Cloud computing and firm growth

Timothy DeStefano, Richard Kneller, Jonathan Timmis 06 May 2020

a

A

Over the last decade a fundamental shift has occurred in the way firms access digital technologies. Traditionally, the acquisition of information and communication technology (ICT) required businesses to make upfront, sunk investments in hardware and software in order to maintain large IT departments. Today, firms can acquire their storage, processing and software needs through cloud computing (Van Ark 2016, OECD 2015).1 Cloud providers offer such services ‘on demand’ via pay-as-you-go subscriptions, accessible on multiple devices in multiple locations. Hence, IT may shift from a centralized sunk cost to a largely variable cost. This has led to changes in firm behaviour that go beyond simply substituting the means of accessing IT (Iansiti and Richards 2011, OECD 2015, Jin and McElheran 2017).

Changing IT to a largely variable cost provides an advantage especially for young firms which are innovating new products and services, as they can now scale operations quickly without the up-front acquisition of a mass of ICT assets, labour and/or establishments. This has become known as ‘scale without mass’.2 Incumbent firms in contrast, have faced the difficulty of migrating legacy software systems on to the cloud, reducing the benefits of this change.

Cloud technologies are likely to primarily affect the organization of production for such firms, reducing the sequencing and coordination problem of working in teams. Employees can work on the same projects simultaneously and be mobile in their work across space and time, reducing the reliance on fixed PCs that are connected to the internal hardware and software in specific firm locations at fixed times. This impacts the spatial distribution of activity and leads to greater geographic dispersion (Bloom et al. 2014).

Evidence from firm-level data

To analyse the implications of cloud computing, we (DeStefano et al. 2020) use newly available data for the UK that allows us to determine the timing of adoption of cloud services by firms, along with the types of cloud services used (including data and storage, email, software etc.). Such detailed measures of cloud adoption at the firm level have not previously been available to researchers on this topic. 

We employ a number of traditional measures of firm performance, such as growth in sales, employment, labour productivity and establishment births and deaths, as well as new measures of geographic dispersion.  These include the unweighted and weighted average distance between establishments and the firm headquarters (weighted by the share of establishment employment in firm employment) and a distance-employment covariance term to measure how employment is distributed across plants.3 To track changes in employment further this is combined with data for a random 1% sample of all workers in the UK.

The paper addresses the endogeneity of cloud adoption using firm-specific instruments on the availability and expected speeds of high-speed fibre broadband - a technological prerequisite for adopting the cloud.  A stable, high-speed broadband connection is required to allow the large flows of data between the cloud service providers and users (ITU 2017).4 

Cloud computing and firm performance

The results are consistent with the idea of differences in the effects of cloud technologies across young and incumbent firms (defined as firms less than and more than 5 years old at the start of the sample period, respectively). The adoption of cloud technologies leads to significant increases in firm scale, measured by employment and sales, for both young and incumbent firms. The effect for young firms is much stronger however (see Figure 1). For example, we find that cloud adoption results in a coefficient of 1.087 for young firms and 0.735 for incumbents, which translates to an annual increase in employment of roughly 28% and 15% for young and incumbent firms, respectively. The results also suggest that that the increase in sales and employment are approximately equal suggesting that there is no significant effect of cloud technologies on labour productivity.

We also find evidence of differences in reorganisation in terms of number of establishments. Younger firms who adopt cloud technologies because they gain access to a fibre connection, are significantly less likely to become multi-establishment firms (see Figure 1). Conversely, we find no effect on the probability of becoming multi-establishment for incumbent firms, but some evidence of experimentation and reorganization through the closure of establishments. For both young and incumbent firms we find no significant effect on the creation of new establishments.

Figure 1 Impact of cloud services on firm growth: young vs incumbents (IV regressions)

Note: The following figure illustrates the effect of cloud adoption on firm performance (for young and incumbent firms) estimated with instrumental variable regressions. All results above are significant at either the 1%, 5% or 10% level. Figure excludes insignificant results. Regressions reflect years 2008, 2013 and 2015.

Cloud computing and geographic organisation

For incumbent firms, cloud adoption leads to geographic dispersion of activity – with the average employee working 25km farther from the headquarters (see average distance (weighted) in Figure 2). This does not occur because of the opening of new plants at farther distances but rather predominantly through a shift in employment across plants reflected by the increase in the Distance-Employment Covariance. Moreover, cloud adoption increases the number of local authorities a firm operates establishments in by roughly 4% annually over the sample period. Interestingly, changes in geographic fragmentation due to cloud adoption do not appear to be driven by a shift towards locations in which average wages or office space costs are lower. For young firms, that tend to be small, there is little impact of cloud adoption on geographic dispersion.

To examine the shift in the distribution of firm employment further, we also analyse changes on the level of the employee using matched employer-employee data.  We find that workers in establishments that adopt cloud services are significantly more likely to move workplaces, compared to workers in establishments that have not yet adopted the technology.  However, we find no systematic evidence of movement of employees towards or away from the headquarters, but rather across plants. This is consistent with the idea that cloud services tend to affect the organization of production rather than the costs of monitoring and communication by management at the headquarters for example.

Figure 2 Impact of cloud services on geographic dispersion of firms: incumbents (IV Regressions)

Note: The figure illustrates the effect of cloud adoption on the geographic dispersion of firms estimated with instrumental variable regressions. All results above are significant at either the 1%, 5% or 10% level. Figure excludes insignificant results. Regressions reflect years 2008, 2013 and 2015.

Conclusion

Consistent with much of the anecdotal evidence, there are differential impacts of cloud adoption for younger and incumbent firms. Cloud services along with the fibre infrastructure enable young firms to scale up without increasing their geographic footprint while incumbents use the technology to reorganize, reduce their costs and increase their geographical dispersion. Moreover, our results show that cloud services enhance worker mobility resulting in the movement of workers across plants. 

This evidence suggests cloud services are distinct from earlier IT technologies, which reinforced the scale advantages of incumbents (see for instance Lashkari et al. 2019).  Cloud adoption is linked to a decline of firm investments in IT capital and software, indicating that cloud services allow firms to substitute away from owning IT equipment. Cloud adoption also decentralizes data availability throughout the firm, going beyond earlier ICT that allowed decentralized information access for specific tasks or workers, such as Enterprise Research Planning and CAD/CAM software (Bloom et al. 2014).  In dispersing economic activity, cloud services are hence similar to previous waves of ICT. 

References

Bloom, N, L Garicano, R Sadun, J Van Reenen (2014), “The distinct effects of information technology and communication technology on firm organization”, Management Science 60: 2859-2885. 

Deloitte (2017), “Technology, Media and Telecommunications Predictions: 2017”, accessed 09 Nov 2019.

DeStefano, T, R Kneller and J Timmis (2018), “Broadband Infrastructure, ICT use and Firm Performance: Evidence for UK firms”, Journal of Economic Behavior and Organization 155: 110 - 139.

DeStefano, T, R Kneller and J Timmis (2020), "Cloud computing and firm growth", Discussion Papers 2020-02, University of Nottingham, GEP.

Eurostat (2018), “Cloud Computing - Statistics on the Use by Enterprises”.

Iansiti, M and G Richards (2011), “Economic Impact of Cloud Computing”, White Paper.

ITU (2017), “Final Report: Question 3/1: Access to cloud computing: Challenges and opportunities for developing countries”.

Jin, W and K McElheran (2017), “Economies Before Scale: Survival and Performance of Young Establishments in the Age of Cloud Computing”, Rotman School of Management Working Paper No. 3112901.

Lashkari, D, A Bauer and J Boussard (2019), “Information Technology and Returns to Scale”, Mimeo.

Lesser, A (2017), “The Cloud Vs. In-House Infrastructure: Deciding Which Is Best For Your Organization”, Forbes, accessed on 09 Nov 2019.

OECD (2015), “OECD Digital Economy Outlook 2015”, OECD Publishing, Paris.

Van Ark, B (2016), “The Productivity Paradox of the New Digital Economy”, International Productivity Monitor 31: 3-18.

Endnotes

1 The adoption of cloud has been rapid. Between 2009-2017 cloud expenditures grew 4.5 times faster than traditional IT investment expenditure (Lesser 2017) representing 25% of firms’ IT budgets (Eurostat 2018, Deloitte 2017).

2 Uber, NetFlix, and Airbnb are often used as examples of the type of business model that have been made possible from cloud computing.

3 This term reflects whether larger establishments tend to be more distant from the headquarters (a positive covariance between distance and establishment size), or whether closer establishments are larger (a negative covariance).

4 While this is the case for most cloud services, an exception is email services, which can be accessed with basic ADSL broadband.

a

A

Topics:  Industrial organisation Labour markets Productivity and Innovation

Tags:  cloud technology, ICT, firm growth

Research Economist, Harvard Business School, Laboratory for Innovation Science Harvard

Professor of Economics, University of Nottingham

Research economist, IFC, World Bank Group and External Affiliate of University of Nottingham's research centre on Globalisation and Economic Policy.

Events

CEPR Policy Research