Why Do High-Tech Firms Offer Perks at Work? The bulk of this work was done by a student. Extended Abstract Research Question and Background We study whether and why high-tech firms rely more heavily on a class of non-wage benefits that includes free meals, transportation subsidies, and athletic facilities ( work-related perks ). Given the levels of competition for skilled labor, managers, policy makers, and researchers have paid growing attention to compensation strategies that best incentivize and attract high-skilled technical workers (Andersson et al. 2009; Bloom et al. 2007). This has also become a topic of interest for the press, who frequently cover employer provision of nontraditional benefits, such as the extended parental leave policies recently announced by Netflix (Rosen 2015) or the growing incidence of benefits such as ping-pong tables, free lunches, on-site dry cleaning, and yoga classes that are meant to improve on-site working conditions at many high-tech companies (Cohen 2015). Specifically, we argue that work-related perks complement IT innovation because they attract and motivate IT workers who can adapt to new technologies. We present a principal-agent model that analyzes the firm s optimal compensation mix (Marino and Zabojnik 2008; Oyer 2008). We argue that offering work-related perks, instead of other types of compensation such as performance pay, enables employers to extract greater profits if workers output is less certain and measurable. Moreover, employers use these perks to attract the types of employees who value this type of compensation, and who therefore may fit better within the employer s work culture. Therefore, we argue that IT innovators, in particular, provide work-related perks to
motivate IT workers to stay on-the-job for longer working hours. This is particularly important for IT workers at firms that are IT innovators because they value on-the-job skill acquisition, and which is most efficient when flexibly switching between work and leisure. The above propositions are captured in the following hypotheses, which we test in our empirical analysis: H1: IT workers derive greater value from work-related perks if their employers are engaged in IT innovation. H2: The value that IT workers derive from work-related perks increases in the value they derive from on-the-job skill acquisition. Data and Measures Our primary data source is from a popular career intelligence website. It spans the years from 2008 to 2014 and contains over a half million reviews of employers and their interview practices. Figure 1 presents a sample employer review from the website. This review describes a company culture characterized by substantial levels of work-related perks and technical work conducted in a fast moving environment. We construct our key variables by labeling the reviews, including Pros, Cons and interview reviews, according to whether they contain the keywords related to our constructs of interests. The basic intuition behind our algorithm is that employees voluntarily mention employer attributes in the Pros section of the reviews if they derive value from them. To measure work-related perks, we adopt a data-centric approach that produces the keywords related to all relevant benefits. Figure 2 plots compensation structure for several typical whitecollar professions. To measure IT innovation, we focus on the incidence of three emerging technologies: data analytics, cloud computing, and mobile computing, and examine questions asked by the employer during technical interviews. We use workers references to Skill
Acquisition in the Pros sections to measure the value derived from on-the-job skill acquisition. We create a similar measure of working time using workers references to working hours in the Cons sections. Analysis We estimate the likelihood that employees mention work-related perks using a discrete choice framework. Suppose worker i derives value from her employer j s n attributes, represented as a vector: v!" (a!, a!, a!, a! ). She will comment on attribute a k in the pros section of an online review if its value passes some threshold a! > τ!", and will comment in the cons section if it falls under some threshold a! < τ!". Prob a!"#"$%&' > τ!"#$#%&'(!" = β! + β! ITInnovation!" +β! Working Time!" + β! Occupation!" +β! Skills Acuisition!" + β! controls ij + γ! + ε!" In Table 1, we report results from logit regressions. Each regression includes firm fixedeffects, and the relevant estimates capture variation among workers within the same firm. The probability of an employee referencing benefits in their review is specified as a function of her occupation, IT innovation, on-the-job skill acquisition, and working time. We find that employers are 9.6% more likely to offer work-related perks if they are IT innovators. The coefficient estimate of the IT occupation variable in Column (1) is 0.147 (z=4.65), and the marginal effect computed using the population average is 8%. Whatever the economic causes, these results suggest that IT workers place higher values on work-related perks and vacation benefits than their colleagues in other occupations from the same firm, after controlling for their labor market conditions. The coefficient on IT innovation is positive and statistically significant (z=6.01). We do not find comparable results on our IT innovation measure for other benefits.
We also analyze the ways in which employees reference time spent at work in different firms and occupations, and find support for the prediction that workplaces where perks are valued require longer hours from their workers. We do not find similar patterns for other types of benefits such as health insurance or retirement. In Column (1), the coefficient of on-the-job skill acquisition is positive and statistically significant (z=5.84). IT workers who value on-the-job skill acquisition derive greater value from work-related perks, confirming this assertion. We do not find similar effects for non-it workers. Work-related perks provide the highest value to workers capable of acquiring new skills quickly. Collectively, the evidence suggests that workrelated perks have become a channel through which to increase efficiency and generate economic surplus through the employer-employee match. Contributions and Future Directions Our results have important managerial implications. First, the existing literature linking compensation to firm performance provides limited evidence about the role of the types of nonwage benefits that we analyze, despite the fact that they comprise a growing proportion of labor costs. Second, our findings suggest that tailoring incentive strategies towards IT investment is promising as it helps firms to cope with rising labor costs due to the scarcity of technical talent in competitive labor markets. There is substantial room for future research in this area. Although the literature on the interplay between organizational practices and IT innovation is growing, we still lack a systematic way to quantify management and HR practices in order to assess how they impact corporate performance. As IT-enabled organizational disruption becomes commonplace, questions about how organizations can adapt to rapid technical change must be addressed. For instance, beyond using the compensation strategies studied in this paper, the ways in which firms
adjust their recruiting and training strategies to manage key IT talent more effectively remain underexplored. The optimal organizational structure to facilitate IT innovation and the optimal ways to organize innovation tasks are both open questions that deserve further attention. References Andersson, F., Freedman, M., Haltiwanger, J., Lane, J., and Shaw, K. 2009. Reaching for the Stars: Who Pays for Talent in Innovative Industries?*, The Economic Journal (119:538), pp. F308 F332. Bloom, N., Sadun, R., and Van Reenen, J. 2007. Americans do IT better: US multinationals and the productivity miracle, National Bureau of Economic Research (available at http://www.nber.org/papers/w13085). Marino, A. M., and Zabojnik, J. 2008. Work-related perks, agency problems, and optimal incentive contracts, The RAND Journal of Economics (39:2), pp. 565 585. Oyer, P. 2008. Salary or benefits?, in Research in Labor Economics (Vol. 28), Bingley: Emerald (MCB UP ), pp. 429 467 (available at http://www.emeraldinsight.com/10.1016/s0147-9121(08)28013-1).
Figure 1: Sample employer review
Figure 2: Salary and benefits in white-collar occupations Figure notes: this figure plots the average salary and share of worker mentioning benefits (WP and H&R) for typical white-collar occupations. Salary and other pay are computed using workers self-reported salary and total income.
Table 1: Logit regressions of benefits on occupation, IT innovation, skill acquisition, and working time (1) (2) (3) (4) (5) (6) DV: P (Benefits) WP FAMILY VACATION HEALTH RETIREMENT WAGES Model FE FE FE FE FE FE IT Innovation 0.305*** 0.0773-0.0953* 0.00673-0.202*** -0.0561** (0.050) (0.074) (0.055) (0.057) (0.076) (0.027) IT Workers 0.147*** 0.0387 0.151*** -0.0501-0.0568-0.0656*** (0.0316) (0.0487) (0.0379) (0.0408) (0.0521) (0.0191) IT Skills Acquisition 1.436*** -0.638 0.0694-0.223-0.268-0.0163 (0.246) (0.413) (0.279) (0.303) (0.386) (0.140) Non-IT Skills Acquisition -0.155** -0.0630-0.289*** -0.112-0.109 0.913*** (0.067) (0.101) (0.086) (0.087) (0.113) (0.033) IT Working Time 1.478*** 0.0316 0.492** 0.483** 0.553* 0.0490 (0.217) (0.326) (0.234) (0.240) (0.309) (0.123) Non-IT Working Time 0.369*** 0.0634 0.265*** 0.253*** 0.585*** 0.391*** (0.050) (0.080) (0.060) (0.068) (0.058) (0.0305) Sales & Marketing -0.241*** -0.146** -0.121** -0.165*** -0.149* -0.0823*** (0.051) (0.060) (0.052) (0.060) (0.079) (0.025) Service 0.108** -0.0738 0.0433 0.00741-0.176* -0.0309 (0.048) (0.070) (0.056) (0.064) (0.091) (0.029) Operations -0.105* 0.0807-0.0968 0.0597 0.0724 0.104*** (0.064) (0.077) (0.064) (0.064) (0.086) (0.031) Financial 0.00336-0.128-0.0961-0.0122 0.0744-0.0341 (0.080) (0.122) (0.090) (0.094) (0.114) (0.046) Teacher 0.347 1.572*** 0.125-0.658-9.797 0.0788 (0.841) (0.476) (0.914) (1.278) (379.7) (0.445) Other Controls X X X X X X City Controls X X X X X X FE Firm Firm Firm Firm Firm Firm Observations 159,937 156,160 159,892 125,051 110,999 136,648 This table reports results from logistic regressions with fixed effects. The dependent variables are whether benefits-related keywords appear in the review. IT Skills and IT Working Time are establishment level shares of IT workers referencing skills and working time keywords in the review. IT Workers, Sales & Marketing, Service, Operation, Financial and Teacher are dummy variables indicating workers occupation group. The occupation groups are classified using workers self-reported job titles. Other controls include whether the worker is currently employed and Log (establishment reviews). Robust standard errors are reported in parentheses. Significance level: *** p<0.01, ** p<0.05, * p<0.1;