This paper investigates the influence of the shares of non-interest income and the diversification, which result from off-balance sheet activities, on the performance of banking in Taiwan. I also use ARCH-M model in this study. The sample period is from January 2000 to December 2011. I find that all sample banks and private banks do not benefit from off-balance sheet activities, but government banks benefit from off-balance sheet activities. In this paper, I also use CUSUM test to find the structural breakpoint and discuss the situation in accordance with it. The result shows that the structural breakpoint is at July 2002, which closes to the founded time of lots of financial holding companies. After the structural breakpoint, the positive impact of shares of non-interest income and diversification on the performance disappear.

This thesis is on the estimation models for the daily global solar irradiation on a horizontal surface at the surface of the Earth. New approaches are developed and applied to 15 locations of Turkey and Germany. The main idea in developing the new models is to link the surface data to satellite imagery. Totally three new modeling approaches are developed and tested. Surface data of daily bright sunshine hours (s) and cloud index (n) values derived from satellite images are the main input parameters. In the first approach, monthly coefficients of Angstr&ouml / m linear relation between daily solar irradiation and daily bright sunshine hours are used to replace the clear sky irradiance calculations of HELIOSAT model. In the second approach new correlations are obtained between daily bright sunshine hours (s) and cloud index (n). Last model that is developed uses a new correlation expression between daily solar irradiation and daily data of s and n and this expression is derived using a physical consideration. The performances of the proposed models are tested against conventional methods (mainly, satellite-based HELIOSAT and ground-based linear Angstr&ouml / m-Prescott type). The results show that the use of sunshine duration together with the cloud index is quite satisfactory in the estimations of daily global solar irradiation. The accuracy of estimations of the combined models is considerably higher than the conventional approaches. Therefore, we propose to use the new approaches to estimate daily global irradiation whenever the data of bright sunshine hours is available for the location of interest or a nearby station. In addition, for a point on the earth surface, depending on the data in hand, suitable and most accurate models of estimations are proposed for that point.

This study investigates whether code-switching practices among Sesotho-English bilinguals promote convergence between Sesotho and English. First, the study identifies different types and patterns of code-switching between Sesotho and English and analyses them using Myers-Scotton’s (1993) Matrix Language Frame model and Myers-Scotton and Jake’s (2000) 4-M model. Second, it applies the ML turnover in order to detect convergence in Sesotho-English code-switching data and to observe which direction it takes. The study also explores other factors contributing to change in the structure of Sesotho, which are not necessarily influenced by convergence. In conducting this study, data was collected through interviews that were held with younger bilingual speakers from different tertiary institutions in and around Maseru (Lesotho) and through recorded youth-centred phone-in radio programmes. Findings from the analysis of data reveal simple to complex Sesotho-English code-switching performance of various types and strategies. Findings also show through the existence of composite language in Sesotho-English code-switching that there is a turnover in the ML, which indicates a development of an asymmetrical convergence between Sesotho and English. It was also discovered that, although other changes in the Sesotho structure are not English influenced, they are enhanced mostly by younger urban bilingual speakers’ frequent “looser” approach to Sesotho. This is an indication that Sesotho’s susceptibility to change correlates strongly with age; that is, both the length of time contact between Sesotho and English has existed, and the generation in which change is mostly found. This thesis adds and documents a different perspective to the previously recorded changes on Sesotho-English contact in Lesotho.

Computing / M. Sc. (Information Systems)

本文係針對臺灣50指數期貨與基金於2003年6月30日上市之後，臺灣期貨及現貨市場報酬率間領先落後關係與波動性的變化來進行探討。研究分為兩部份，第一部份是觀察臺灣50指數期貨與現貨之間的關聯性，並探討臺灣加權股價指數、金融保險類股股價指數及電子類股股價指數期貨與現貨市場間的變化；第二部份是採用可模擬現貨走勢的臺灣50指數基金、國泰金及臺積電的股價來做為現貨的替代變數，觀察其與期貨之間的關連性是否與第一部份的結果類似，若是實證結果極為相同，則相關機構與一般投資人將可運用各期貨與其標的指數中市值最大的股票來進行套利操作。此外，本文在進行模型估計時，首度採用一階段估計法，來聯合估計雙變量GARCH模型中的條件平均數方程式與條件變異數方程式，以避免過去相關文獻將兩條方程式個別估計時所造成的估計誤差。 實證結果所獲得的重要結論如下：首先，臺灣期貨市場的發展仍未趨成熟，並不具有價格發現的功能，在考慮風險溢酬方面，僅有臺灣50指數期貨與現貨的投資人會在報酬率之外，額外要求用以補償的風險溢酬，再者，臺灣50指數期貨與基金的上市，並沒有對臺灣現有的期貨與現貨市場造成顯著的影響，然而，替代變數並不能完全取代現貨指數，但相較之下，國泰金在臺灣50指數期貨與基金上市之後的那段期間模擬成效最好。 / This paper investigates the change of lead-lag relationship in returns and volatilities in Taiwan futures and spot markets after the introduction of Taiwan 50 Futures and Taiwan Top50 Tracker Fund (TTT) on June 30, 2003. The study divides into two parts. The first part examines the relationship between Taiwan 50 Futures and spot markets, and also discusses the change of Taiwan Stock Exchange Capitalization Weighted Stock Index, Taiwan Stock Exchange Banking and Insurance Sector Index, and Taiwan Stock Exchange Electronic Sector Index in futures and spot markets. Another part uses the stock price of TTT, Cathay Financial Holding Company and Taiwan Semiconductor Manufacturing Company as the substitutive variables of spot index and goes a step further to examine the relationships between them and futures individually. Additionally, this research used One-Pass Method for first time to estimate jointly the conditional mean equation and conditional variance equation of Bivariate GARCH Model to avoid estimating error in previous relative studies with Two-Pass Method. The major empirical results are as follows: first, the development of Taiwan futures market is incomplete. The futures market does not play the price discovery role to the spot market. Second, under the consideration of risk premium, only investors in Taiwan 50 Futures and spot markets would ask for compensated risk premium excepting returns. Third, the opening of Taiwan 50 Futures and TTT does not influence significantly Taiwan futures and spot markets. Last but not least, these substitutive variables can not replace spot index perfectly. However, comparing with others, the stock price of Cathay Financial Holding Company is the very model of Taiwan Stock Exchange Banking and Insurance Sector Index after the introduction of Taiwan 50 Futures and TTT.

本研究第一部份為確定性的資產訂價模型，探討證券交易所得稅與證券交易稅對股票價格的影響。如同資產價格方程式，其均衡價格為未來各期股利之貼現值的總和。所不同的是一般資產價格方程式之貼現率即為市場利率，但本研究特別引進資產存量動態調整方程式與政府預算限制式，說明除了市場利率以外，強調證券交易所得稅、證券交易稅與代表性個人之週轉率均為影響貼現率的主要因素。由於模型描述代表性個人具有遠瞻(forward-looking)的特性，所以當期暫時性的稅率變動不會影響股票價格，會影響股票價格的只有未來的租稅政策。另外，本研究亦得到與李嘉圖定理相似的中立性假說：只要租稅現值保持固定不變，即使資本利得稅與證券交易稅如何調整與變動，都不會影響其均衡價格。 為了瞭解股票週轉率的函數型態。假設其為下一期股票報酬率與租稅的函數，以台灣地區1982年1月至2001年1月的月資料，使用兩階段的研究方法分別探討租稅對股票報酬率與週轉率的影響，目的是確認：租稅除了會直接的影響股票價格外，是否會透過週轉率而間接的影響股票價格。因此本研究第二部份摒除週轉率與政府政策為外生的假設，另以實證分析租稅對股票報酬率與週轉率的影響，以補基本理論模型之不足，並進行有趣議題之分析。 首先由Ganger因果關係檢定得知：週轉率領先於政府訂定的證券交易稅稅率，亦即週轉率過去的落遲項會影響到當期的證券交易稅稅率；且由相關係數得知二者呈正相關。故政府訂定的證券交易稅稅率視過去的週轉率而定，過去的週轉率愈高，政府當期訂定的證券交易稅稅率愈高，以企圖抑制股市的不健全發展。再者，由於台灣股票報酬率的時間序列資料具有自我迴歸條件非均齊變異數(autoregressive condition heteroskedasticity，簡稱ARCH)的現象，故異於其他相關研究的實證方法，另經由比較各種ARCH族模型的實證結果，選取出一套考慮風險貼水(risk premium)之最佳配適的兩階段ARCH(1)-M模型進行實證分析。 實證結果顯示：任何一個ARCH(1)-M的模型，皆存在顯著的風險貼水效果。而預期未來的與當期的證券交易稅對當期的股票報酬率均為直接的負面影響，但以前者影響較為顯著，後者的影響則視顯著水準而定，因此政府調降或調高當期的證券交易稅稅率，與股市之榮枯並無確定性的直接關係，此結果與一般學者的觀點不謀而合。此外，當期的週轉率租稅彈性為-0.0646，其值顯著為負，且絕對值小於一，表示缺乏彈性，雖然租稅會間接地透過週轉率影響股票價格，但影響並不大，而且政府若鑒於過去的週轉率太高，要以租稅抑制當期的股市投機風氣，其成效亦不彰。至於證券交易所得稅因宣告(或施行)的期間極為短暫，故對當期的股票報酬率影響不顯著。換言之，風險趨避程度、證券交易稅、預期報酬率、週轉率與股票價格之間具有相互連動的關係。是故，證券交易稅對股票價格的影響，視風險趨避程度、證券交易稅稅率、週轉率之租稅彈性與週轉率之預期報酬彈性而定。 除此以外，本研究第三部份亦考慮政策的不確定性與政策的變動以建立不確定性的資產訂價模型，模擬分析結果顯示：若原先政府施行的租稅過程愈具持續性，則一旦政府的租稅政策瞬間改變時，投資人愈無法將股票價格-股利比值對租稅的反應，事先地考慮於股票投資的決策當中，所以租稅政策衝擊的反應較大。簡言之，投資人預期的稅率與實際施行的新稅率愈相近時，租稅政策衝擊的反應最小。因此第三部份的結果亦與第一部份的基本數學模型(認為當期暫時性之稅率變動不會影響當期的股票價格)，以及第二部份的實證結果(發現當期證券交易稅稅率變動與當期股市之榮枯並無確定性之關係)相一致。 第一章 緒論……………………………………………………………………001 1.1 研究動機與目的………………………………………………………001 1.2 研究方法………………………………………………………………004 1.3 研究結構………………………………………………………………011 第二章 我國證券交易課稅制度沿革…………………………………………014 2.1 我國證券交易所得稅制度沿革………………………………………014 2.2 我國證券交易稅制度沿革……………………………………………021 2.3 各國證券交易課稅制度比較…………………………………………024 第三章 證券交易課稅的文獻探討……………………………………………031 3.1 證券交易所得稅的文獻探討…………………………………………031 3.1.1 關於證券市場的文獻探討……………………………………031 3.1.2 關於投資及儲蓄的文獻探討…………………………………040 3.1.3 關於經濟效益的文獻探討……………………………………042 3.1.4 關於稅制及稅收的文獻探討…………………………………046 3.2 證券交易稅的文獻探討………………………………………………050 3.2.1 國外文獻探討…………………………………………………050 3.2.2 國內文獻探討…………………………………………………053 第四章 確定性的資產訂價模型……………………………………………… 058 4.1 基本理論模型…………………………………………………………058 4.2 租稅效果………………………………………………………………064 4.3 中立性的假說…………………………………………………………069 第五章 台灣實證研究—資料處理與分析…………………………………… 073 5.1 資料來源與說明………………………………………………………074 5.2 檢定……………………………………………………………………076 5.2.1 單根檢定—ADF 檢定…………………………………………076 5.2.2 常態分配檢定…………………………………………………080 5.2.3 自我相關檢定—Ljung-Box Q(L-B Q)檢定…………………081 5.2.4 序列相關 LM 檢定……………………………………………082 5.2.5 ARCH LM 檢定…………………………………………………082 5.2.6 Granger 因果關係檢定………………………………………086 5.2.7 當期股票報酬率與未來各期證券交易稅稅率的相關係數…093 第六章 台灣實證研究—兩階段 ARCH 族實證模型………………………… 095 6.1 基本條件平均數方程式的設定與分析………………………………095 6.1.1 傳統自我迴歸模型……………………………………………095 6.1.2 自我迴歸模型的殘差分析……………………………………096 6.2 條件變異數方程式的設定……………………………………………101 6.2.1 ARCH 模型……………………………………………………102 6.2.2 ARCH-M 模型…………………………………………………103 6.3 估計方法與檢定………………………………………………………105 6.4 第一階段之股票報酬率的 ARCH 族模型估計與分析………………106 6.4.1 實證模型………………………………………………………106 6.4.2 實證結果與分析………………………………………………113 6.5 兩階段模型的估計與分析……………………………………………114 6.5.1 實證模型………………………………………………………114 6.5.2 實證結果與分析………………………………………………119 第七章 不確定性的資產訂價模型…………………………………………… 130 7.1 不確定性的租稅政策與政策變動……………………………………130 7.2 模型的模擬與分析……………………………………………………136 第八章 結論與未來的研究方向………………………………………………151 8.1 結論……………………………………………………………………151 8.2 未來的研究方向………………………………………………………155 附錄A 股票價格函數滿足終極條件的證明…………………………………158 附錄 B 股票價格函數為單一固定點的證明…………………………………160 參考文獻…………………………………………………………………………163 / This paper works out the effect of capital gains tax and stock transaction tax on share prices. In the generality of cases, the equilibrium share price is shown to equal the discounted sum of future dividends. In this paper, we emphasize that tax policy is a driving force in determining the discount rate and market turnover rate also plays an important role in determining the effects of policy. We show that temporary changes in policy have little effect on current stock price; only perceived policy in the future has price effect. We also show that a permanent shift in tax policy exerts only level effect, but no growth effect, on stock price. Using monthly data from Taiwan, our empirical results indicate that changes in stock turnover rate often leads changes in transaction tax, and both are positively correlated. This result implies that government policy often reacts to market volatility instead of the other way around. Since the data exhibit autoregressive conditional heteroskedasticity, we employ a two stage ARCH(1)-M model as our empirical strategy. Our results show that the price effect of transaction tax is significant, with the expected change in future tax especially so. The results also indicate that expected stock returns have positive and important effects on market turnover rate. Overall, the policy effect on stock price depends on the tax rate, the degree of risk aversion, sensitivity of turnover rate to policy and the expected stock return. The analysis so far assumes that the tax policy is certain, when in fact it is hardly the case. The 1989 fiasco and the recent flip-flop of the government are vivid testimonies of the volatile nature of the government policy. The third part extends the model to a stochastic environment and examines the consequences when a long-standing unequivocal tax policy suddenly and unexpectedly shifts to a stochastic regime. This type of regime shift seems to characterize the recent experience in Taiwan. Our results show that the shape of the pricing function under stochastic regimes depend on the persistence of the tax process. For tax rates that are lower than the unconditional mean of the process, the price is lower than the certainty case because the expected tax rate is higher over this interval. As the tax process becomes more persistent, the expected duration of the tax rate lengthens and the pricing function becomes negatively sloped. Notice that the slope of the pricing function (i.e., the tax elasticity of the price-dividend ratio) is flatter than the certainty case because in a stochastic regime consumers always take into account the possibility of tax changes, no matter how small the probability is.

Computing / M. Sc. (Information Systems)