在許多的交通現象中，個體的行為可能會受到「社會互動」的影響；然而，回顧過去做 為政策評估基礎的交通行為模式，多半視個體行為係獨立於他人行為的脈絡，此將使得社會 互動存在下的交通政策評估產生偏誤。有鑑於此，本研究選定了在台灣各市中心造成嚴重問 題的「機慢車路邊停車行為」進行考察，以台南市中心街區做為實證地區，對機慢車使用者 實施敘述性偏好問卷的面對面訪問調查，透過系統隨機抽樣的方式，共獲取了 192 個有效樣 本。本研究立足於「互動模式」理論，利用蒐集的實證資料來建構「導入社會互動效果的停 車地點選擇行為模式」，此模式的特點在於，其能彈性地處理內生性的識別問題，並可整合 傳統的離散選擇模式。模式校估結果顯示，「團體平均選擇水準」會對個體的決策行為產生 顯著的正向影響，其代表了機慢車使用者於「停車地點選擇行為」上有著遵從他人行為的傾 向，而支持了於該行為中社會互動的存在。另外，於模式校估結果的基礎上，所進行之「考 量社會互動下的政策變數效果分析」中顯示：若依一般的效果評估方式，當減少 1% 的「停 車場至目的地之距離」時，選擇「路外停車」的「機率變動百分比」為0.81%，但在納入了 「社會互動的連鎖效果」後，「機率變動百分比」則增為 3.38%；當減少 1% 的「使用停車 場之固定費用」時，若依一般的效果評估方式，選擇「路外停車」的「機率變動百分比」為 1.76%，而在納入「社會互動的連鎖效果」後則增為 7.33%。此結果意味著，在針對可能存在 社會互動的交通行為進行分析時，若未將社會互動納入考量，則易使相關政策的評估產生偏誤。

The effect of social interaction on individual behavior possibly exists in many types of travel phenomenon. However, the previous travel behavior models, used to be bases for policy evaluation, assume that individual behavior is independent of others, and therefore lead to the bias in the evaluation of transportation policies under the circumstance that social interaction exists. In view of this, this study examines the on-street motorcycle and bicycle parking behavior which has resulted in a severe negative effect in the city centers of Taiwan. The city centre of Tainan City is designated as the study area. The stated preference questionnaire survey aimed at the motorcycle and bicycle users is conducted by face-to-face interview. 192 samples are acquired through the systematic sampling with a random start. Subsequently, the parking location choice behavior model incorporating social interaction effect is constructed based on the theory of interaction-based model which is capable of resolving the endogeneity problem flexibly and integrating traditional discrete choice model. The estimation result shows average choice level of group has the significant positive effect on decision of individual. This result represents that motorcycle and bicycle users tend to conform to the others’ behavior in terms of parking location choice. It is recognized that social interaction among users’ behavior exists. The policy analysis is also employed. First, according to the method of general effect evaluation, which is often employed in previous studies, the percentage change of the probability of individual choosing off-street parking would be 0.81% given the distance (from the parking lot to the destination) decreases 1%, whereas given the same circumstance, the percentage change of the probability would be 3.38% according to the proposed method incorporating the chain effect of social interaction. Secondly, according to the method of general effect evaluation, the percentage change of the probability of individual choosing off-street parking would be 1.76% given the fixed cost (of using the parking lot) decreases 1%, whereas given the same circumstance, the percentage change of the probability would be 7.33% according to the proposed method. The results imply that the biases exist in the evaluation of related policies if the social interaction effect is not considered in the analysis on the travel behavior which is influenced by the social interaction.