The different effects of hardware and software on production interdependence in manufacturing

Abstract

Make or buy decisions are often decisions between investing in capital for production or building supply chain relationships upstream. These are strategic decisions about production interdependence: the degree to which materials and services are provided internally versus purchased externally. Information technology (IT) has increased productivity with hardware and software such as robotics and flexible manufacturing systems. IT has also improved information sharing and coordination along the supply chain by integrating business processes with software. Within manufacturing, we examine whether hardware and software are related to choices of make versus buy differently. From a transaction cost perspective this can be due to differential impacts on reducing internal production costs versus external coordination costs. We find that in U.S. manufacturing industries from 1998 to 2016, hardware favors internal production suggesting hardware reduces costs of internal provision more; software increased purchases from upstream suppliers suggesting software reduces costs of external provision more. This result shows that hardware and software complementarity has limits in that each has distinct productivity targets, and that empirically the decision of make versus buy is manifested by investments in these different types of IT capital. That is, when facing strategic decisions about make versus buy in manufacturing, hardware and software have opposite impacts.

Introduction

One of the most fundamental decisions in manufacturing is how much of the inputs to make and how much to buy. This determines firm's production interdependence: the degree to which materials and services used to produce final goods are provided internally versus purchased externally. The result leads to a level of production interdependence, which is an important dimension of the structure of supply chains and the economy in aggregate: it defines the boundary of firms and of industries, it determines bargaining power and competition, and it dictates the depth (i.e., number of tiers) of supply chains. It also determines the distribution of value added across the supply chain. Thus, make versus buy decisions are strategic.

As part of these decisions industries have invested heavily in information technology (IT). According to the U.S. Bureau of Labor Statistics (BLS), IT capital stock by the manufacturing sector climbed steadily from $73.04 billion in 1987, reaching $305.09 billion in 2018 (in 2012 dollars from capital tables released March 24, 2020.). This IT capital, which from an acquisition perspective can be separated into hardware and software where the former accounts for over 50% in manufacturing, reflects how industries have developed best practices to engineer operations and business processes.

IT capital has also changed how production is organized. Previous studies have found that IT has an impact on various measures of organizational structure. For example, IT leads to smaller firm size [11], and IT is related to decreases in vertical integration and increases in diversification [38]. These studies use IT capital as an aggregate. In the context of manufacturing, we presume that IT hardware and software may have different impacts on how production is organized, making different IT investments strategic in a way that is distinct from making IT investments with the goal of straightforward increases in productivity.

Every system is a combination of hardware and software. However, some are predominantly hardware and others are predominantly software. On one hand, IT implementations that are predominantly hardware in manufacturing like industrial process instruments, robotics, sensors, radio-frequency identification (RFID), computer numerical control (CNC), and flexible manufacturing systems (FMS), together with more general computers and communications, have yielded increased productivity on the shop floor. This hardware allows flexibility in production and scalability at low cost; links functional unit information systems to shop floor manufacturing technology; increases productivity through automation of routine tasks; improves designer productivity through computer-aided design; and enables greater process consistency and reliability, thereby improving product quality using CNC and FMS [66].¹ Consequently, as hardware in manufacturing favors internal production, we expect that more hardware is associated with more internal provision, which translates into less interdependence.

On the other hand, IT implementations that are predominantly software, such as interorganizational systems (IOS) that enable strategies like just-in-time (JIT) production, have supported industries becoming increasingly integrated with their suppliers' business processes [23]. The range of software includes tools like data analytics and machine learning incorporated in organization procedures and management methods. These are embedded in software-supported processes such as total quality management, JIT production, materials requirements planning (MRP), and supply chain management (SCM) – all of which are used to connect customers and suppliers [12]. Consequently, industries in a manufacturing supply chain may have become more interdependent as software investments have grown.

To help guide IT investments in support of make versus buy decisions and to better understand the impact of hardware and software subcomponents of IT on production interdependence, we examine how an industry's IT capital impacts its production dependence measured as direct backward linkage (DBL) with upstream suppliers. We disaggregate IT capital into hardware and software, and we examine their differential effects on DBL. Our definition of hardware and software is developed by the BLS. Hardware includes generic IT hardware such as computers and communication equipment, as well as manufacturing technology consisting of computer-aided industrial process instruments and related machines in manufacturing, such as those described above; the former accounts for roughly one third and the latter two thirds of hardware investment. Software includes pre-packaged, custom, and own-account software that includes application software such as ERP, IOS, SCM, etc., also described above. Some system software is firmware and, as such, is bundled with hardware.

We use production theory and develop an estimation model where we separate the direct effects of IT capital on DBL and the indirect effects of different IT capital subcomponents on DBL through intermediate inputs. To capture the indirect effects we allow the output elasticity of intermediate inputs to depend on different IT capital subcomponents, yielding a set of specifications. We then estimate that model with U.S. manufacturing industry-level data over 19 years: 1998 to 2016. As Aigner and Chu [2] explain, any firm production function can be obtained from optimal parameter values at the industry-level, and estimation of an industry-level production function represents that of an average firm in the industry. This is certainly the case when firms follow industry best-practice.

First, we find that IT capital in aggregate has a negative direct effect on DBL across our different specifications. This is because IT capital expands output (make) relative to intermediate inputs (buy), corresponding to lower interdependence.

Next, disaggregating IT capital into hardware and software subcomponents, we develop a specification that isolates the indirect effects of hardware and of software through intermediate inputs. We find that the indirect effect of hardware reduces DBL, and that, in contrast, the indirect effect of software increases DBL. Moreover, the indirect effect of software is greater than the direct effect of aggregate IT capital. Thus, we find that hardware capital favors internal production and reduces interdependence, whereas software capital favors external purchasing and increases interdependence with supply chains upstream. This seemly surprising result shows that hardware and software complementarity where investments in hardware are viewed as necessitating investments in software and vice versa, has limits in that each has distinct productivity targets, and that empirically the decision of make versus buy is manifested by these different IT capital investments. In other words, when it comes to impacts on interdependence, hardware and software have opposite effects.

Further, we disaggregate hardware capital into separate measures of general computers and communications, and of manufacturing technology. We find that the indirect effects of both hardware measures relate to lower interdependence, indicating complementarity between different categories of hardware persists in their effects on interdependence, and the effects remain opposite to those of software.

The remaining sections are organized as follows: Section 2 explores literature related to our work; Section 3 develops our conceptual, theoretical, and estimation models; Section 4 presents our empirical estimation including a description of the data, variables, and econometric adjustments; Section 5 presents our estimation results; and Section 6 provides our conclusions.