ITC HS harmonisation method for trade data analysis - a proposal

(I have used LaTeX in this post for math symbols. If you are using a mobile browser, you may consider requesting a desktop version of the page in order to view the table and formulae properly)

The International Trade Classification - Harmonised system (ITC HS) codes are used for classification purposes in international trade. All products traded internationally are represented though these codes. The codes extend upto 8 digits in India, first 6 of which are harmonised with the international system. Most of the countries are harmonised at 6 digits which implies that the product classification upto 6 digits is common. However, the next two digits are country specific and this sub classification or further bifurcation is a function of statistical and industry needs. Some countries have further two digits, taking the number of digits to 10 to help further statistical purposes. In India, it stops at 8 digits. First two digits in the classification are known as chapter number (01 to 99), the third and fourth digits make up the 'heading', the fifth and sixth digits are called 'sub headings'. The last two digits are called 'tariff headings'. A tariff heading is usually given a number ending with 00 in order to classify all 'other' products falling under the sub-heading after important ones are allotted various numbers starting with 11 and running upto theoretically 99. At times, this 'other' which ends with '00' grows big enough to warrant it's own tariff heading. I have seen tariff headings where the data under 'other' warrants a split to get granularity in data, but it has not been done. At other times, I have seen tariff headings that take up a huge bandwidth (a disproportionate share of the trade under the heading) and yet they have not been split to show further granularity. At times, it feels that the process could be automated. However, till now as I understand, the Govt. depends on various representations from industry and requests from statistical bodies for a split in the sub-headings. The requests are not always very scientific and arise out of ad-hoc needs. 

This I feel may be avoided by automating the function of splitting sub-headings of ITC HS. The proposal below suggests a method to decide if a particular tariff heading needs further bifurcation into lower levels, e.g. from 6 digits to 8 digits, or from 8 digit heading ending with 90 or 00, to further divisions and so on. 

The proposal would use the same technique used in calculating Herfindahl­ Hirschman index for market concentration in antitrust/competition law cases. The proposal doesn’t apply to cases where the new heading is needed due to environment/public interest and so on. Such headings can be considered on ad­hoc basis. The proposal utilizes the relative weight and not absolute value method. The absolute values in terms of trade in ‘X’ crore rupees might become cumbersome as the relative importance is missed out. e.g. a 20 crore exports under a particular heading under handicrafts is significant enough to have a separate heading whereas even 200 crore is negligible in the case of diamond exports. Hence a technique of harmonization that looks at relative importance with respect to parent heading might prove superior to a method of looking at absolute value. The core idea behind this proposal is to develop a system that can be programmed, thus obliviating the need of human intervention in drilling down the headings. The working is illustrated with an example below, in which a case of 6 digits HS codes at sub-heading level are being considered for further harmonization up to 8 digits.

Steps:

Step 1

List down the 6 digit codes under consideration, with the total trade value of previous year and the corresponding share in their parent heading at 4 digit. Let’s take the heading 8413, under which the eight 6 digits subheadings are:

ITC HS Code	Trade Value	Fractional Share in Parent 4 digits (8413)
841311	A1	$S1 = \frac{A1}{\sum A}$
841319	A2	$S2 = \frac{A2}{\sum A}$
841330	A3	$S3 = \frac{A3}{\sum A}$
841350	A4	$S4 = \frac{A4}{\sum A}$
841360	A5	$S5 = \frac{A5}{\sum A}$
841370	A6	$S6 = \frac{A6}{\sum A}$
841381	A7	$S7 = \frac{A7}{\sum A}$
841391	A8	$S8 = \frac{A8}{\sum A}$

Step 2 

$Index number =  S1^2 + S2^2 + S3^2 + … + S8^2 = \sum S^2 $

Step 3

If Index number is greater than 0.25, there is a case for further bifurcation of that subheading at 6 digit which has the highest share in trade into 8 digits. 

Step 4 

The process is repeated till all high value 6 digits are bifurcated into 8 digits. The highest possible value for the $Index Number$ is 1 (only one heading with 100% share). $Index Number$ below 0.01 indicates a highly distributed trade. $Index Number$ below 0.15 indicates an unconcentrated trade. $Index Number$ between 0.15 to 0.25 indicates moderate concentration. $Index Number$ above 0.25 indicates high concentration of trade requiring further granularity. The concentration begins somewhere at the point where the $Index Number$ reaches 0.15 and crosses the threshold once the number reaches 0.25, at which point it should be further drilled down into lower headings. 

A worked example with some numbers is given below: 

Let’s take a case where there are 16 sub­headings at 6 digit level. Now, we will consider two cases in which the six largest 6 digit headings (out of those 16) cover 90% of the trade value: 

Case 1: Six sub-headings share 15% of trade each, 5 sub-heading share 2% each, and rest 10 sub-headings share 1% each. 

Case 2: One sub-heading covers 80% while five other sub-heading cover 2% each, and the rest 10 subheadings share a minor 1% each like Case 1. 

By inspection (and common sense) we can say that the first case would be fine at 6 digits without further harmonisation, whereas the second case is apt for further bifurcation/harmonisation. 

The Index Number for these two situations makes it strikingly clear: 

Case 1: 

$Index Number = (0.15^2+0.15^2+0.15^2+0.15^2+0.15^2+0.15^2) + (10 \times 0.01^2) = 0.136 $

Case 2: 

$Index Number = 0.80^2 + (5 \times 0.02^2) + (10 \times 0.01^2) = 0.643 $

The squaring of the shares this way in the index number calculation penalises bigger shares more than the smaller ones, giving additional weight to headings with larger size. The threshold value 0.25 is based on popular usage of this threshold in antitrust/competition law cases of market monopoly. In a way, any higher share of one among many is a kind of monopoly, hence calling of use of such an index number. The above steps can be easily programmed in simple tools like excel. Any request that comes for addition of a sub­heading at 8 digit can be evaluated using this method and if deemed fit, may be taken up for harmonisation.