Abstract
This paper assesses the impact of covid-19 pandemic, measured through work mobility reduction, and e-commerce growth on the labour market using data from Indonesian labour force surveys and e-commerce transaction values. The findings confirm that the pandemic adversely affects workers’ employment prospects, work hours, total earnings, and hourly earnings. E-commerce growth does not counteract the adverse impact of the pandemic as expected, but it plays a role as an employment buffer during the crisis, although it tends to suppress workers’ earnings. Our results imply that more efforts are needed to improve the productivity of workers involved in e-commerce.
Acknowledgments
With the usual disclaimer, we are grateful to Bank Indonesia’s Department of Statistics for helping us by providing the e-commerce transaction values data.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 Indonesian E-commerce Association (idEA) is a communication forum for Indonesian E-commerce industry players. For further details, see: https://idea.or.id
2 Statistics Indonesia (BPS) defines formal sector workers as consisting of employers assisted by permanent/paid workers, and employees, while informal sector workers as consisting of own-account workers, employers assisted by temporary/unpaid workers, casual employees in agriculture and non-agriculture, and family/unpaid workers. For further details, see: https://sirusa.bps.go.id/sirusa/index.php/variabel/8482 (definition of formal sector) and https://sirusa.bps.go.id/sirusa/index.php/variabel/8483 (definition of informal sector).
3 For further details of Google Mobility Report, see: https://www.google.com/covid19/mobility
4 This data is publicly unavailable.
5 According to Abadie et al. (Citation2017), three factors determine the correct clustering strategy for the standard error. First, the data should have enough variation across samples in each cluster. Second, there is enough variation in different units being observed. Finally, the treatment assignment across units has enough variation. In our case, we rely on the variation of the mobility index at the province levels. Nonetheless, across provinces, the variation in the mobility index is not substantial. Because all provinces faced a similar mobility restriction during the COVID-19 pandemic, thus if we clustered at the province levels, it would reduce the variation in our dataset. Moreover, we only have 34 provinces, and henceforth, we have a problem with a small degree of freedom. Therefore, we use the robust standard error to overcome the potential heteroscedasticity in our data due to these issues.
6 During the normal situation, workers have a choice whether to work in the formal sector, to work in the informal sector, or to be unemployed. In such a situation, the choice of where to work is endogenous and hence the appropriate estimation method is the Multinomial Probit. However, the period of analysis in this paper is during a crisis situation, where the formal sector was shrinking and hence workers did not really have a choice where to work. Therefore, we estimate the model for the formal and the informal sectors separately using the Probit method.
7 The share of the top-4 marketplace is almost 80% of the total market capitalisation of e-commerce in Indonesia.
8 These indicators were calculated from the SAKERNAS microdata. There are some slight differences from the official figures published by BPS.
9 Training for the unemployed represents the Prakerja (Pre-employment) Programme. This is part of the National Economic Recovery (PEN) Programme launched by the Government of Indonesia to counter the adverse effect of the pandemic on the economy (https://pen.kemenkeu.go.id/in/post/mengapa-program-pen). The PEN Programme includes many specific programmes in six clusters: health, social protection, business incentives, support to micro, small, and medium enterprises (MSME), corporate financing, and regional governments and sectoral ministries. Information on participation in the Prakerja Programme is available in the SAKERNAS data. For the PEN-MSME programme, however, we did not find data that is suitable to be merged with the data analysed in this study as most of the benefits of this programme were distributed through banks.
10 We use some bounding assumptions on Rmax and develop a set of bounds for δ. According to Oster (Citation2019), the appropriate upper bound for δ is 1, which indicates that the unobservables are equally important as the observables. We define some bounds for δ: one side of the bound is Rmax = R, δ = 0 and the other bound is Rmax = 1.3R, δ = 1.
Additional information
Funding
Notes on contributors
Masagus M. Ridhwan
Masagus M. Ridhwan, Ph.D is a deputy director at the Bank Indonesia Institute, Bank Indonesia. He holds a doctorate in economics from Vrije Universiteit Amsterdam in the Netherlands.
Asep Suryahadi
Asep Suryahadi, Ph.D is a senior research fellow at the SMERU Research Institute, Indonesia. He holds a doctorate in economics from the Australian National University in Australia.
Jahen F. Rezki
Jahen F. Rezki, PhD is a lecturer at the University of Indonesia and the Vice Director for Research at the Institute for Economic and Social Research (LPEM-FEB UI). He holds a doctorate in economics from the University of York, United Kingdom.
Dinda Thalia Andariesta
Dinda Thalia Andariesta, MSM is a researcher at the Bank Indonesia Institute, Bank Indonesia. She holds a master degree in management science from Institut Teknologi Bandung, Indonesia.