Fiscal slack or environmental pressures: which matters more for technological innovation assimilation? A configurational approach

Abstract

Prior studies suggest that fiscal slack may positively affect technological innovation in the public sector but yield inconsistency between theoretical expectations and empirical findings. To address this research gap, this study develops a configurational explanation combining fiscal slack and environmental pressures by employing fs-QCA in the context of Chinese provincial governments’ online public service provision. The findings reveal that the role of fiscal slack is environmentally dependent on explaining technological innovation and that environmental pressures seem to be more important. That is, fiscal slack is likely irrelevant to a high degree of technological innovation assimilation, while the demanding environments of multiple environmental pressures matter. Relative fiscal scarcity may be somewhat relevant to a low degree of technological innovation assimilation only when combined with a more relaxed environment. Several theoretical and practical implications are discussed.

Notes

Acknowledgments

An earlier version of this article was presented at Renmin University of China in 2017. We would like to thank Tom Christensen, Ran Chen, Liang Ma, Ning Leng, Yixin Dai, Zhilin Liu, and Ke Meng for their helpful comments and suggestions.

Notes

1 Walker (2014) draws the conclusion based on the support score. The support score refers to “the number of tests that are consistent with a publicness hypothesis as a percentage of all the tests that are reported in a study” (Boyne 2002:105). Researchers can combine all of the support scores derived from each study to obtain an aggregate support score that performs as an indicator for researchers to judge whether a certain theoretical relationship has been tested by most empirical studies. According to Boyne (2002:113), the value of 50% is the anchor for supporting a hypothesis by most empirical tests. However, Walker’s (2014) literature review shows that the value is far below 50% to support the hypothesis of the positive relationship between slack resources and innovation. The value to support a nonsignificant relationship between slack resources and innovation is above 50%.

2 We acknowledge that there is a potential nonlinear relationship between complexity and innovation. However, according to an empirical test by Walker et al. (2015), this relationship is not robustly supported.

3 The provincial governments include 23 provinces (including Taiwan), four municipalities, and five autonomous regions. However, due to Taiwan’s data limit, the analysis of this article focuses on the other 31 provincial governments.

4 According to Ragin (2008:44), consistency refers to “the degree to which the cases sharing a given combination of conditions agree in displaying the outcome in question,” and coverage refers to “the degree to which a cause or causal combination accounts for instance of an outcome.” These two measures can be applied to both necessity and sufficiency analyses. Moreover, scholars should conduct consistency tests before coverage tests to assess the quality of causal inference in QCA. If the causal inference fails to pass the consistency test, then there is no need for a further coverage test.

5 In this step, the minimum cutoff threshold of consistency is 0.75. Scholars can generally choose among 0.8, 0.85, 0.9, and 0.95 in practice based on the trade-off between consistency and coverage in the final result. However, in some situations, 0.8 as the cutoff threshold may reduce the consistency of the configurations to the outcome of interest, and 0.95 may reduce the coverage of the configurations such that fewer meaningful cases would be excluded in the QCA analysis. In this study, we choose 0.9 as the cutoff threshold for the analysis of a high degree of assimilation and 0.85 for a low degree of assimilation. These two values are set in consideration of the best trade-off between consistency and coverage in this study.

6 According to Fiss (2011), in the simplification process, researchers can obtain a parsimonious solution by including counterfactuals (without distinguishing between easy or difficult counterfactuals) or an intermediate solution by including easy counterfactuals. The intermediate solution is suggested by Ragin (2008:166).

7 “Core conditions” refer to the conditions that appear in both parsimonious and intermediate solutions. Conditions that appear only in intermediate solutions but not in parsimonious solutions are “complementary conditions” (Ragin and Fiss 2008:204). Core conditions present stronger relationships with the given outcome than complementary conditions.

8 This measurement is preferable in two ways. First, it is the only annually conducted performance assessment score that we could find regarding OPSP in China. In addition, this performance score has also been authorized and employed by the central state, although several other third-party institutes, such as the China Software Testing Center and Guomai Institute, annually assessed government websites based mainly on the criteria incorporating government information accessibility, online participation, and software applications. However, the objects of these assessments are government websites, which are different from online public service platforms. Second, one research team at Tsinghua University once conducted a similar performance assessment of OPSP; however, the time span is constrained to 2016 and there was no additional assessment in other years. In addition, the performance scores derived by the China National School of Administration and Tsinghua University are highly correlated (r = 0.657, p < 0.001), which we think could provide additional support of the measurement validity. In addition, among other publicly reported data, we also note that the government website performance score derived by Guomai Institute incorporates a subindicator regarding online service on government website. Even though it is not an ideal measurement to capture the whole conditions of the scope and quality of OPSP, it can help us further conduct the validity check. The result shows that these two measurements are also highly correlated (r = 0.715, p < 0.01). Thus, while we are constrained to the performance score derived by one scholar institute, we do believe that the composite score we employ in this article could capture something important regarding the performance of OPSP.

9 This index is developed by Lieberson (1969) to measure diversity within a population. Additionally, its formula is equivalent to the Blau index employed in several public management studies.

10 We acknowledge that this approach may fall short in measuring the “learning-model effect” between local peers. However, according to Zhu (2014), it is possible that local governments in China may refuse to completely imitate their peers’ innovative policy instruments and prefer to develop their own innovative policy instruments in the hope of being pioneers in innovation adoption and obtaining praise from the upper authority. This is because complete learning from peers may potentially highlight their rivals’ ranks in evaluation-based tournaments. In addition, it may be difficult for researchers to develop a measure that can unarguably discern the mechanism of learning and competition among local peers (Grimmelikhuijsen and Feeney 2017). Several empirical studies in the field of policy diffusion apply the “directed dyad-year Event History Analysis (EHA)” to test “learning-model effect” among local peers (Volden 2006; Zhang and Zhu 2018). However, as a specific type of EHA, this approach requires special data structures and is not suitable for QCA analysis. Perhaps this is one of the reasons for researchers’ employment of the approach of international learning channels to describe learning in several other studies. In this regard, we call for improvements in this measurement.

11 This program was jointly launched by the Central Compilation & Translation Bureau, Party School of the Central Committee, and Peking University in 2000. Furthermore, it is an important material source for scholars to explore the innovation of the Chinese public sector (Zhao 2012; Wu et al. 2013).

12 Ragin (2008:85) suggests two approaches for calibration: the “direct method” and the “indirect method.” The direct method is widely used in fs-QCA analysis (Frambach et al. 2016; Andrews et al. 2016; Wang 2016).

13 For example, there is one exception. According to the results of Tosmana, the crossover point for environmental complexity is 0.67, and one case has this value. In this situation, if we set this value as a crossover point for environmental complexity, it would be difficult for researchers to further assess whether the case with this value is more in than out of the set. Thus, we further set 0.7 as the crossover point with the help of the authors’ qualitative knowledge because there is a qualitative difference between the cases with values of environmental complexity ranging from 0.69 to 0.71, and there is no case with this value. Most of the cases with values above 0.7 are municipalities directly under the central state or eastern coastal provinces that have a large number of labor immigrants from other middle-western provinces. Compared to those in other provinces, governments in these areas serve much more heterogeneous populations.

14 The threshold of consistency for necessity analysis is 0.9. However, even if a condition satisfies the threshold of consistency in necessity analysis, it is still required to conduct a coverage test for trivialness (irrelevance). In this regard, the threshold of coverage should be relatively high. Researchers can also employ an XY plot in which the X axis refers to the membership score of the condition and the Y axis refers to the membership score of the outcome, in which there are many cases clustering close to the right Y axis, which indicates trivialness (Schneider and Wagemann 2012:144–7). In this article, we find that although three conditions pass the consistency test, they fail the coverage test of trivialness. More details are provided in the Appendix.

15 Solution consistency above 0.8 is supposed to present an acceptable fit (Fiss 2011). Although there is no suggested threshold for solution coverage (Schneider and Wagemann 2012:139), a value of solution coverage of fs-QCA ranging from 0.5 to 0.7 is relatively satisfactory in many published peer-reviewed papers (Thomann 2015; Frambach et al. 2016; Andrews et al. 2016).

16 Due to space limitations, we do not elaborate on the meanings of solution consistency and coverage in the section on low degree of technological innovation assimilation.

17 Due to space limitations, we do not elaborate on the meanings of consistency, raw coverage, and unique coverage for each configuration.

Additional information

Notes on contributors

Ziteng Fan

Ziteng Fan ([email protected]) is a Ph.D. candidate at the School of Public Policy & Management in Tsinghua University. His research mainly focuses on e-government, public sector innovation, organizational theory, and citizen-state interactions.

Qingguo Meng

Qingguo Meng ([email protected]) is a professor at the School of Public Policy & Management in Tsinghua University. His research mainly focuses on e-government and performance management.

Na Wei

Na Wei ([email protected]) is a post-doctoral fellow at the School of Public Policy & Management in Tsinghua University. Her research mainly concentrates on e-government and collaborative governance.