Impact of TBTs on Trade in ICT Goods: Differentiating by Regulatory Objectives

Globalisation and digitalisation are intertwined. On the one hand, the globalisation process in recent decades has stimulated trade in the information and communications technology (ICT) sector (known as the digital sector). On the other hand, the digitalisation created by the ICT sector has reduced trade costs, thus stimulating trade and the globalisation process (Ahmad et al., 2011; Mattes et al., 2012; Yushkova, 2014; Nath & Liu, 2017; Xing, 2018; Ozcan, 2018). According to World Bank data,

https://data.worldbank.org/indicator/TM.VAL.ICTG.ZS.UN

As Figure 1 shows, the total import of ICT goods peaked in 2013 at USD 1.74 trillion. That indicates the importance of the trade in digital goods in the global economy. While tariffs imposed on ICT goods have been substantially reduced over the past two decades, non-tariff measures (NTMs) covering product regulations and standards imposed on goods have proliferated. Technical barriers to trade (TBTs) are among the most common NTMs imposed on ICT goods. Therefore, this paper analyses the impact of TBTs on imports of ICT goods over the period of 1996–2018.

Figure 1.

With the establishment of the World Trade Organization (WTO) in 1995, tariffs and traditional barriers to trade have been substantially reduced. In December 1995, members of the WTO concluded the Information Technology Agreement, which envisaged the elimination of tariffs on ICT goods. In December 2015, 50 members of this agreement resolved to extend the agreement to more than 200 products. As the right-hand side (rhs) axis of Figure 1 indicates, the simple average tariffs (in percentages) imposed on ICT goods have been gradually decreasing since 1997, while the import-weighted tariffs have been much lower than the simple average tariffs. This indicates that import flows are directed at tariff lines with lower tariffs in this sector. However, the use of NTMs has been on the rise. In fact, the average number of TBTs imposed on the import of ICT goods increased by 43 times, from about 0.3 at the start of the period to 12.3 in 2018.

This highlights the importance of the regulatory NTMs on the trade pattern of ICT goods, which is the focus of this study. Many studies in the literature have shown the heterogeneity of regulatory NTMs and their diverse implications for the trade in goods (Kee et al., 2009; Beghin et al., 2015; Ghodsi et al., 2017; Bratt, 2017; Niu et al., 2018). TBTs imposed by a large, growing economy, such as China, may stimulate imports to facilitate the organisation of supply chains and imports of intermediate materials that are used in other stages of production domestically. By contrast, certain other regulatory NTMs, such as sanitary and phytosanitary (SPS) measures imposed on the import of consumer goods, may significantly hamper trade (Disdier et al., 2008). ICT goods are used as capital, intermediate inputs and consumer goods. Therefore, the implications of NTMs for trade in ICT goods may be very far-reaching. In other words, the impact of regulations and standards embedded within TBTs on the use of goods, demand and price may be very heterogeneous.

The statistics department of the United Nations Conference on Trade and Development (UNCTAD) considers 93 goods at the six-digit level of the Harmonised System (HS) 2012 version, placing them in five broad categories.

The classification of ICT goods by UNCTADstat can be found here: https://unctadstat.unctad.org/en/Classifications/DimHS2012Products_Ict_Hierarchy.pdf

These goods correspond to 77 product codes in the HS 1996 version, for which data exist for a longer time span (since 1996). These goods and the related categories defined by UNCTAD are presented in Table A1 in the Appendix.

TBTs are imposed to regulate the market for legitimate reasons that are in line with all WTO agreements, and specifically with the TBT agreement of the WTO. According to the WTO’s Integrated Trade Intelligence Portal (I-TIP) data, 1,351 TBT notifications were imposed on these 77 goods between 1996 and 2018. According to the descriptions of these notifications, the NTMs were imposed in pursuit of 30 different regulatory objectives (suggested by their key words). Figure 2 presents the number of TBT notifications in force in 2018 for each of the 30 objectives, on the basis of the key words cited. Some 95% of the notifications provided at least one key word; for the remaining 5% of the TBTs imposed on ICT goods, no key word was specified (‘not specified’ ranks tenth among all the key words cited in TBT notifications to the WTO—see Figure 2). About 46% of the TBTs imposed on ICT goods had only one key word (including those with ‘not specified’); thus, 54% of TBTs provided at least two key words (and up to eight).

Figure 2.

The most common key word used in the description of a TBT imposed on ICT goods and notified to the WTO was ‘safety,’ with 422 notifications. The second most important category, with 383 TBTs imposed on ICT goods, had the key word ‘telecommunication/radiocommunication’ (which is uncommon for other types of goods). ‘Protection of human health or safety’ was the third most common key word, mentioned in 379 TBT notifications. ‘Prevention of deceptive practices & consumer protection’, ‘protection of the environment’, ‘human health’ and ‘quality requirements’ are the next most commonly cited key words, in 174, 173, 137 and 95 TBT notifications, respectively.

Each of these TBTs may target more than one product for regulation across a border. For instance, many TBTs are imposed on all products classified under chapter 85 of HS, which covers all ‘electrical machinery and equipment and parts thereof; sound recorders and reproducers, television image and sound recorders and reproducers, and parts and accessories of such articles’, which also includes non-ICT goods. Therefore, these TBTs may be very effective in controlling imports across the whole sector, if complying with them becomes too costly. Compliance costs may increase simply because of the bureaucracy or ‘red tape’: extra staff may be required to handle it. However, the costs may decline thanks to digitalisation and the automation of many customs offices. To reduce costs, some countries may regulate the market using ‘harmonisation’, ‘trade facilitation’ and ‘conformity assessment’ TBTs, with the relevant key words cited in, respectively, 43, 35 and 18 TBT notifications that target ICT goods. When the cost of compliance across the border increases due to the burden of the numerous regulations, it becomes an example of red tape hindering trade.

It is important to note that previous studies, which are referenced both above and below, have not distinguished between different types of TBTs. In fact, many studies have grouped all TBTs, or even all NTMs, under a single variable. The major contribution of this paper is in differentiating between various types of TBTs based on their objectives. However, it should be noted that some TBTs have a general objective and apply universally to all goods. Consequently, when previous research analysed specific TBTs targeting a particular product, these general TBTs were also included in their analysis. This is evident in the NTM data notified to the WTO, where we observe that some broadly targeted TBTs impact ICT goods. Key words such as ‘pesticides’, ‘protection of animal and plant life or health’ and ‘nutrition information’ appear in TBTs aimed at a wide range of products, or even universally. This demonstrates how many objectives cited within TBT notifications target all products. Prior studies have categorised these TBTs as affecting ICT goods, even though in practice, they should not specifically target such goods. The regulations behind the borders ensure that these general TBTs are complied with for all regulated products. Therefore, this study makes a significant contribution to the literature by analysing whether both related and unrelated regulations impact the trade value and volume of products. It delves deeper into the objectives and key words of TBTs notified to the WTO. One major hypothesis of this study is that objectives related to ICT goods, such as ‘safety’ or ‘telecommunication/radiocommunication’, should strongly affect the trade of these goods. In contrast, unrelated objectives such as ‘pesticides,’ ‘protection of animal and plant life or health’ and ‘nutrition information’ should not affect the trade of these goods, as they might be misreported as TBTs targeting these goods.

Therefore, this paper analyses the impact of different types of regulatory TBTs on the import of ICT goods during the period of 1996–2018. More precisely, the key words mentioned in the regulations notified to the WTO will be used to provide greater insight into what types of TBT have what kind of impact on imports of ICT goods. For instance, while ‘safety’ was the most frequently cited key word in TBTs imposed on ICT goods, the results show that these TBTs did not have any significant impact on the value or the volume of imports across many econometric specifications. However, TBTs that pursued a ‘conformity assessment’ played a stimulative and facilitating role in terms of the value of imports of ICT goods globally. These effects will be differentiated on the basis of the categories determined by UNCTAD statistics in July 2019. Both the value of trade and its volume will be analysed to provide greater insight into the impact of TBTs on import prices, value and volume. Since TBTs are regulations that are imposed unilaterally on all exporters, unilateral import values of 77 HS six-digit ICT goods during the period of 1996–2018 are used. However, because of mutual recognition and harmonisation of standards, regulations and NTMs across the single market of the European Union (EU), unilateral imports are calculated in such a way that intra-EU trade and intra-EU trade policy measures are excluded from the analysis. The organisation of the rest of the paper is as follows. The next section describes the methodology and the data used in the analysis, section 3 provides the results of the analysis and section 4 offers a conclusion.

As mentioned above, TBTs are imposed unilaterally on imports of goods from all exporting countries. Therefore, to measure the impact of TBTs on imported goods, one can choose a unilateral setting, which is also used in earlier studies, such as Kee et al. (2009). However, there are zero trade flows for some imported products to certain countries, which may cause sample selection bias in the estimation of the impact of NTMs, especially if the NTM imposed on a product prohibits trade completely. To control for zero trade flows in the import data, the Pseudo Poisson Maximum Likelihood (PPML) is used that is implemented in Stata by Santos Silva and Tenreyro (2006) (see also Head & Ries, 2008; Head & Mayer, 2014; Yotov et al., 2016; Larch et al., 2019), further developed by Correia et al. (2019a,b) to run multidimensional fixed effects. The equation for the estimation of unilateral trade is as follows: (1) miht=exp[α0+α1 arc Tiht+α2TBTiht+ωit+ωih]+εiht {m_{iht}} = {exp ^{\left[ {{\alpha _0} + {\alpha _1}\;{\rm{arc}}\;{T_{iht}} + {\alpha _2}{TBT_{iht}} + {\omega _{it}} + {\omega _{ih}}} \right]}} + {\varepsilon _{iht}} where m_iht is total import values m^v_iht (or import volumes m^q_iht) of product h to country i in year t; arcTBT_iht is the hyperbolic sine transformation (Bellemare and Wichman, 2020) of simple average tariffs (in per cent) levied by country i on imports of product h in year t; TBT_iht is the stock number of TBTs imposed by country i on the imports of products h and in force in year t, which also includes TBTs imposed in earlier years that are not withdrawn;

It is important to note that according to the WTO I-TIP data, no TBT is reported as having been withdrawn. Many earlier studies in the literature have relied on a simple dummy variable to indicate the existence of NTMs, which overlooks details about their extensiveness and stringency. However, it could be argued that some TBTs may provide information about the reduction in barriers. Whereas previous studies treat the existence of any such information merely as a binary variable, this study quantifies them. This approach allows for a more nuanced analysis of their final impact on trade flows, illuminating whether their objective is to reduce barriers or to increase the stringency of regulations.

It is widely discussed in the literature that, due to reverse causality of TBTs and the dependent variable, there is a potential endogeneity bias in the estimation of TBTs, as TBTs may be correlated with the error term E(TBT_iht ɛ_iht)≠0. Such an endogeneity bias is controlled for using the instrumental variable approach in the literature (Kee et al., 2009; Beghin et al., 2015; Bratt, 2017; Niu et al., 2018; Ghodsi, 2020). Moreover, other sources of endogeneity bias could be measurement error and the omitted variable bias. While the latter is controlled for using the appropriate fixed effects in equation (1), the former could also be controlled for using the instrumental approach. Following the seminal work of Kee et al. (2009), past growth in imports Δm_iht-1=arcm_iht-1-arcm_iht-2, and total exports arcx_iht of product h from country i in year t are two exogenous variables that may affect the imposition of TBTs, but do not correlate with the dependent variable in year t. Moreover, Kee et al. (2009) also use the GDP-weighted average TBT of the five geographically closest countries to importing country i as the third exogenous instrument for TBT, to control for the measurement error in the calculation of TBT. However, the five closest countries may be quite arbitrary: in a globalised world, distant countries may impose new TBTs, following other distant trading partners, to harmonise their standards and regulation settings. For instance, China is very distant from the EU and the US, but their global value chains have been intensively intertwined in the past two decades. This means that regulations and standards in Chinese manufacturing sectors may have converged with those in other countries, so that foreign intermediate inputs can be imported and used in Chinese value chains. Therefore, as the third exogenous instrument, the average stock number of TBTs imposed by countries other than the importing country i under estimation TBT^w_i‘ht is used, which is weighted by the per capita GDP of countries. The reason for choosing per capita GDP, rather than GDP, is that the level of development may be a better indicator of a qualitative and regulatory framework for the implementation of TBTs than simply the size of the economy.

The GDP weights instead of per capita GDP weights are also used in separate specifications, the results of which are available on request.

After estimating equation (2) and after retrieving the fitted value of TBT^iht {\widehat {TBT}_{iht}} , this variable is included in the second-stage estimation, as follows: (3) miht=exp[γ0+γ1 arc Tiht+γ2TBT^iht+ωit+ωih]+ϑiht {m_{iht}} = {\exp ^{\left[ {{\gamma _0} + {\gamma _1}\;{\rm{arc}}\;{T_{iht}} + {\gamma _2}{{\widehat {TBT}}_{iht}} + {\omega _{it}} + {\omega _{ih}}} \right]}} + {\vartheta _{iht}} where the fitted value of TBT is not correlated with the error term E(TBT^iht ϑiht)=0 E\left( {{{\widehat {TBT}}_{iht}}\;{\vartheta _{iht}}} \right) = 0 . It is important to note that because multiple fixed effects are used in the second stage, it is not possible to use a bootstrapping estimation. However, the error term is clustered by importer-product ih to control for the heteroskedasticity and autocorrelation in the error term.

The benchmark estimation of equation (3) is done on the sample of all 77 ICT goods in the HS 1996 version, which gives the average global impact of TBTs on the value and volume of imports across all these ICT goods. However, as noted above and discussed in the literature, different products may be differently affected by TBTs. Therefore, to observe the impact of TBTs across the five product categories defined by UNCTAD, the fitted TBT in equation (3) interacted with five product category dummies to obtain five point estimates and confidence intervals for a TBT’s impact on the imports of the five product categories. A list of these categories and the ICT goods classified is provided in Table A1 in the Appendix. One can estimate the impact of TBTs on different product categories using separate samples of estimations. However, estimating all the products in the same sample would yield results that are more robust. The main reason for that is that all the available information on all products is used in one single estimation, which not only provides a greater degree of freedom for the calculation of statistics, but also uses more information for the estimation of parameters. Moreover, this methodology gives standard deviation and point estimates of each coefficient of TBT for each product category conditional on the coefficients of TBT for other product categories in the same estimation.

Furthermore, the TBT variable used in the benchmark specification includes all types of TBT notified to the WTO that target ICT goods. Thirty separate variables are generated to take account of the 30 different key words cited in the TBTs imposed on these goods. These 30 variables are estimated in separate regressions to infer their impact on trade flows in ICT goods and their product categories.

Table 1 presents the second-stage instrumental variable estimation results using PPML for imports of all ICT goods during the period of 1996–2018. Estimation results using the stock number of TBTs in levels are presented in the left-hand panel (the two left-hand columns) of Table 1.

The number of TBTs imposed by an importing country that was a member of the WTO in 2018 has a statistically positive effect on the value of its imports of ICT goods m^v_iht at the 5% level of significance. The coefficient suggests that an additional TBT imposed each year on ICT goods increases the value of imports by 2.2%. However, the elasticity of import values with respect to the number of TBTs is statistically insignificant; this result is presented in the right-hand panel of Table 1, which uses hyperbolic sine transformation of the number of TBTs and is defined as arc TBT^iht {\rm{arc}}\;{\widehat {TBT}_{iht}} . The results using an instrumental variable (IV) approach and the fitted values of TBTs in the second stage are very similar to those obtained from the estimations using the actual values of TBTs in the models presented in Table A2 in the Appendix. The difference between the two results suggests that the bias that is corrected by the exogenous instruments yields results that underestimate the impact of endogenous TBTs.

PPML first-stage estimation results on TBTs are presented in Table A3. From the three exogeneous instruments used, only the average global number of TBTs weighted by the per capita GDP of each country that imposes them is statistically significant. Past growth in import values (volumes) Δm^v_iht-1 (Δm^q_iht-1) or export values (volumes) arcx^v_iht (arcx^q_iht) are not correlated with the endogenous TBT_iht, and only the TBTs that are imposed globally are significantly related to the number of TBTs that a country imposes on each ICT good. This result comes despite the inclusion of country–product fixed effects that control for any regulatory measure and characteristics that are persistent in each country that demands or supplies one of these ICT goods. As can be observed from the results of the first-stage IV regressions (Table A3), the number of observations decreased to 63,509 (from 181,971 in the non-IV regression presented in Table A2). The reason is that dropped observations are either singletons or are separated by a fixed effect, as Correia (2015) also explains. This means that those dropped observations for TBTs will be controlled for by the fixed effects if they are included in the estimations as explanatory variables. However, the inclusion of these fixed effects gave statistically insignificant coefficients for other exogenous instruments in the first stage. Therefore, when the fitted value of a TBT is used in the second stage, the actual values of TBTs could replace the missing observations from the fitted values of the first-stage estimation, simply because those actual values of TBTs are controlled for by the fixed effects in the second-stage estimation.

Another interesting result appears to be the insignificant impact of tariffs on the value or volume of imports in any specification. As explained above, tariffs were reduced to a low level (or zero) by many WTO members after the Information Technology Agreement was concluded in December 1995. The simple average global tariffs levied on these products declined from 7.4% in 1997 to 2.1% in 2018. In December 2015, 50 members of the WTO that were committed to the Information Technology Agreement implemented a reduction in tariffs to zero on the import of further ICT goods.

Even more interestingly, the results of the first-stage estimation (Table A2) illustrate that a 1% reduction in tariffs is correlated in a statistically significant way with a 0.032% increase in the number of TBTs imposed, which serves to indicate the substitutability of TBTs for tariffs. The reason is that, while tariffs fell by 58% from 1997 to 2018, the number of TBTs grew by 48 times over the same period. Meanwhile, the impact is estimated to be 0.032%, controlling for fixed effects and other variables in the estimation. The substitutability of NTMs for other types of trade policy measures, such as traditional tariffs or other NTMs, has been widely studied in the literature (Rosendorff, 1996; Yu, 2000; Moore & Zanardi, 2011; Aisbett & Pearson, 2012; Orefice, 2017).

In the modern digitalised world, ICT goods are the engine of the global economy. According to the data provided by UNCTAD to the World Bank, trade in ICT goods was about 12.9% of total imports of goods in 2019, which indicates their importance in the global economy. While tariffs levied on ICT goods have been eliminated or significantly reduced since the conclusion of the WTO’s Information Technology Agreement in December 1995, NTMs have become more relevant and important for trade in these goods. As the most important trade policy measure imposed on imports of ICT goods, TBTs form the focus of this study. This paper analyses how the different types of TBT imposed on imports of ICT goods globally affected the value and volume of imports of these goods at the six-digit level of the Harmonised System during the period of 1996–2018. Key words cited in the TBTs that are notified to the WTO indicate the aims behind the imposition of the TBTs. Therefore, as a major contribution to the literature, this study delves deeper into the impact of 30 types of TBTs, using the key words cited in them. The methodology used in the paper controls for zero trade flows by using the PPML technique and for endogeneity bias by using the exogenous instrumental variable approach. Moreover, the paper provides more detailed analysis of the impact of TBTs across five ICT product categories that were defined by UNCTAD in July 2019.

The results indicate that TBTs have a generally strong positive impact on the value of imports of ICT goods. Moreover, this positive impact is stronger and statistically more significant for the ‘communication equipment’ category of ICT goods. Furthermore, TBTs that cite the key words ‘conformity assessment’, ‘crime protection’, ‘electromagnetic compatibility’, ‘harmonisation’, ‘national security requirements’, ‘other’ and ‘protection of animal or plant life or health’ lead to a statistically significant stimulation of the value of imports of ICT goods. While TBTs that cite ‘nutrition information’ which are irrelevant for ICT trade significantly reduce the value of imports, other TBTs do not have any statistically significant impact on the import values of ICT goods. Most importantly, TBTs that cite ‘quality requirements’ stimulate the volume of imports of ICT goods in a statistically significant way. The imposition of this important TBT—one that improves the quality of traded goods—could be an important way of stimulating trade volume, while the imposition of other TBTs mentioned above can increase export revenues.

Overall, the results presented here could give policy makers a better insight into how to adjust trade in ICT goods by implementing smart policies to regulate the market, when the market cannot automatically adjust and reduce the negative externalities. For instance, TBTs that cite ‘conformity assessment’ stimulate the value of imports of all ICT goods in all five product categories. TBTs in pursuit of ‘harmonisation’ or ‘plant health’ may also improve trade values across many categories. However, some other TBTs—such as those that pursue ‘labelling’ or ‘packaging’—may result in a statistically significant reduction in the volume of imports, as the costs associated with compliance increase at the plant level where the ICT good is produced or packed, but do not necessarily induce positive externalities to stimulate demand or export revenues.