本研究選取台灣2010年第1季至2024年第2季之季資料，涵蓋六大都會區（台 北、新北、桃園、新竹、台中與高雄），結合LASSO（LeastAbsoluteShrinkageand Selection Operator）進行變數選取，並採用分位數迴歸模型（QuantileRegression）探 討總體經濟變數與LTV政策在不同房價區段下之影響差異。此方法同時兼顧高維資 料處理能力與分布異質性辨識，有助於補足傳統OLS模型的限制。 實證結果顯示，匯率（TWD/USD）變動在所有城市與各分位數中皆呈穩定負向 效果，意即台幣升值期間，房價普遍上揚，顯示資本流動與購屋預期同步影響市場 價格。物價變數（CPI）與貨幣供給（M2）則展現地區性與價格分位之異質性，例 如CPI滯後在桃園、新竹與台中的高價市場表現顯著正向，而M2在新北與新竹高 價區段之推升效果尤為明顯。政策面方面，LTV限貸措施普遍於高分位數顯著負向， 顯示其對高價市場確實發揮抑制效果，且不影響中低價位購屋族群，符合精準調控 原則。此外，不同城市間房價反應機制亦存明顯差異。以新竹市為例，受科技產業 聚落與高所得群體影響，其房價對M2與CPI滯後項特別敏感；高雄則在LTV政策 下反應較為顯著。 整體而言，本研究證實房價並非均質市場，總體經濟變數與政策措施對不同價格 層與地區市場具備顯著異質性。分位數迴歸結合LASSO模型可有效揭示此一結構， 亦為落實分區調控與差異化政策提供實證依據。建議未來政府在制定房市政策時， 應考量市場內部的結構性差異，精準對應各區段之風險與動態，提升政策效率並維 持住宅市場之穩定與公平。;This study investigates the impact of macroeconomic variables and loan-to-value (LTV) policy on housing prices in Taiwan. The dataset covers quarterly data from Q1 2010 to Q2 2024 across six major metropolitan areas: Taipei, New Taipei, Taoyuan, Hsinchu, Taichung, and Kaohsiung. Employing the Least Absolute Shrinkage and Selection Operator (LASSO) for variable selection and combining it with quantile regression, this research captures the heterogeneous effects of economic and policy variables across different housing price seg ments. Empirical findings reveal that the TWD/USD exchange rate has a consistently significant and negative effect on housing prices across all regions and quantiles, indicating that TWD appreciation leads to higher housing prices. Inflation (CPI) and money supply (M2) show varying effects across cities and price levels, with the strongest impact observed inthemid-to highquantiles. Onthepolicyfront, theLTVrestrictionpolicyexhibitsasignificantlynegative effect in the upper quantiles, particularly in high-price markets, suggesting that differentiated credit control measures are effective in mitigating housing market overheating. Overall, the study finds that Taiwan’s housing prices are jointly influenced by macroe conomic conditions and policy instruments, and that these effects are heterogeneous across regions and price distributions. The results provide empirical support for targeted policy interventions and regionally adaptive housing market regulation.
