Business dynamics statistics of High Tech industries

• Published date: 01 January 2020
• DOI: http://doi.org/10.1111/jems.12334
• Date: 01 January 2020
• Author: Nathan Goldschlag,Javier Miranda

Published 2019. This article is a U.S. Government work and is in the public domain in the USA

J Econ Manage Strat. 2020;29:3–30. wileyonlinelibrary.com/journal/jems

|

3

Received: 20 March 2018

|

Revised: 5 August 2019

|

Accepted: 24 October 2019

DOI: 10.1111/jems.12334

ORIGINAL ARTICLE

Business dynamics statistics of High Tech industries

Nathan Goldschlag

1

|

Javier Miranda

2

1

Center for Economic Studies, U.S.

Census Bureau, Suitland, Maryland

2

Economy‐Wide Statistics Division, U.S.

Census Bureau, Suitland, Maryland

Correspondence

Nathan Goldschlag, 4600 Silver Hill Road,

Washington DC 20233.

Email: nathan.goldschlag@census.gov

Abstract

Modern market economies are characterized by the reallocation of resources

from less productive, less valuable activities to more productive, more valuable

ones. Businesses in the High Tech sector play a particularly important role in

this reallocation by introducing new products and services that impact the

entire economy. In this paper we describe an extension to the Census Bureau’s

Business Dynamics Statistics that tracks job creation, job destruction, startups,

and exits by firm and establishment characteristics, including sector, firm age,

and firm size in the High Tech sector. We preview the resulting statistics,

showing the structural shifts in the High Tech sector over the past 30 years,

including the surge of entry and young firm activity in the 1990s that reversed

abruptly in the early‐2000s.

KEYWORDS

employment flows, firm dynamics, High Tech, startups

1

|

INTRODUCTION

Modern market economies are characterized by the reallocation of resources from less productive, less valuable

activities to more productive, more valuable ones. Businesses that innovate grow and replace those that do not

(Schumpeter, 1942). The innovations that underlie this reallocation process are ultimately responsible for increased

productivity growth and increasing standards of living. Businesses in the High Tech sector play a particularly important

role in this regard, generating new products and services that fuel this reallocation. Not surprisingly, developing a

deeper understanding of these businesses, their inception, early life cycle dynamics, and constraints to growth are the

focus of intense attention (Acemoglu, Akcigit, Alp, Bloom, & Kerr, 2018; Decker, Haltiwanger, Jarmin, & Miranda,

2016a, 2016b; Stoneman & Battisti, 2010). However, our understanding of the dynamics of innovative firms remains

limited. One reason for this is that, until recently, longitudinal data sets of establishments and firms covering the US

economy did not exist. The development and introduction of the US Census Bureau’s Longitudinal Business Database

(LBD) changed that. The LBD tracks all establishments and firms in the private nonfarm economy with paid employees

starting in 1976 and is updated annually.

1

The LBD, however, is a confidential microdata set only accessible to Census

Bureau employees and researchers with an approved “need to know”and therefore access to the data is necessarily

limited. In this paper, we describe the use of the confidential microdata in the LBD to preview a series of public use data

products that capture the dynamics of firms and establishments in High Tech industries. This effort is consistent with

the goals of the Census Bureau to produce up‐to‐date economic and social measures to advance informed decision‐

making in business and society through the development of public use tabulations while also protecting the

confidentiality of the underlying data (Jarmin, Louis, & Miranda, 2014).

The existing Business Dynamics Statistics (BDS) tabulations, built upon the LBD, contain information on the

number of firms and establishments, firm openings and closings, and job creation and destruction by relevant firm

---------------------------------------------------------------------------------------------------------

characteristics, including firm age, size, state, metropolitan statistical area, and sector. To examine the dynamics of the

High Tech sector we must first identify High Tech businesses in the LBD, which is a nontrivial exercise. In this paper,

we review the existing methodologies used to identify High Tech industries, paying special attention to their application

and use in BDS data products. These methodologies rely on the identification of inputs and outputs of the innovation

process across detailed industries, such as R&D investments, proportion of STEM workers, or the technological content

of the products.

2

A critical element in the evaluation of these methodologies is the ease with which they can be updated

and replicated. Relevant factors in this regard include the timelines and coverage of the underlying data that the High

Tech classification methodology employs. Some classification methods would require third‐party data over which the

Census Bureau has little control. An equally important consideration when choosing a methodology for identifying

High Tech industries is that it produces stable indicators. A methodology that identifies a volatile set of industries is of

little use for longitudinal analyses. Finally, an important requirement is that the methodology be flexible enough to

identify High Tech industries across the entire economy since it is increasingly the case that industries outside of

manufacturing can reasonably be considered High Tech.

We find considerable differences in the set of High Tech industries identified by different methodologies. Output‐

and R&D‐based methods almost exclusively identify manufacturing industries due to data limitations. Ad hoc

extensions to the methodology attempt to compensate for this inherent limitation but yield a very broad listing of

industries outside of manufacturing. Moreover, output‐based methods with such extensions are relatively difficult to

replicate and maintain. Input‐based methods, on the other hand, are relatively easy to maintain and replicate given the

survey collections they rely upon. Amongst these methodologies, the one based on occupation data (what workers do) is

the most promising. The availability of occupation data across all industries makes it possible to identify a list of High

Tech industries that spans all sectors, including information and services industries, using a consistent framework.

Despite the significant difference between the methods, there is nevertheless significant overlap in the set of High Tech

industries that emerge, particularly in the manufacturing sector. Comparing measures of business dynamism derived

from alternative High Tech definitions shows similar overall patterns in the manufacturing component of High Tech.

Consistent with previous findings in the literature, we find overall declines in business dynamism in the High Tech

sector after 2000. We find High Tech intensity is directly related to the magnitude of the decline. More intensely High

Tech industries exhibit declines in dynamism that is 20% greater than those experienced by the less intensely High Tech

industries.

On the basis of concerns about coverage, stability, interpretation, and replicability, we define High Tech industries

based on the union of industries with the highest proportion of STEM employment in 2005, 2012, and 2014.

3

Using this

classification, we compute the Business Dynamics Statistics of High Tech industries (BDS‐HT). These data capture

flows of firms, establishments, and employment for High Tech and non‐High Tech industries. High Tech industries

make up a relatively small portion of the economy, about 4% of firms and 6% of employment in 2014. One of the most

pronounced features of the BDS‐HT data is the significant increase in entry and young firm activity in the 1990s, which

reverses sharply after 2000.

The rest of the paper proceeds as follows. Section 2 describes different methods for identifying High Tech industries.

We evaluate their reproducibility, ease of maintenance, and overall strengths and weaknesses. Section 3 compares the

list of High Tech industries that emerge across methodologies. We compare selected measures of business dynamism

economy wide and for the manufacturing sector. Section 4 presents a preview of the BDS‐HT tables that result from our

preferred High Tech classification. Section 5 concludes.

2

|

IDENTIFYING HIGH TECH INDUSTRIES

An interagency workshop held by US statistical agencies in 2004 identified a set of important factors that contribute to

the concept of High Tech. These include disproportionate utilization of STEM workers, disproportionately high

employment of R&D workers and capital, the production of High Tech products, and the use of High Tech production

methods, including the use of High Tech capital goods and services.

4

Operationalizing these definitions involves

classifying economic activity based on either the use of High Tech inputs or the production of High Tech outputs. Each

approach has advantages and disadvantages with respect to their use in measuring the dynamics of High Tech

industries. These differences are primarily related to the availability of data and the robustness of criteria and

thresholds used.

5

Table 1 summarizes the different methods of classification and their associated literature. In the

following sections we detail each method as well as their pros and cons when used in the development of BDS‐HT data.

4

|

GOLDSCHLAG AND MIRANDA