本論文的研究主軸為本國市場效果的實證研究, 文章除驗證新貿易理論對創意商品的適用性, 並解析貿易成本與需求規模對本國市場效果的交互影響與其間的非線性關係。實證資料以經創意商品(creative goods)貿易為例。選擇創意商品,除因其具有文化創意特色,可歸為異質產品並且可凸顯無形貿易成本的影響, 且此類商品根據國際商品統一分類代碼(Harmonized system code
; HS code) 六位碼編組而成, 商品分類可謂細緻, 可避免如Schumacher and Siliverstovs (2006)以及Hallak (2010) 指出以總和商品可能相互抵消效果而致的合成謬誤。並且根據UNCTAD(2008) , 此類商品受景氣波動的干擾較少。由於全球創意商品貿易以經濟合作開發組織(The rganization
for economic cooperation and development;OECD) 國家為主流, 且OECD 國家創意商品貿易具有產業內貿易特性, 故而樣本國家以OECD 國家為主, 並於林德效果檢測時擴及至全球創意商品貿易。
實證估計式以Hsu et al. (2012) 推導的引力模型結構式為理論基礎, 加入一個由貿易成本與相對需求組成的非線性本國市場效果設定項,驗證本國市場效果的相關假說。此一本國市場效果設定項除具有理論基礎之外, 並具有實證意涵與對應的假說檢定, 可為實證工作提供重要基石。
估計式中除加入例行的需求規模、貿易成本以及控制貿易成本對商品流動的多邊阻力(multilateral resistance; MR)外,並加入需求與貿易成本的非線性交互作用項。相較過去文獻如Feenstra et al. (1998, 2001) 或Hanson and Xiang (2004) 側重需求與貿易型態關係的確認, 加入此項不
但可以匡正實證估計是錯誤設定造成估計偏誤, 得以正確驗證本國市場效果, 並且能解析本國市場效果成因的作用機制與政策意涵。
實證過程圍繞兩個命題:
命題1 : 如果本國市場效果存在, 但引力模型忽略或未控制本國市場效果, 將使估計產生偏誤。
命題2 : 相對需求與貿易成本對本國市場效果具有非線性影響關係。相對需求增加, 本國市場效果將強化。相對需求對本國市場效果的邊際影響, 會隨相對需求上升及貿易成本下降而遞減。貿易成本增加, 本國市場效果將弱化。隨著貿易成本愈高其對本國市場效果的邊際影響力遞減。
根據2000年至2005年OECD創意商品貿易資料的實證結果,除確認本國市場效果顯著外,各項相關假說檢定並與Hsu et al. (2012) 的理論預期一致。包括實證估計式忽略本國市場效果設定項, 將使估計結果產生偏誤; 貿易成本、相對需求對本國市場效果以顯著的非線性方式交互影響本國市場效果。而由比較靜態分析的實務案例, 再次確認推論與實證結果的一致性。
為進一步檢測命題2 , 確認相對需求與貿易成本對本國市場效果的非線性影響關係。在此以貿易雙方的相對需求與相對貿易強度進行分量迴歸(quantile regression model ; QR) 估計, 捕捉不同貿易強度相對需求的非線性變化特徵, 並檢定相對需求的邊際影響是否隨不同分量而有顯著的差異。在此, 除創意商品貿易資料外, 並加入OECD 會員國的國際貿易商品統計(international trade by commodity Statistics;ITCS) 國際商品統一分類編碼(HS Code) 二位碼的製造業商品貿易資料為估計樣本。實證結果再次應證相對需求與貿易成本對本國市場效果的非線性影響關係, 並且不論是根據分量迴歸檢定結果, 或以估計係數的信賴區間圖示, 都顯示不同分量下, 相對需求對本國市場效果的邊際影響確實存在非線性影響關係, 並且貿易強度與相對需求間呈現正向的變化關係, 約有90% 以上的商品類別, 相對需求的邊際變動軌跡呈現遞增型式, 符合理論預期的顯著非線性變化關係。
最後以偏好相似理論Linder (1961) ,即林德假說(Linder hppothesis) ; 以需求結構因素解析國際產業內貿易之可能原因。林德假說提出重疊需求(overlapping demand) 的概念, 認為影響一國需求結構的主要因素是所得水準, 即一國的人均所得水準決定了該國特定的偏好模式, 若兩國之間收入水準相似, 則兩國偏好模式可能愈相似, 需求結構也將愈相近, 也就是說重疊需求的量愈大, 從而兩國間貿易量也就愈大。反之, 兩國所得水準相差越大, 需求結構差異也就越大, 從而相互間貿易強度也就越低。本文以Hallak (2010)之一般均衡理論與期建構的實證結構式為基礎, 以UNCTAD(2008) 全球及OECD 國家創意商品貿易資料配合分量迴歸估計, 結果顯示林德假說確實適用於解釋OECD 以及全球創意商品貿易現象, 並且根據貿易流量高低之不同分量係數檢定結果也顯示林德效果確實存有差異, 不同分量下林德效果的邊際效果確實有顯著差異。

This thesis proposes and extends theory-based gravity equations to test the presence of the ‘home market effect’ in the global trade in creative goods. Traditional neoclassical models based on comparative advantage suggest that, all else equal, a country with idiosyncratically strong demand for a good will result in it being an importer of that good. Contrary to traditional trade theory, the home market effect emphasizes the advantages of a large home market as a foundation for exports of a good. The HME is the most obvious characteristic of new trade theory.
The home market effect is caused by increasing returns to scale and trade costs. When it is more profitable for a firm to operate in a single country because of increasing returns to scale, the firm will base itself in the country where most of its products are consumed in order to minimize trade cost costs. The home market effect implies a link between market size and exports.
Here, we imply the framework derived from Hsu’s (2012) model, in which an HME term is appended to the traditional gravity model to capture the home market effect. It is different from Feenstra (1998, 2001) and Hanson (2004) who focused on the linkage between the relative size and the direction of trade. The HME term consists of both the relative market size of the domestic market versus the relative proximity to foreign markets, trade costs and their interactions. It means that except for the traditional gravity equation that includes regular variables such as market size, trade cost and multilateral resistance, which are used to control the obstruction of trade costs on goods traded as the regressors, the extended gravity model appends an HME term which is based on a theoretical foundation that can guide the hypothesis testing and own its empirical content, and can provide a fundamental guideline for empirical study as well as quantitative scenarios.
The extended model can reinvestigate the adaptations to the new trade theory for the trade in creative goods, and analyze and infer the intersection and non-linear relationship for the trade cost and relative market size. We apply the bilateral trade in creative goods in OECD countries as an empirical case study. The creative goods have certain characteristics. The first is that they contain creative and cultural features which can be treated as differentiated goods so that they can withstand the impact of the trade cost. They consist of an HS code of 6 digits, and the classification is sophisticated so that it can avert the fallacious synthesis of Schumacher (2006) and Hallak (2010). Besides, the creative goods fluctuate smoothly as the business cycle is depressed (UNCTAD, 2008).
The empirical issues surround two hypotheses:
Hypothesis One: There will be biased estimates if there exists an HME, but it has been ignored or has been not controlled well in empirical studies.
Hypothesis Two: There exists a non-linear relationship between relative market size and trade costs with the HME. A rise in the relative market size will strengthen the HME and the marginal effect of the relative market size will decline if the relative market size increases or the trade cost decreases. Trade cost increases will weaken the HME and the marginal effect of the trade cost will go up if the trade cost falls.
The estimation results of applying the data for the OECD trade in creative goods from 2000 to 2005 show that the HME exists and thus confirm the hypotheses mentioned above.
In order to investigate Hypothesis Two more precisely, we apply quantile regression (QR) to re-examine and capture the non-linear relationship between the relative market size and trade intensity. Furthermore, we adopt bootstrapping, a non-parametric approach used to construct statistical inferences, to test whether the marginal effect of the relative market size on trade intensity will be significantly different for different percentiles.
One advantage of QR, relative to OLS (ordinary least squares), is that the QR estimates are more robust against outliers in the response measurements. However, the main attraction of quantile regression goes beyond that. In practice, we often prefer using different measures of central tendency and statistical dispersion to obtain a more comprehensive analysis of the relationship between variables. QR has been attributed to the complexity of interactions between different factors leading to data with an unequal variation in one variable for different ranges of another variable.
We employ UNCTAD (2008) trade in creative goods and ITCS databanks for the classification of manufacturing goods at the HS code 2-digit level for OECD countries. The results support the non-linear relationship between relative market size and trade intensity, and confirm that the marginal effects of relative market size on trade intensity are obviously different and show they are positively related, so that the higher the ratio of relative market size, the greater that the intensity of bilateral trade will be.
Contrary to the HME focus on the supply approach, the Linder hypothesis is based on the demand approach used to examine the intra-industry trade. The Linder hypothesis is a conjecture based on trade patterns, that proposes that the more similar the demand structures of countries are, the more the countries will trade with each other.
Furthermore, international trade will also be vigorous if there are similar demand structures between two countries, even if they possess similar factor endowments and technologies.
We also apply the data for trade in creative goods but extend the sample from the OECD countries to the global countries. Based on the general equilibrium framework of Hallak (2010), we use Newey-West estimators and QR as empirical methodology. Examinations of the Linder hypothesis have led to the observation of a “Linder effect” that is consistent with the hypothesis. Econometric tests of the hypothesis usually serve as a proxy for the demand structure in a country based on its per capita income (per capita GDP which is denoted as yit, where i stands for the country index, and t is for time). It is convenient to assume that the closer are the levels of per capita GDP, the closer will be the consumer preferences.
Here, we use the gaps in income between countries such as (ln yit-ln yjt), ln |yit-yjt| and product in income such as ln (yit*yjt) as Linder effects. Neither the results in the income gap or income product show that the Linder hypothesis explains the trade in creative goods for the OECD and Global countries well, and the marginal effects for the Linder effect for different levels of trade intensity are dissimilar. Thus the larger the income gap, the more sensitive is the trade intensity.