• No results found

Review of Methodologies Used in Previous Studies

CHAPTER THREE LITERATURE REVIEW

3.2 Review of Methodologies Used in Previous Studies

The review of methodology will be sub-divided into the following: separating trade flows into intermediate and final products, measurement of intra-industry, disentangling IIT into vertically and horizontally differentiated products, measurement of some explanatory variables and model estimations adopted in the previous studies.

3.2.1 Review of the Methods of Separating Trade Flows into Intermediate and Final Products

The separation of trade flows into final and intermediate products depends on the digit level of the trade flows. In trade data, numbering systems (codes) are used to identify commodities. These commodity codes are hierarchical, in that the longer the digit of the code the more specific is the commodity. There are three approaches often used in the literature to select the intermediate goods from the total trade flows. First, Yeats (2001), Schuler (1995), Keller (1999), and Kol and Rayment (1989) propose that trade in goods identified as parts or components should be considered to be trade in intermediate goods.

Another method is to focuses only on individual SITC 7 group to measure the growing importance of trade in components in international trade, because this industry group consists solely of parts and components (Yeats (2001)). The last method adopted by Turkcan (2003) is to use the Broad Economic Categories Classification Scheme (BEC, 1986). The BEC includes 19 basic categories, including: Categories classified as capital goods, consumption goods, and intermediate goods (111, 121, 2, 3, 42, and 53) etc. BEC scheme has a little limitation. That is some of the categories such as food products (112, 122), fuel goods (321), and capital goods (41, 51) could be consumed directly by consumers, or used as intermediates in the related industry. However, Turkan (2003) maintained that the use of BEC remained the best method of identifying intermediate products.

UNIVERSITY OF IBADAN LIBRARY

59

Table 3.1 Broad Economic Categories Classification Scheme (BEC, 1986)

Commodity Categories Classes

111. Mainly for industry Intermediate goods 112. Mainly for household consumption Final Goods 121. Mainly for industry Intermediate goods 122. Mainly for household consumption Final Goods

21. Primary Intermediate goods

22. Processed Intermediate goods

31. Primary Intermediate goods

32. Processed Intermediate goods

321. Motor Spirit Intermediate goods and Final goods

322. Other Intermediate goods

41. Capital goods (except transport equipment) Final Goods 42. Parts and accessories Intermediate goods

51. Passenger motor cars Intermediate goods and Final goods 53. Parts and accessories Final Goods

61. Durable Final Goods

62. Semi-durable Final Goods

63 Non-durable Final Goods

Sources: John Haveman’s web page:

http://www.macalester.edu/research/economics/PAGE/HAVEMAN, Retrieved 09/10/12

UNIVERSITY OF IBADAN LIBRARY

60

3.2.2 Measurement of IIT

Early attempts to measure the phenomenon of IIT was led by Verdoorn (1960), Balassa (1966), and Grubel and Lloyd (1975) (Grubel 1981). Verdoorn (1960) pioneered the development of the index used to measure IIT. In what he described as the Verdoorn‘s index, he used equation (10), which is the ratio of exports to imports of the same product group to measure IIT.

That is: i i

i

V X

M (10)

where, Vi is the Verdoorn‘s index, Xi is the exports of commodity group i, and Mi is the imports of commodity group i. If the Verdoorn‘s index is closer to 1, it indicates that the commodity group is involved in higher levels of IIT. The argument against Verdoorn index is that it does not identify the extent of IIT in a particular product group.

Balassa (1966) proposed an index of IIT that measured the degree of trade overlap - simultaneous import and export - of goods within an industry:

ii ii

X M

Bj X M

 

 (11)

where i commodity within industry j. This index, the ratio of net trade to gross trade, ranging from 0 to 1, with 0 representing ―perfect‖ trade overlap, and therefore pure IIT, while 1 represents pure ITER. In order to calculate the degree of IIT for all industries (country level), Balassa took an unweighted average for each Bj:

B 1 Bj

n

(12) where n  number of industries. This can be generalised to be a weighted index:

j j

B

w B (13)

UNIVERSITY OF IBADAN LIBRARY

61

where wj  industry j‘s share of total trade.

However, Grubel and Lloyd (1975) argue that Balassa‘s index of IIT should be treated as a measure of ITER. Since, it fails to take into account the individual industries‘ share in total trade or to correct for aggregate trade imbalances. Hence, they proposed an alternative index to measure the extent of IIT.

The groundbreaking work of Grubel and Lloyd (1975) developed the Grubel and Lloyd index. They calculated IIT based on the difference between the trade balance (difference between exports X, and imports, M) of the industry or product i, (Xi – Mi) and the total trade of the same industry or product (Xi + Mi). The G-L index measures the share of IIT of industry i for a given country j as:

 

1

1

n ij ij

ij i

i ij ij

X M

GLIIT GL

X M

   

(14)

where Xij and Mij are home country‘s exports of industry i to country j and home country‘s imports of industry i from country j, respectively. Thus, GLIITij index in (14) measures the intensity or proportions of IIT in industry i. If all trade in industry i is IIT, that is, Xij = M ij , then GLIITij = 1. Similarly, if all trade in industry i is ITER, that is, either Xij = 0 or M ij = 0, then GLIITij = 0. Thus, the index of IIT takes values from 0 to 1 as the extent of IIT increases, that is, 0 ≤ GLIITij ≤ 1. The IIT index in Eq. (14) can be modified to measure the IIT in all products in country j as a weighted measure of the GLIITij ‘s and can be written as

1 1

1

( )

( )

n n

ij ij ij ij

i i

ij n

ij ij

i

X M X M

GLIIT

X M

  

 

(15)

where n is the number of industries at a chosen level of aggregation.

Several criticisms have been made against the original GL index. A common criticism stems from their definition of what constitutes an industry. Early studies on IIT choose some digit level of the Standard International Trade Classification (SITC) to define their

UNIVERSITY OF IBADAN LIBRARY

62

industries. When IIT was measured this way there were several critics such as Finger (1975), Lipsey (1976), Gray (1988) and Rayment (1983) who regarded the observed IIT, or the greater part of it as spurious as a result of inappropriate statistical aggregation.

Aquino (1978) and Balassa (1986a), among others, have suggested adjusted measures to correct this deficiency of Grubel and Lloyd (1975), although many empirical economists prefer and continue to use the Grubel and Lloyd index, otherwise known as unadjusted G-L index. The bias of the unadjusted G-L index includes its ignoring trade imbalances, and not having to choose the ‗correct‘ disaggregation level of data. Aquino (1978) suggests that the G-L index be adjusted with estimates of what the values of exports and imports of each commodity would have been if total exports had been equal to total imports. Aquino (1978) and Balassa (1986), proposed to adjust the index by incorporating overall trade imbalance as follows:

1 1

1 1 1

( )

( )

n n

ij ij ij ij

i i

ij n n n

ij ij ij ij

i i i

X M X M

AGLIIT

X M X M

  

  

 

  

(16)

where AGLIITij is the adjusted IIT index. Since this adjusted measure of IIT index incorporates the total trade imbalance, it is measured with respect to total balanced trade.

However, no consensus exists among scholars on how to adjust for trade imbalance when measuring IIT hence the continuous use of the unadjusted G-L index.

3.2.3 Review of the Methods of Disentangling IIT in Vertically/Horizontally Differentiated Products

Thus far, we have only differentiated between one- and two-way trade types. We now move to disentangle HIIT and VIIT. Within a given commodity classification that experiences two-way trade, products may or may not differ in their quality. In models of IIT, horizontal product differentiation is characterized by products with similar quality levels, with different attributes, while vertical differentiation is characterised by products with significantly different quality levels. According to Stiglitz (1987), empirical work that has disentangled IIT has assumed that prices represent quality, even under imperfect

UNIVERSITY OF IBADAN LIBRARY

63

information. From this assumption, differences in the unit values (UV) or prices of these commodities can be assumed to represent these quality differences. Unit values have been defined for each commodity classification as the value of trade divided by the quantity traded, giving an average price of the goods traded in this category. Clearly, the more disaggregated the classification system, the better this method will be in capturing the price of the commodities. A classification system such as the 6-digit Harmonised Tariff Schedule with 6000 commodity classifications will capture this well. The categories are so specific that different commodities will have different quantity measures: litres, kilogramme, number, etc. while the SITC classification system is more general and uses tonnes as its quantity variable for all commodity categories.

Abd-el-Rahman (1991) pioneered the study of how to separate IIT into vertical and horizontal. He divided trade flows into two types: IIT in vertically differentiated products, and IIT in horizontally differentiated products they separated trade flows into vertical and horizontal on the basis of calculated unit values21 of the involved products.

According to him, trade flows are defined as vertically differentiated when relative unit values are outside this range 0.85 and 1.15. Sometimes, a higher than 15%, difference in unit values is accepted for calculations. IIT is considered to be a VIIT if the following criteria are met:

1

x i

m i

UV

UV   or 1

x i

m i

UV

UV   (17)

IIT is horizontal trade when:

1 1

x i

m i

UV

UV

    (18)

21 Unit value is calculated by dividing the monetary value of ECOWAS country‘s imports and export from EU by their corresponding quantities.

UNIVERSITY OF IBADAN LIBRARY

64

x

UVi : unit value of exports for a product from industry i UVim: unit value of imports for a product from industry i. α is the threshold for the range,

x i

m i

UV UV

 

 

  is the deviation of relative unit values of exports.

Abd-el-Rahman‘s (1991) study was followed by the seminal works of Fontagné and Freudenberg (1997) done for the European Commission (1997). Fontagn´e, and Freudenberg (1997) have suggested a modified criteria that preserves the relative nature of the threshold:

1 1

1

x i

m i

UV

UV

 

 (19)

for horizontal product differentiation, and:

1

x i

m i

UV

UV  or 1

1

x i

m i

UV

UV  (20)

In terms of the choice of the threshold, Fontagné and Freudenberg (1997) and Abd-el-Rahman (1991) also differ a little. Fontagné and Freudenberg (1997) used 15 per cent threshold with the assumption that price differences reflect only differences in quality (the assumption of perfect information), such that a consumer will not purchase a similar, or lower, quality good at a higher price. However, Greenaway, Hine, and Milner, (1998) emphasised the case of imperfect information and that the 15 per cent threshold may be too narrow hence, the choice of the 25 per cent threshold.

3.2.4 Measurement of some the Explanatory Variables 3.2.4.1 Product Differentiation

An empirical measure of product differentiation in international trade flows was first suggested and used by Hufbauer (1970). Theoretical and empirical studies of IIT have

UNIVERSITY OF IBADAN LIBRARY

65

stressed the importance of product differentiation as one of the determinants of IIT (Byun and Lee, 2005; Faustino and Leitão, 2007; Chang, 2009). A difference can be made between vertical and horizontal product differentiation (Faustino and Leitãlo, 2007). Balassa and Bawens (1987) and Hu and Ma (1990) use the Hufbauer (1970) index as a proxy for the degree of product differentiation.

ij ij

The Hufbauer index = 

(21)

where ij stands for the standard deviation of export unit values for shipments of good i to country j , and ij represents the unweighted mean of those unit values. The Hufbauer index has been modified by Fontágne, et. al., (1997) as follows:

The Hufbauer index =

1

( )

( )

n

ij ij

i j ij

Value MAX UV Value MIN UV

  

  

  

 

(22)

where The Hufbauer index= degree of product differentiation, Valueij = export value of host country, that is value of trade for good I in industry j, Valuej = unit value of exports, that is value of trade in industry j, MAX(UVij) = the highest unit value of export of good i in industry j, while MIN(UVij) = the lower unit value of export of good I in industry j.

The computed degree of the Hufbauer index measure is equal to or greater than 1, where values close to 1 indicate low degrees of product differentiation and values further away from 1 is conversant with higher degrees of product differentiation (vertical). According to Fontágne, et. al., (1997), the index provides an average unit value dispersion of export unit values for a given product aggregated over the sum of all products within a given industry and is a measure of vertical differentiation of a product

3.2.4.2 Geographic distance (DIST)

Geographic distance is typically used as a proxy for transport costs, insurance costs, delivery times and market access barriers. Many studies use kilometres or miles to

UNIVERSITY OF IBADAN LIBRARY

66

measure geographic distance between the capital cities of trading partners. Since the commonly used distance variable (Lee, 1992; Hu and Ma, 1999; Sharma, 2004;

Veeramani, 2007) is time invariant, it could not be used in fixed effects (FE) models.

The alternative is to adopt the weighted distance variable which varies over time (Balassa, 1986; Stone and Lee, 1995) as a proxy for geographical distance between countries i and j, where the weight is the ratio of GDP of country j to the sum of total GDPs of all its trading partners and is computed as follows:

1

ij* ij

t j

DIST GDP WDISTij

GDP

23

3.2.5 Review of Methods of Analysis

Various estimation techniques have been employed in examining the determinants of IIT. Studies, such as Burange and Chaddha (2008), McMahon (2003), Havrylyshyn and Civan (1985) and Havrylyshyn and Kunzel (1997) employed descriptive methods of analysis. A handful of studies have used the ordinary least squares (OLS).

However, since the dependent variable is the Grubel Llord index which ranges between 0 and 1, using OLS will cause some econometric problems. The regression equation estimated using the OLS could not predict values outside the intra-industry index range and there could be the problem of heteroscedasticity. Therefore, a number of studies ((Lee and Lee (1993), Musonda (1997) and Tharakan and Kerstens (1995)) have argued that a logistic transformation is appropriate since the dependent variable varies from 0 to 1. However, OLS with logistic transformation also has some problems. First, if the GL index is equal to zero or one, then the dependent variable is not defined, and there are missing values in the dependent variable. This characteristic of the data would make OLS with a logistic transformation awkward because the estimation method would cause much of the data to be lost. Second, it is difficult to interpret the coefficient estimates of explanatory variables even if there are no missing values in the dependent variable.

Papke and Wooldridge (1996) proposed an alternative estimation method: the Factional Logit Regression Model (FLRM). They designed this model to capture the

UNIVERSITY OF IBADAN LIBRARY

67

characteristics of the dependent variables that are bounded between 0 and 1. Most studies that were between regions and their trading partners adopted panel data analysis techniques. Kandogan (2003), Manrique (1987), Shahbaz and Leitao (2010), Sichei and Harmse (2004), Zhang, Witteloostuijn and Zhou (2005), all adopted panel data.