以風險中立機率分配探討復徵證所稅之影響

Abstract

政府在推行許多政策時都可能造成金融市場的波動，並對證券市場造成深刻的影響。2012年初，前經建會主委劉憶如即轉任財政部長，並提出復徵證券交易所得稅之議題，與此同時如何穩定社會大眾對金融市場的信心，也是稅制改革須考量的因素。本研究透過研究文獻中對復徵證所稅事件做了詳細的探討，並且歸納和整理出了八個事件與其對應的實證假說。 本研究以選擇權價格隱含的市場資訊為研究對象，藉由風險中立機率分配與隨機優勢檢驗理論之結合，提出一簡便計算且容易操作之實證模型。透過風險中立機率分配的消減變化，能推測市場對於事件所產生之反應。實證結果呼應Reese(1998)主張資本利得稅率調整之行為，會對股票市場報酬分配造成影響。 在假說一、二、三、五、七和八中，本研究經風險中立機率分配理論推測對市場會造成影響，並經由隨機優勢檢驗後說明其為顯著，正好呼應Somers (1948) 、Stiglitz (1983)認為課徵資本利得稅是造成經濟波動的因素，且加深股價波動之程度。此外，假說七、八顯著證明當復徵證所稅相關事件產生對投資大眾有利之訊息時，對證券市場會構成正面影響，也呼應了Lodhi and Shah (2013) 對巴基斯坦證券市場的看法，認為當緩徵、停徵資本利得稅等對投資大眾有利之消息一出，將會有效恢復投資人之信心。整體而言，本研究透過更易操作之實證模型，做出相較於相關文獻更深入之探討。

The former Finance Minister of Taiwan, Christina Liu, proposed to regain the capital gains tax on Taiwan stock market in the beginning of 2012, and there was a sharp decline in stock index subsequently. In this study the events of imposing the capital gains tax on Taiwan stock market was analyzed in detail and the eight events with their empirical hypotheses were summarized. Each market event was analyzed by the combination of risk-neutral probability density functions and was tested for stochastic dominance. In hypotheses, no. 1, 2, 3, 5, 7, and 8 have significantly impacts on the stock market, which exactly corresponded with the findings of Somers (1948) and Stiglitz (1983) that considered the imposing capital gains tax on securities was the factor which leads to economic fluctuations and the increasing of the volatilities of stock market. In addition, hypothesis no. 7 and 8 prominently proved that when market events of imposing capital gains tax on Taiwan stock market are beneficial to the investors, a positive impact would be imposed in the stock market. It corresponded with the results of Lodhi and Shah (2013) who analyzed Pakistan stock market and found that a market event, such as the delay or the suspension of imposing capital gains tax on securities can cause the investors to restore their confidence. From an overall perspective, this study has conducted a relatively decent research than other past research papers through the use of a more user-friendly empirical model.