Have the economic policy packages reduced import costs in Indonesia? A quantitative assessment of two sets of measures1 The World Bank The purpose of this note is to evaluate the outcome of two sets of actions, included in the first of Indonesia’s economic policy packages issued in September 2015 and supported by the Logistics DPL I, aiming to reduce the costs of importing: 1) “The Minister of Trade has eliminated Registered Importer and Producer Importer licenses for eight (8) categories of products, which accounted for 10.6% of non-oil imports in 2015, as evidenced through various Minister of Trade Regulations.” (action nr. 7 of Logistics DPL) 2) “BPOM has required the application of Import Declaration Letter for food and drugs to be submitted electronically, as evidenced through Head of BPOM Regulations 12 and 13 of 2015.” (action nr. 10 of Logistics DPL) Our estimates suggest that the elimination of IP/IT and the introduction of the compulsory use of e-BPOM have substantially reduced the costs of importing for the products to which they were applied. The effect is equivalent to a reduction of the import price by 6.7% and by 8.4% respectively. Rationalization of import requirements Documentary compliance for international trade in Indonesian takes considerably longer than in neighboring countries, which contributes to long overall clearance times. As per 2013, between 50 and 60 percent of the dwell time - the time a cargo spends within port limits - in Tanjung Priok port was due time spent from discharge to submission of customs declaration, which was mainly associated with the processing of import licenses not obtained prior to cargo arrival. Import licenses include importer identification number, product-specific import licenses, and shipment-specific import approvals. The Ministry of Trade is one the ministries most responsible for import requirements and in the context of the reform packages it has started this rationalization process. Specifically it has revised or repealed 16 regulations eliminating Registered Importer (IT) and/or Producer Importer (IP) licenses for 16 categories of restricted products spanning several sectors with the following 8 being the most important: horticulture products, tires, compact discs and related machinery, textile, wood products, certain types of finished products, color photocopiers, animal products. These licenses require various types of documents to be presented (from 4 different documents for cloves’ IT license to 11 for tires’ IP license). The revised or repealed regulations cover products that comprise about 10.6 percent of total non-oil and gas import value in 2015 and 19 percent of tariff lines at the 10 digit HS-code level. The immediate impact of this simplification should be a reduction in the time and cost of complying with import requirements for those products. Indeed initial interviews with private 1 This note is prepared by Massimiliano Calì with contributions from Agnesia Adhissa, Lamiaa Bennis, Laura Puzzello and Hazmi Shidqi. 1 sector firms suggest that the time to process import licenses decreased as a result of this simplification. Ultimately, these reforms are expected to reduce the overall time and cost associated with meeting regulatory requirements associated with trade. To the extent that the elasticity of imports with respect to trade costs is positive, this reduction in cost may also translate in some increase in imports. Expediting documentary submission Potential gains in pre-clearance time are constrained by the fact that until recently several agencies and ministries still require the submission of the hard copy of certain trade documents, which slows down the approval process. This contributes to the slowness in compliance with documentary requirements in Indonesia relative to more competitive countries in the region. Electronic processing of trade documentation has indeed been shown to yield significant reductions in processing time and even to increases in number of exporters and export volumes.2 The government has made progress on the electronic submission and processing in key border agencies. Ministries and agencies, including Customs, the Ministry of Trade and the Food and Drugs Agency (BPOM) have developed their own electronic systems to process declarations and import/export-related licenses. BPOM has recently issued two regulations that automate the process for applying for an Import Declaration Letter (SKI) for both food and drug products and inputs. The BPOM regulations cover food and drug products and materials that comprise 6 percent of total non-oil and gas import value in 2015 and that have typically higher processing times than average. The electronic processing of BPOM import documentation is expected to lead to a substantial reduction in the processing time and costs for BPOM related permits, which would also likely be reflected in improvement in the overall processing time and costs. Indeed the empirical evidence suggests that these time and cost reductions would also translate into increased imports and exports, leading to overall gains for the economy. 3 Indeed our interviews suggest that the rolling out of E-BPOM has significantly improved the processing time of BPOM’s permit.4 A quantitative assessment We examine more rigorously the trade impact of the IP/IT elimination and of the introduction of the compulsory electronic submission to BPOM, by employing a simple difference-in- difference analysis of Indonesian imports. In the case of the BPOM reforms for instance this 2 Carballo, J., G. Schaur and C. Volpe Martincus (2015). “The Border Labyrinth: Information Technologies and Trade in the Presence of Multiple Agencies”, Inter -American Development Bank, mimeo. 3 See for instance Djankov et al. (2010), “Trading on time” Rev iew of Economics and Statistics; Carballo, J., G. Schaur and C. Volpe Martincus (2015). “The Border Labyrinth: Information Technologies and Trade in the Presence of Multiple Agencies”, Inter -American Development Bank, mimeo. 4 According to the priority importers association before the introduction of the electronic submission, the application could take between 2 to 5 working days, whereas it takes 1 to 2 days with E-BPOM, depending on the time the application is submitted and subject to payment of BPOM’s service fee. 2 empirical strategy compares the changes in imports of products which are subject to BPOM permits (treated product group) before and after the introduction of the e-BPOM (i.e. November 2015) with the same changes for the other products (control product group). The group of treated products is essentially the entire universe of food and drugs products. As all of these products are subject to the policy change (i.e. the government has not selected certain sub-samples), it is less likely that issues of self-selection may arise that could bias the estimation. A similar strategy is also applied to the IP/IT reforms, with the difference being that these reforms had different timing across different product categories. In addition, the IP/IT treated products are potentially more subject to selection bias as these are sub-samples of larger product categories and their selection may be related to other factors (e.g. ability of producers to lobby the government) which may also directly influence imports. To the extent that such bias may arise, the estimation of the IP/IT effects would have to be treated with caution. We work with highly disaggregated product classification, i.e. HS-10 digit, which is the same the government use to classify imports. We match quarterly merchandise import data (from Customs) since 2008 with the IP/IT and e-BPOM policy variables. Each variable takes the value of 1 for all of the products affected by the relevant policy for each quarter from its introduction onwards and the value of zero otherwise. We run the analysis between first quarter 2008 and the second quarter 2016. We also control for all of the other non tariff measures applied to each product category exploiting the extensive data collection by ERIA updated by the World Bank.5 The formal estimation model is presented in more detail in Annex 1. Table 1 presents the main results of the empirical analysis. Both policies have a positive and highly significant effect on imports. That is true whether modelling their effect separately (columns 1 and 2) or at the same time (column 3). The elasticity of imports to these measures is large. Taking our preferred specification in column (3), our results suggest that the elimination of IP/IT has increased imports by 17.4% on average, and the compulsory electronic submission through e-BPOM has raised imports by 22%. 6 It is worth noting that the BPOM result is generally more robust than the IP/IT one, as the latter does not pass some of the robustness tests that we carry out. 7 These large effects suggest that both measures have reduced the cost of importing those products quite significantly. While we do not have data to test through what channels have these costs been reduced, our interviews suggest that the time savings and the greater certainty 5 See more details in Calì and Puzzello (2017). “The welfare impact of non tariff measures on Indonesia”, World Bank. 6 These elasticities are obtained by transforming the coefficients as follows: (e Coeff - 1). 7 We run placebo tests for the effects of both policy variables by randomly changing the quarter in which those policies were introduced. Specifically we anticipate the starting date of the policies by 1, 2, 3 and 4 years. If the coefficients we estimate in Table 1 are picking up genuine policy induced effects as opposed to just random trends, then the placebo coefficients IP/IT and BPOM should be not significant or at least significant smaller in magnitude than the coefficients in Table 1. That is the case for all of the placebo coefficients we estimate except when the start date of the IP/IT variable is anticipated by one year (results not shown here but available upon request). 3 in the processing time have been key to realize the reduction in trading costs. We use these estimates along with the elasticities of imports with respect to the price of imports to compute the estimated reduction in import costs induced by these two sets of measures (see Annex 2 for a formal derivation of this computation). Our calculations suggest that the elimination of IP/IT and the introduction of the compulsory use of e-BPOM have substantially reduced the costs of importing for the products to which they were applied. The effect is equivalent to a reduction of the import price by 6.7% and by 8.4% respectively. Table 1: The impact of trade facilitation reforms on imports (1) (2) (3) Imports Imports Imports 0.158*** 0.160*** IP/IT (0.047) (0.047) 0.190*** 0.198*** E-BPOM (0.063) (0.064) -0.195*** -0.197*** -0.196*** NTM (0.032) (0.032) (0.032) Observations 243,849 243,849 243,849 R-squared 0.032 0.032 0.032 Note. The dependent variable is: − ̂ (1 + ). Standard errors are clustered at the HS10 product level. *** Significant at 1 percent. All specifications include product and quarter-year fixed effects. 4 Annex 1: Methodology for the estimation of the IP/IT and e-BPOM effects on imports We follow a similar methodology as in Calì and Puzzello (2017) and estimate the following specification: = 1 + 2 + + (1 + ) + + + (1) where mpqy is the import value of product p (at HS-10 digit) in quarter q of year y, and are the two relevant policy variables as defined above; is an indicator variable that takes on 1 if at least an NTM applies on p in a given q-y; is the import elasticity of product p; is the ad-valorem tariff on good p in a given q-y; is a product dummy that allows to controlling for product-specific characteristics; is a quarter-year dummy that accounts for shocks specific to a given quarter-year and common across products; and is the error term. The coefficients of interest in equation (1) are β1 and β2, which capture the average impact of an NTM on Indonesian imports. The inclusion of product and quarter-year dummies partially control for factors that simultaneously affect imports and NTMs use, and it implies that is identified by changes in the use of NTMs within and across products over time. The effect of tariffs on imports in equation (1) depends exclusively on the import elasticities, which are available at the Harmonised System (HS) six-digit level for Indonesia in the modelling tool SMART. To avoid estimating these elasticities at the HS10 product level, we assume all HS10 products within an HS6 category have the same import elasticity. Doing so allows us to move the term ̂ (1 + ) to left-hand-side of equation (1), thus addressing the endogeneity of tariffs. The use of estimated elasticities introduces some noise in specification (1) changing the error structure. We cluster standard errors at the product level to ensure their asymptotic robustness. Formally, the specification we estimate is: − ̂ (1 + ) = 1 + 2 + + + + (2) where is the new error term. The results reported above are also robust of our estimates to the use of product-quarter and year dummies, or product-quarter and year-quarter dummies in place of and . 5 Annex 2: How to derive the cost of trading from the estimation of the elasticity of imports to the trade measures Taking the example of BPOM, we have estimated through equation (2) the average elasticity of ̂ ̂ , expressed here as ( imports with respect to BPOM reform, i.e. ). We can re-express this estimated value as: ̂ ( ) = [( )×( )] = [ ] × [ ] where p is the import price. The last equality assumes that and are independent (which are likely to be). Getting read of the expectation operator to save clutter we then have: [ ]×[ ] = [∑ ( )×( )] × [ ] Where i is the import product. So that: ̂ ( ) ̂ 2 − 1 [ ]= = ̂ ̂ ∑ [( ) × ()] [∑ ( ) × ( )] ̂ ̂ Where 2 is our estimated reduced form coefficient, ( ) is the estimated elasticity of import i with respect to import tariff rate i (from Kee at al.) and is the share of sector i in total imports (across the entire period). As the elasticity is available at the HS-6 digit level that is the level of disaggregation that we should employ. 6