi(t− 1) + β5 APFi(t− 1) + β6 RLi(t− 1)
+β7 EPi(t− 1) + β8 TOi(t− 1) + β9 GEi(t− 1) + β10 timet + U1i + U2i timet + εit
4. Results
where
( ) (( ) ( )) The gist of the results is shown in Table 13. Foremost, the negative
( )
U1i
∼ MVN
0
,
τ11 τ12
and εit ∼ N 0, σ 2
(8) correlation between intercepts and time slopes indicates that the rate of
U2i 0 τ21 τ22 increase in Corruption Control is low for countries that may have
We note that, on including the time-varying covariates, the average already attained their near saturation Corruption Control levels. On the
slope for Corruption Control flattens (i.e., the value reduces) in all the contrary, countries at lower Corruption Control levels can improve upon
three models pertaining to Eqs. (6), 7, and 8. In this case, on average, their status at a quicker rate till they reach close to their saturation levels
countries enter the observation period at Corruption Control levels of when the rate tapers off. In this case, including nonlinearity did not seem
0.428±0.025, 0.479±0.033, and 0.453±0.027 for Eqs. (6), 7, and 8, compelling to us due to the following reason. The residual versus fitted
respectively. Further, from Eq. (6), the rate of increase in the Corruption plot (Appendix 3) of the Random Intercepts and Slope Model (Table 13)
Control level is 0.076 units for every unit increase in the EGOV level. TI did not show any specific patterns, and the error terms were found to be
is consistently found to have a significant positive effect on Corruption randomly distributed, potentially ruling out the presence of nonlinearity
Control with an effect size of 0.125 and 0.096 for Eqs. (7) and 8, in the covariates [92]. In addition to the visual assessment, we also
respectively. HC was again found to have a negative effect on Corruption conducted a ‘sign test’ on the conditional residual values obtained and
Control with effect size obtained as − 0.053 (not significant at 95% found the residuals randomly distributed, indicating linearity [93].
significance level) and − 0.084 for Eqs. (7) and 8, respectively. OS was Further, we have tested for the influence of near saturation Corruption
not found to have any significant effect on Corruption Control. The Control by including an additional model configuration (Saturation
relevant results are summarized in Tables 9, 10, and 11. Dummy Model) in Table 13. The Saturation Dummy Model includes a
saturation dummy variable (similar to [94]), which takes the value of 1
for a country-year if the country’s growth in Corruption Control is less
3.3. POST-HOC analysis than 5% in that year as well as for all subsequent years hence. Accord
ingly, the saturation dummy variable captures whether a country has
To better understand HC’s negative effect on Corruption Control and attained near saturation Corruption Control level in the previous year.
the insignificance of OS on Corruption Control, we have further However, the effect of the saturation dummy was not found to be sig
analyzed the Random Intercepts and Slope Model of Eq. (8) with nificant. As shown in Table 13, the inclusion of time-varying covariates
increased time-lags. In particular, we have run the model for time-lags of has improved the model fit significantly, which is also visible from the
two, three, and four years, in addition to the one-year lag already improvements in parameter estimates, Akaike Information Criterion
10
P. Silal et al. Information & Management 60 (2023) 103735
Table 10
Random Intercepts and Slope Model for HC/TI/OS-CPI.
Eq. (7) No. of obs.: 612 Group Variable: Country No. of groups: 102
Log-likelihood: 1038.395 Akaike Information Criterion (AIC): − 2095.158
Coef. Std. Err. t-stat Sig. 95% Confidence Interval
Table 11
Random Intercepts and Slope Model for HC/TI/OS-CPI with control variables.
Eq. (8) No. of obs.: 612 Group Variable: Country No. of groups: 102
Log-likelihood: 1146.1 Akaike Information Criterion (AIC): − 2339.304
Coef. Std. Err. t-stat Sig. 95% Confidence Interval
Table 12
Lag Analysis for Random Intercepts and Slope Model for HC/TI/OS-CPI.
Eq. (8) 1-year lag 2-year lag 3-year lag 4-year lag
Coef. Sig. Coef. Sig. Coef. Sig. Coef. Sig.
(AIC), and log-likelihood value. Based on these parameter values, we can Intercepts and Slope Models along with the results from the post-hoc
conclude that the Random Intercepts and Slope Models have a better fit analysis to test our hypothesized relationships.
over the Random Intercepts Models. Also, the Random Intercepts and The hypothesis test results (Table 14) are mostly in line with our
Slope Models seem to be more efficient, as they have smaller residuals arguments. EGOV exhibits a significant positive effect on Corruption
and narrower confidence intervals. Therefore, we use the Random Control over time, even after allowing for heterogeneity among
11
P. Silal et al. Information & Management 60 (2023) 103735
Table 13
Model Comparison for HC/TI/OS-CPI.
Model Comparison Unconditional Random Random Unconditional Random Intercepts and Random Intercepts and Saturation Dummy
Intercepts Intercepts Slope Slope Model
Coef. Sig. Coef. Sig. Coef. Sig. Coef. Sig. Coef. Sig.
12
P. Silal et al. Information & Management 60 (2023) 103735
H4 is not supported, which is in stark disagreement with the existing compared to Uruguay. TII seems to be the major differentiating factor in
stream of literature that strongly supports the role of service digitali both cases, thereby making a case for the importance of understanding
zation in controlling corruption [5,8,15]. However, those studies have the EGOV-COR relationship at the subindices level. While our study
analyzed only the isolated effect of OS (connoted as “E-Government attempts to provide preliminary explanations for these anomalous ob
maturity”) on Corruption Control [8,15]. Thus, in the presence of HC servations by highlighting the dissimilarities in the individual effects of
and TI, the effect of OS on Corruption Control loses significance or be EGOV subindices on Corruption Control, yet there is ample scope for
comes negative. We explain this phenomenon in the subsequent section. more work to uncover the underlying country-specific nuances.
In Table 15, we have provided a comparison between our study and
5. Discussion other related studies in this area. Notably, our study provides empirical
evidence pertaining to the following three themes: (i) the three EGOV
While the broad focus of our study has been to understand the subindices differ in their effectiveness on Corruption Control, (ii) HC has
relationship between EGOV and Corruption Control and its stability a delayed effect on Corruption Control, and (iii) OS exhibits relative
across time, our secondary intention was to delineate EGOV’s role in ineffectiveness on Corruption Control. In a sense, our study comple
controlling corruption by exploring the individual effects of EGOV ments the work of Elbahnasawy [5] not only because of the recent time
subindices and developing a more nuanced understanding of the rela period considered but also because of the highlight on how the sub
tionship. Our novelty specifically lies in our analytical approach, which indices interplay as EGOV matures. Interestingly, HC needs at least a
enables us to cater to the limitations inherent in current analytic ap two-year gestation period for positively influencing corruption; there
proaches, including the widely used cross-sectional analytical tech fore, deploying HC for corruption requires a more proactive planning
niques as well as the traditional panel estimation techniques. In approach than TI, which starts influencing Corruption Control almost
particular, our analytical model helps us to assess the effects of within- immediately. We specifically contribute to the existing body of literature
country time-invariant factors while simultaneously allowing us to by uncovering the dissimilarities in the influence of the three EGOV
model the efficient and “maximal” error structures of the random effects subindices on Corruption Control and the temporal variations in their
estimation technique, thereby enabling us to combine the best of both impact. We elaborate more on this in the remainder of this section.
fixed effects and random effects estimation techniques [19]. The avail
ability of panel data also allows us to explore the relationships between 5.1. Heterogeneities in the E-Government corruption control relationship
HC/TI/OS and Corruption Control for various time-lags, which is not
possible with cross-sectional data. Most importantly, our approach seeks Our study reveals the following heterogeneities: (i) variations among
to extend the available empirical literature on the EGOV-COR rela countries in terms of their initial Corruption Control levels (captured by
tionship by accounting for country-level variations in both initial cor intercepts) and the improvement rate (captured by slope) of Corruption
ruption levels as well as their longitudinal corruption trajectories while Control, and (ii) dissimilarities in the individual effects of three EGOV
analyzing the EGOV-COR relationship. subindices on Corruption Control. These variations imply that countries
Our study, thus, adds to the existing body of literature by investi at lower initial Corruption Control levels can improve easily and quickly
gating the EGOV-COR relationship at the subindices level while also with little EGOV development effort. However, incremental improve
assessing its stability over time. These findings particularly help explain ments at higher levels of Corruption Control require extensive and
the stark observations discussed earlier in the Introduction section. focused EGOV effort by the governments.
Although Australia and Estonia have witnessed similar EGOV growth Further, we are also able to substantiate our initial apprehension that
rates between 2008 and 2018, Australia has experienced an 11% in dissimilarities exist in the influence of the three EGOV subindices on
crease in corruption in stark contrast to a 12% reduction in corruption Corruption Control. TI is consistent in having a positive effect on Cor
observed in the case of Estonia within the same time period. Despite ruption Control, HC has a delayed positive effect on Corruption Control,
identical EGOV growth, a closer look uncovered stark differences in the and OS is ineffective in reducing corruption. Accordingly, our study
TII change between the two countries. At the same time, HCI and OSI reveals that there exists dissimilarity in impact by HC, TI, and OS on
experienced similar growth trends for the same time period. Estonia’s controlling corruption. For instance, with a one-year lag, HC has a
TII grew at 27%, while Australia’s TII levels grew at a mere 8%. Again, negative impact, TI has a strong positive impact, and OS has no impact
Uruguay and Kazakhstan show stark differences in their corruption on Corruption Control. Our analysis also exhibits that the extent of
levels despite similar EGOV levels. A closer investigation uncovers a dissimilarity is not uniform over the various time-lags studied. For
relatively low TII contribution to the EGOV index for Kazakhstan instance, when we move from one-year lag to two-year lag,
Table 15
Comparison of Related Research.
Paper Andersen 2009 Krishnan et al. 2013 Elbahnasawy 2014 [5] Srivastava et al. 2016 [32] Nam 2018 [15] This Paper
[14] [8]
13
P. Silal et al. Information & Management 60 (2023) 103735
interestingly, HC shows a positive impact, TI as usual shows positive improved processes, which may lead to positive long-term impacts
impact, but the strength has gone down, and OS continues to have no compared to the relatively superficial efficiency-focused initiatives
impact on Corruption Control. Again, for a three-year lag, the positive [104,106].
impact of HC diminishes, TI continues having a positive impact, and OS
now has a negative impact on Corruption Control. Thus, our study 5.4. Implications, limitations, and future works
further indicates that for some subindices, the variation is negligible
across time-lags, whereas for others, the change is substantial. These Our study helps derive several implications for theory and praxis.
differences highlight the need for policymakers and future researchers to Specifically, we seek to contribute to the management and consumption
appropriately account for the unequal time horizon of HC and TI’s in of information in the public service delivery systems by conceptualizing
fluences on Corruption Control while implementing EGOV for anticor and empirically validating the role of EGOV and its subindices on Cor
ruption or while exploring the EGOV-COR relationship. ruption Control. We outline these contributions and implications in the
sections below, followed by some discussions on study limitations and
5.2. Human capital works but with a delay future research avenues.
Given no previous guidelines on the exact amount of time delay 5.4.1. Contributions to theory
required for HC to positively influence Corruption Control, we relied on Our study contributes to the existing body of literature on EGOV and
prior scholarly guidance [89] and started with a one-year lag during the Corruption Control. Firstly, we adopt the RBV perspective to highlight
hypothesis formulation stage while subsequently moving on to higher EGOV and its subindices, namely HC, TI, and OS, as distinct compe
time-lags at the post-hoc analysis stage. Our results highlight the effec tencies. Building on the distinct nature of their competencies, we
tiveness of HC in controlling corruption after a two-year delay. In fact, at conceptualize the unique mechanisms via which they contribute to
a one-year lag, HC has a negative effect on Corruption Control. For a effective information management and consumption, thereby stream
three-year lag, the positive effect continues, though at a smaller effect lining public service delivery by contributing to Corruption Control.
size, finally losing significance at a four-year lag. In particular, the Toward this, we have drawn from the theoretical perspectives under
two-year delay may be attributed to the additional cognitive load and lying Agency Theory, Theory of Media Literacy, Information Quality
hence the time required to process the new information and the proper literature, and Transaction Cost Theory to conceptually link EGDI and its
assimilation of knowledge inputs imparted through interventions tar subindices with Corruption Control. Using this assemblage of theories
geted at HC improvements [95,96]. This observation is in line with helps capture greater complexities inherent in the EGOV-COR relation
evidence from allied domains such as education and human resource ship while also helping circumvent RBV limitations owing to its input-
management, where some lapse of time has been shown to be imperative focused emphasis. Secondly, our study uses the most recent dataset
before positive impacts of literacy and training interventions materialize and employs an analytical approach to allow for country-specific in
[97,98]. As can be inferred from the negative effect of HC on Corruption tercepts and slopes as we attempt to uncover the heterogeneities implicit
Control at a one-year lag, any premature application of acquired in the EGOV-COR relationship across countries and subindices. This
knowledge may suffer from the problems of erroneous beliefs or approach enables us to empirically capture the dissimilarities countries
misinformation resulting in suboptimal or detrimental outcomes [99]. exhibit in their initial corruption levels as well as their decay/growth
The subsequent reduction of HC’s effect post the two-year lag may be rates across time, thereby helping us obtain EGOV-COR estimates robust
attributed to relative obsolescence and decay in the value of knowledge to country-specific differences in their Corruption Control tendencies.
acquired two years back [100]. This finding points to the need for Also, we empirically demonstrate the heterogeneity in the individual
continuous updates on disseminated knowledge from time to time to effects of EGOV subindices on Corruption Control. While TI’s influence
preserve or enhance the value of accumulated knowledge [100,101]. on Corruption Control was positive and stable for all the time-lags
Thus, policymakers may strategize that HC improvements need to begin considered, HC required a longer time-lag to impact Corruption Con
with a two-year head start with handholding during the first two years trol positively. Meanwhile, our findings help unearth the insignificance
and also plan for adequate interventions to update existing knowledge of OS on Corruption Control, which is in stark contrast to previous
from time to time. findings. Thirdly, we add to the time-lag scholarship on the EGOV-COR
relationship by highlighting the variations in impact by EGOV sub
5.3. Service digitalization may not be enough indices over different time-lags. Using all possible time-lags (up to four-
year lag) within our dataset, we observe the maximum effect of TI on
Compared to HC and TI, OS or the digitalization of public service Corruption Control for one-year lag, while HC showed the maximum
delivery infrastructure is found to be insignificant in contributing to effect on Corruption Control for a two-year lag. These identified time-
ward Corruption Control. While this is in stark contrast to the larger lags can serve as an initial reference for future EGOV-COR studies.
body of literature espousing a positive significance of OS in controlling
for corruption, it finds support in an alternative stream of literature. 5.4.2. Practitioner implications
These studies emphasize e-readiness and appropriate high importance to Based on our findings, we also extracted a few actionable items for
enabling factors, such as HC and TI, as a prerequisite to ensuring the policymakers. Firstly, while deciding on their EGOV agenda, govern
effectiveness of digitalizing (often partly) offline public service channels ments should acknowledge the dissimilarities across EGOV subindices in
[102–104]. In fact, some authors note the futility of public service their influence on Corruption Control. Specifically, given the significant
digitalization, or OS, are pointless if citizens are unable to appropriately influence of HC and TI, the EGOV agenda might focus more on devel
use it due to a lack of knowledge or access [102]. Again, the insignifi oping supporting infrastructure in the form of TI and HC for increasing
cance can also be attributed to the nature of digitalization, specifically effectiveness toward mitigating corruption. Secondly, governments
whether the digitalization initiative focuses on improving the effec could appropriately consider the variations in the time taken by EGOV
tiveness of the process or whether it is just limited to increasing the subindices in positively influencing Corruption Control. For instance, TI
efficiency of the process. If the digitalization initiative focuses on effi can lead to positive changes in Corruption Control within a year, while
ciency or being quick “ICT fixes” only, then the emergent OS may further HC requires a two-year lag in positively influencing Corruption Control.
amplify pre-existing inefficiencies prevalent in the offline processes Again, we found that the effect of HC on Corruption Control loses sig
[105]. Conversely, a digitalization initiative targeted toward improving nificance after around four years. Attributing such behavior to knowl
process effectiveness necessitates radical changes, such as edge decay, we suggest that HC strategies may include proper
re-engineering old ineffective and outdated processes to usher in handholding of citizens and ensure continuous update of accumulated
14
P. Silal et al. Information & Management 60 (2023) 103735
knowledge. Thirdly, relative to HC and TI, OS may not be as effective in by including the moderating effect of these government quality mea
positively contributing to Corruption Control as earlier thought, which sures in the EGOV-COR empirical framework. Future works could
may be attributed to the relative importance of the supporting infra particularly assess the moderating effects of RL, EP, and GE indicated by
structure in the form of HC and TI vis-à-vis OS. Alternatively, the our post-hoc moderation analysis (Appendix 2). Finally, given our study
insignificance could also stem from OS being used as a window-dressing focus, we have not explicitly hypothesized for the following: (i)
strategy [107,108] while ignoring service delivery effectiveness. To nonlinear relationships, including the higher marginal growth of Cor
ward this, we suggest that OS strategies, through process re-engineering ruption Control observed for countries with lower initial Corruption
exercise, would focus more on increasing service delivery effectiveness Control levels and (ii) structural breaks, including the Corruption Con
as an integral part of their digitalization strategy instead of limiting their trol saturation points for different countries. Accordingly, future studies
digitalization focus on increasing process efficiency. could include nonlinearity into their models to better explicate the
decreasing marginal growth of Corruption Control with increasing
5.4.3. Limitations and future research avenues Corruption Control levels. Also, identifying Corruption Control satura
While extreme care was taken in conducting the study, there are few tion points for countries would immensely enrich our understanding of
limitations that provide new opportunities for future research. Firstly, the phenomenon.
our study was limited to 102 countries out of 190 odd countries
worldwide, as complete data was not available for the remaining 6. Conclusion
countries. The nonavailability of data is particularly evident for those
countries reeling under political crises or war-like situations. Further Against the backdrop of a dearth of studies exploring the relationship
more, there is a time discontinuity in the dataset used, as EGDI data for between EGOV and Corruption Control at the EGOV subindices level,
some of the years were not available (UN did not publish EGDI data for our study contributes to the body of literature by thoroughly investi
these years). We also acknowledge that the scope of the validity of our gating the relative effectiveness of the three EGOV subindices in con
results is limited to the time period of the dataset under consideration. trolling corruption. In doing so, we also assess the temporal stability of
Accordingly, reassessments of priorly established relationships are these relationships over a considerable period of about a decade. To
required for capturing the time-induced changes. Secondly, the tradi ward this, we draw upon the theoretical insights from RBV, Theory of
tional measure used for OS is limited in the sense that it captures only Media Literacy, Agency Theory, Information Quality literature, and
superficial aspects, such as the number of services, while ignoring Transaction Cost Theory to derive a set of testable hypotheses. Using a
quality measures such as the extent of internal process improvements or panel dataset for 102 countries pertaining to the time period of
service orientations. This limitation may have undermined the effec 2008–2019, we employ an LME estimation method, which enables ac
tiveness of OS as a construct. Future studies may explore the possible curate modeling of reality by allowing the inclusion of individual in
inclusion of these quality measures when constructing the OSI (Online tercepts and slopes for countries. Based on our analysis, TI was found to
Service Index) and re-examine our hypothesized relationships using have a significant positive effect on Corruption Control, while OS was
such advanced measures. Thirdly, our findings could further be com ineffective in controlling corruption. Interestingly, HC was found to
plemented via supporting qualitative studies that examine our hypoth have a delayed impact on Corruption Control, with its effect peaking at
esized relationships at a more micro-level. Besides, recent evidence the two-year lag point. Besides the theoretical implications, our findings
shows wide variations in the implementation of EGOV across political seek to guide policymakers in prioritizing among the three EGOV sub
regimes. As the analysis of political systems was outside the scope of this indices while directing the national EGOV development initiatives more
present study, future studies could examine our model under different effectively toward Corruption Control.
political regimes. Fourth, our study does not include the possible
moderating effects of government quality measures, such as RL, EP, or Declaration of Competing Interest
GE, within the scope of this study. While we have controlled for their
effects in our analyses, future studies could further enhance our findings None
Continent Countries
Africa (20) Algeria; Botswana; Cameroon; Egypt; Gambia; Ghana; Kenya; Madagascar; Malawi; Mali; Mauritius; Morocco; Mozambique; Namibia; Nigeria; Senegal; South
Africa; Tanzania; Uganda; Zimbabwe
Asia (21) Bangladesh; China; India; Indonesia; Israel; Japan; Jordan; Kazakhstan; Republic of Korea; Kuwait; Kyrgyz Republic; Malaysia; Pakistan; Philippines; Qatar;
Saudi Arabia; Singapore; Sri Lanka; Thailand; United Arab Emirates; Vietnam
Europe (39) Albania; Armenia; Austria; Azerbaijan; Belgium; Bulgaria; Croatia; Cyprus; Czech Republic; Denmark; Estonia; Finland; France; Georgia; Germany; Greece;
Hungary; Iceland; Ireland; Italy; Latvia; Lithuania; Luxembourg; Moldova; Netherlands; North Macedonia; Norway; Poland; Portugal; Romania; Russian
Federation; Slovak Republic; Slovenia; Spain; Sweden; Switzerland; Turkey; Ukraine; United Kingdom
North America Canada; Costa Rica; Dominican Republic; El Salvador; Guatemala; Honduras; Jamaica; Mexico; Nicaragua; Panama; Trinidad and Tobago; United States
(12)
Oceania (2) Australia; New Zealand
South America (8) Argentina; Bolivia; Brazil; Chile; Colombia; Paraguay; Peru; Uruguay
15
P. Silal et al. Information & Management 60 (2023) 103735
(continued )
Model Comparison Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Note: Model 1: Random Intercepts and Slope Model with covariates and no moderation; Model 2: Random Intercepts and Slope Model with
covariates and Anti-Press Freedom (APF) as a moderator; Model 3: Random Intercepts and Slope Model with covariates and Rule of Law (RL) as a
moderator; Model 4: Random Intercepts and Slope Model with covariates and Economic Prosperity (EP) as a moderator; Model 5: Random Intercepts
and Slope Model with covariates and Trade Openness (TO) as a moderator; Model 6: Random Intercepts and Slope Model with covariates and
Government Effectiveness (GE) as a moderator.
Appendix 3: Model Residual versus Model Fitted Values Plot for Random Intercepts and Slope Model
16
P. Silal et al. Information & Management 60 (2023) 103735
17
P. Silal et al. Information & Management 60 (2023) 103735
[60] S. Brown, S.A. Hillegeist, How disclosure quality affects the level of information [91] P. Cohen, S.G. West, L.S. Aiken, Applied Multiple Regression/Correlation Analysis
asymmetry, Rev. Account. Stud. 12 (2007) 443–477, https://doi.org/10.1007/ For the Behavioral Sciences, Psychology press, 2014.
s11142-007-9032-5. [92] S.V.C. de Souza, R.G. Junqueira, A procedure to assess linearity by ordinary least
[61] O.E. Barron, H. Qu, Information Asymmetry and the Ex Ante Impact of Public squares method, Anal. Chim. Acta. 552 (2005) 25–35, https://doi.org/10.1016/j.
Disclosure Quality on Price Efficiency and the Cost of Capital: Evidence from a aca.2005.07.043.
Laboratory Market, Account. Rev. 89 (2014) 1269–1297, https://doi.org/ [93] J. Martin, D.D.R. de Adana, A.G. Asuero, Fitting Models to Data: Residual
10.2308/accr-50715. Analysis, a Primer, in: Uncertain. Quantif. Model Calibration, InTech, 2017.
[62] V. Venkatesh, J.Y.L. Thong, F.K.Y. Chan, P.J.H. Hu, Managing citizens’ https://doi.org/10.5772/68049.
uncertainty in e-government services: The mediating and moderating roles of [94] J. Cylus, P. Mladovsky, M. McKee, Is There a Statistical Relationship between
transparency and trust, Inf. Syst. Res. 27 (2016) 87–111, https://doi.org/ Economic Crises and Changes in Government Health Expenditure Growth? An
10.1287/isre.2015.0612. Analysis of Twenty-Four European Countries, Health Serv. Res. 47 (2012)
[63] A. Prasad, S. Shivarajan, Understanding the role of technology in reducing 2204–2224, https://doi.org/10.1111/j.1475-6773.2012.01428.x.
corruption: a transaction cost approach, J. Public Aff. 15 (2015) 22–39, https:// [95] E. Pollock, P. Chandler, J. Sweller, Assimilating complex information, Learn.
doi.org/10.1002/pa.1484. Instr. 12 (2002) 61–86, https://doi.org/10.1016/S0959-4752(01)00016-0.
[64] J. Fledderus, T. Brandsen, M.E. Honingh, User co-production of public service [96] Y. Yamauchi, E.B. Swanson, Local assimilation of an enterprise system: Situated
delivery: An uncertainty approach, Public Policy Adm. 30 (2015) 145–164, learning by means of familiarity pockets, Inf. Organ. 20 (2010) 187–206, https://
https://doi.org/10.1177/0952076715572362. doi.org/10.1016/j.infoandorg.2010.09.002.
[65] Transparency International, Corruption Perceptions Index 2019, 2019. [97] D. Watson, O. Tregaskis, C. Gedikli, O. Vaughn, A. Semkina, Well-being through
https://www.transparency.org/en/cpi/2019. learning: a systematic review of learning interventions in the workplace and their
[66] S. Krishnan, T.S.H. Teo, J. Lymm, Determinants of electronic participation and impact on well-being, Eur. J. Work Organ. Psychol. 27 (2018) 247–268, https://
electronic government maturity: Insights from cross-country data, Int. J. Inf. doi.org/10.1080/1359432X.2018.1435529.
Manage. 37 (2017) 297–312, https://doi.org/10.1016/j.ijinfomgt.2017.03.002. [98] R.L. Bangert-Drowns, M.M. Hurley, B. Wilkinson, The Effects of School-Based
[67] D.A. Kanyam, G. Kostandini, S. Ferreira, The Mobile Phone Revolution: Have Writing-to-Learn Interventions on Academic Achievement: A Meta-Analysis, Rev.
Mobile Phones and the Internet Reduced Corruption in Sub-Saharan Africa? Educ. Res. 74 (2004) 29–58, https://doi.org/10.3102/00346543074001029.
World Dev. 99 (2017) 271–284, https://doi.org/10.1016/j. [99] D.P. Hunt, The concept of knowledge and how to measure it, J. Intellect. Cap. 4
worlddev.2017.05.022. (2003) 100–113, https://doi.org/10.1108/14691930310455414.
[68] P. Dutt, Trade Protection and Bureaucratic Corruption : An Empirical [100] J. Van Loo, A. de Grip, M. de Steur, Skills obsolescence: Causes and cures, Int. J.
Investigation, Can. J. Econ. 42 (2009) 155–183. Manpow. 22 (2001) 121–138, https://doi.org/10.1108/01437720110386430.
[69] S.P. Huntington, Political Order in Changing Societies, Yale University Press, [101] I.P. Murillo, Human capital obsolescence: Some evidence for Spain, Int. J.
1968. Manpow. 32 (2011) 426–445, https://doi.org/10.1108/01437721111148540.
[70] I. Kolstad, A. Wiig, Is Transparency the Key to Reducing Corruption in Resource- [102] D.H. Brown, S. Thompson, Priorities, policies and practice of e-government in a
Rich Countries ? World Dev. 37 (2009) 521–532, https://doi.org/10.1016/j. developing country context: ICT infrastructure and diffusion in Jamaica, Eur. J.
worlddev.2008.07.002. Inf. Syst. 20 (2011) 329–342, https://doi.org/10.1057/ejis.2011.3.
[71] D. Serra, Empirical determinants of corruption: A sensitivity analysis, Public [103] D.H. Brown, N. Lockett, Potential of critical e-applications for engaging SMEs in e-
Choice 126 (2006) 225–256, https://doi.org/10.1007/s11127-006-0286-4. business: A provider perspective, Eur. J. Inf. Syst. 13 (2004) 21–34, https://doi.
[72] D. Treisman, What Have We Learned About the Causes of Corruption from Ten org/10.1057/palgrave.ejis.3000480.
Years of Cross-National Empirical Research? Annu. Rev. Polit. Sci. 10 (2007) [104] W.R. Rose, G.G. Grant, Critical issues pertaining to the planning and
211–244, https://doi.org/10.1146/annurev.polisci.10.081205.095418. implementation of E-Government initiatives, Gov. Inf. Q. 27 (2010) 26–33,
[73] The World Bank, Worldwide Governance Indicators, (2018). https://datacatalog. https://doi.org/10.1016/j.giq.2009.06.002.
worldbank.org/dataset/worldwide-governance-indicators. [105] C. Ciborra, Interpreting e-government and development: Efficiency, transparency
[74] D. Kaufmann, A. Kraay, M. Mastruzzi, The Worldwide Governance Indicators: A or governance at a distance? Inf. Technol. People 18 (2005) 260–279, https://doi.
Summary of Methodology, Data and Analytical Issues, 2010. https://doi.org/ org/10.1108/09593840510615879.
10.1017/S1876404511200046. [106] A. Groznik, A. Kovacic, P. Trkman, The role of business renovation and
[75] X. Zhao, H.D. Xu, E-Government and Corruption : A Longitudinal Analysis of information in E-government, J. Comput. Inf. Syst. 49 (2008) 81–89, https://doi.
Countries, Int. J. Public Adm. 38 (2015) 410–421, https://doi.org/10.1080/ org/10.1080/08874417.2008.11645309.
01900692.2014.942736. [107] L. Torres, V. Pina, B. Acerete, E-government developments on delivering public
[76] S.A. Asongu, Fighting corruption in Africa: do existing corruption-control levels services among EU cities, Gov. Inf. Q. 22 (2005) 217–238, https://doi.org/
matter? Int. J. Dev. Issues 12 (2013) 36–52, https://doi.org/10.1108/ 10.1016/j.giq.2005.02.004.
14468951311322109. [108] S. Rubino-Hallman, N.K. Hanna, New Technologies for Public Sector
[77] The World Bank, World Development Indicators, 2018. http://databank.worldba Transformation, J. E-Government. 3 (2007) 3–39, https://doi.org/10.1300/
nk.org/data/source/world-development-indicators. J399v03n03_02.
[78] Freedom House, Freedom of the Press, 2018. https://freedomhouse.org/report/
freedom-press/freedom-press-2017.
Prakrit Silal: Prakrit Silal is an Assistant Professor in the Information Technology Systems
[79] R.M. O’Brien, A caution regarding rules of thumb for variance inflation factors,
& Analytics Area at Indian Institute of Management Jammu. He has recently completed his
Qual. Quant. 41 (2007) 673–690, https://doi.org/10.1007/s11135-006-9018-6.
Ph.D. from the Department of Management Information Systems at the Indian Institute of
[80] C. Hsiao, Panel data analysis-advantages and challenges, Test 16 (2007) 1–22,
Management, Calcutta. His research interests include Digital Government, E-Participation,
https://doi.org/10.1007/s11749-007-0046-x.
and Information and Communication Technology for Development (ICT4D). His research
[81] B.H. Baltagi, Econometric Analysis of Panel Data, 5th Edition, John Wiley and
publications have appeared in Government Information Quarterly, Marketing Intelligence
Sons., Chichester, 2013. https://www.wiley.com/en-us/Econometric+Analysis+
& Planning, Journal of Global Information Technology Management, Health Marketing
of+Panel+Data%2C+5th+Edition-p-9781118672327.
Quarterly, and International Journal of Technology Diffusion. Some of his works have
[82] C.M. Angst, K.D. Wowak, S.M. Handley, K. Kelley, Antecedents of Information
been presented in international reputed conferences including International Conference on
Systems Sourcing Strategies in U.S. Hospitals: A Longitudinal Study, MIS Q 41
Information Systems (ICIS) and Academy of Management (AOM) Annual Meetings.
(2017) 1129–1152, https://doi.org/10.25300/MISQ/2017/41.4.06.
[83] A.C. Johnston, M. Warkentin, M. McBride, L. Carter, Dispositional and situational
factors: Influences on information security policy violations, Eur. J. Inf. Syst. 25 Ashutosh Jha: Ashutosh Jha is an Assistant Professor in the Information Management
(2016) 231–251, https://doi.org/10.1057/ejis.2015.15. Group at SPJIMR, Mumbai. His research interests include Adoption and Diffusion of Mo
[84] A. Hardin, C.A. Looney, G.D. Moody, Assessing the Credibility of Decisional bile Information Systems, Techno-Economics of Next Generation Networks, and E-
Guidance Delivered by Information Systems, J. Manag. Inf. Syst. 34 (2017) Governance. His research publications have appeared in Information Systems Frontiers,
1143–1168, https://doi.org/10.1080/07421222.2017.1394073. Technological Forecasting and Social Change, IEEE Access, Applied Economics Letters,
[85] D.J. Barr, R. Levy, C. Scheepers, H.J. Tily, Random effects structure for IIMB Management Review, and Decision. His works have also been presented in several
confirmatory hypothesis testing: Keep it maximal, J. Mem. Lang. 68 (2013) international conferences, including International Conference on Information Systems
255–278, https://doi.org/10.1016/j.jml.2012.11.001. (ICIS), European Conference on Information Systems (ECIS), Hawaii International Con
[86] F. Falkenström, S. Finkel, R. Sandell, J.A. Rubel, R. Holmqvist, Dynamic models ference on System Sciences (HICSS), International Association for Management of Tech
of individual change in psychotherapy process research, J. Consult. Clin. Psychol. nology (IAMOT), and IEEE, among others.
85 (2017) 537–549, https://doi.org/10.1037/ccp0000203.
[87] B. Winter, Linear models and linear mixed effects models in R with linguistic
Debashis Saha: Debashis Saha, currently an Endowed Chair Professor in the MIS area in
applications, (2013) 1–22. http://arxiv.org/abs/1308.5499.
IIM Calcutta, has been teaching IT and Digital for more than 30 years now. His research
[88] A. Bell, M. Fairbrother, K. Jones, Fixed and random effects models: making an
interests include Fintech, ICT4D, Emerging IT paradigms, Digital Disruption, E-Gover
informed choice, Qual. Quant. 53 (2019) 1051–1074, https://doi.org/10.1007/
nance, Digital Business Transformation, IT strategy and governance, and Business-driven
s11135-018-0802-x.
IT. He has co-supervised 17 doctoral theses, published about 300 research papers in
[89] Y. Griep, I. Vranjes, J.M. Kraak, L. Dudda, Y. Li, Start Small, not Random: Why
various journals/conferences, and directed four funded projects on IT. His research has
does Justifying your Time-Lag Matter? Span. J. Psychol. 24 (2021) e45, https://
appeared in major journals including Decision Support Systems, OMEGA-The Interna
doi.org/10.1017/SJP.2021.42.
tional Journal of Management Science, Information System Frontiers, among others. He
[90] C. Dormann, M.A. Griffin, Optimal time lags in panel studies, Psychol. Methods
has also served on the editorial board of selected international journals.
20 (2015) 489–505, https://doi.org/10.1037/met0000041.
18