Gender differences in performance-driven managerial innovation: evidence from US nursing homes

Abstract

The management literature has highlighted the role of a manager’s gender in adopting and practicing managerial innovation. The conditions that affect female (or male) managers’ decision making on innovations, however, have been less explored. Using a national survey of top-level administrators in US nursing homes and archival nursing home quality data, this study examines how performance information shapes gender differences in managerial innovation adoption. We find that female managers are more likely to adopt innovations relative to male managers, particularly when they perform better than they have in past years. Our findings, however, do not support a gender difference in innovation adoption when a nursing home performs worse than other competing organizations. The findings provide important implications on how a manager’s gender produces systematic differences in innovation adoption related to performance information.

Notes

1 Specifically, studies suggest that a manager’s age and tenure are negatively associated with innovation because older and experienced managers have been socialized into accepting institutionalized managerial practices and organizational routines (Hambrick and Mason 1984; Huber et al. 1993). Education, in contrast, is widely known to enhance the adoption of innovation (Damanpour and Schneider 2006). Highly educated managers have more knowledge and expertise, which inspires them to accept new ideas and changes. Young et al. (2001), for example, find that younger and more educated managers in hospitals are more likely to adopt an innovative strategy such as Total Quality Management (TQM).

2 This perspective reflects the traditional view on gender differences in managerial styles. The early literature describes the typical features of men’s managerial style as “‘directive’, ‘self-centered’, ‘self-interested’, ‘decisive’, ‘aggressive’, and ‘task-oriented’” while adjectives used to describe women’s styles include “‘participative’, ‘collaborative’, ‘co-operative’, ‘coaching style’, ‘people-oriented’, and ‘caring’” (Wajcman 1996, 342).

3 These studies demonstrate gender differences in the self-perceptions of innovation rather than differences in actual innovation outcomes. Although perceptual judgments of innovation do not perfectly predict innovative behaviors, we expect that these two are positively related.

4 Historical and social performance gaps not only influence managers’ decision-making but also shape citizens’ assessments of performance (see Olsen 2017).

5 Using a relative risk model, Nicholson-Crotty, Nicholson-Crotty, and Fernandez (2017) test the quadratic U-shaped relationship between performance and innovation. They theorize that public managers will be less risk averse and more innovative when they think their organizations either fall short of or exceed performance. Their results support this theoretical argument.

6 Nählinder (2010) also argues that women and men pursue innovation for different reasons, and cultural expectations of how a woman is supposed to act creates more pressure for women not to pursue new ideas even in women-dominated occupations.

7 This sample is representative of the population of nursing homes in terms of both key variables and control variables. The sample also has a similar distribution to the population in terms of performance indicators.

8 The star ratings make a comprehensive performance measure because they incorporate a wide range of nursing home quality indicators, including health deficiencies, staffing, and healthcare quality. Furthermore, the star ratings cover longer time spans of performance, compared to other performance indicators. The star ratings reported each year are constructed using three years of cumulative performance of health deficiencies, which means that star ratings in 2013 actually reflect the 3-year estimated performance from year 2011 to 2013. This measurement method helps us to capture more comprehensive historical and social gaps.

9 The star ratings are constructed using three performance dimensions—health deficiencies, staffing, and quality ratings—which have different reporting time periods given the year. Particularly, health deficiencies in star ratings are constructed using recent three years of health deficiency reports and revisits. In calculating the total weighted score of health deficiencies, the most recent survey is weighted more heavily than earlier surveys; the weighting factor assigned to each survey cycle are 1/2 (most recent year), 1/3 (the previous year), and 1/6 (the second prior year), respectively. Given this weighting method, our historical performance gap between 2012 and 2013 star ratings actually reflects the 3-year estimates health deficiencies of 2010–2012 and 2011–2013 as a part of the star ratings. For more about the five-star quality rating system, see CMS (2018).

10 We use the county as a reference unit for the social gap because elderly healthcare is long-term care in community-based settings. According to the National Council for Aging Care (2017), proximity to family is one of the most important factors when selecting a nursing home (see also Reinardy and Kane 1999). Research also often defines a county as a geographic market for elderly healthcare (Amirkhanyan et al. 2018; Bowblis 2012).

11 Additionally, we conduct several diagnostic tests to check whether our models meet the assumptions of regression. Even the most elaborate model (with interactions between gender and performance gaps) meets all regression assumptions. Breusch-Pagan test results show that the null hypothesis of homoskedasticity is not rejected (${\chi }^2$ = 0.105, p-value = 0.746). Ramsey's regression specification error test shows that the model has an appropriate functional form (F = 1.737, p-value = 0.158). Tests show our model is correctly specified (linktest t = 0.735, p-value = 0.463). There are no influential observations (Cook's distance), and residuals are also normally distributed (Shapiro-Wilk normality test). Overall multicollinearity is modest with a mean VIF of 1.73.

12 For the robustness checks, we conduct additional analyses. First, there are cases where only one nursing home exists in a county, and in this case, the value of our social gap will be 0. To test whether the monopoly market structure influences our results, we exclude monopoly nursing homes from our analyses. The results remain the same. Second, we account for state variation by adding state fixed effects and estimate models using standard errors clustered at the county level. The results are also similar to those reported here.

13 The guide to choosing a nursing home published by CMS indicates that the first step in choosing a facility is finding one in your area (CMS 2017). Nursing Home Compare also guides potential residents to choose a facility based on their zip code, and allows them to expand the searching area from 1 mile to 200 miles.

14 For instance, health inspection, the main quality measure in the five-star quality rating, is measured based on the number, scope and severity of facility deficiencies identified by state investigators. The nursing homes with high deficiencies are required to have revisits to ensure their compliance with the regulations. The number of deficiencies and revisits are key to determine nursing home closures; therefore, only low-performing nursing homes at the bottom may react to the change in ratings, seeing it as a threat to their market survival.

15 Recognizing that performance gaps across nursing homes in a county might not be very high, we also considered a larger unit when calculating social performance gaps. We consider both metropolitan areas and states, but the main results are not significantly different from that of the county-level social gap results.