This paper attempts to examine which theory is best at explaining the geographic concentration in Java, an island in which most of the Indonesia’s large and medium manufacturing industries have located overwhelmingly. Using the regional specialization index as a measure of geographic concentration of manufacturing industry and pooling data over the period 1991-1996, our econometric analysis integrates the perspectives of industry, region (space), and time. The most striking results is that most of the NCT (Neo-Classical Theory) hypotheses can be rejected. Moreover, most of the findings support the NTT (New Trade Theory) and NEG (New Economic Geography). Our findings suggest that manufacturing firms in Java seek to locate in more populous and densely populated areas to enjoy both localization economies and urbanization economies, as shown by the significance of scale economies and income per capita. The interplay of agglomeration economies is intensified by the imperfect competition of Java’s market structure. This paper gives empirical evidence with respect to path dependency hypotheses. This finding supports the NEG’s belief that history matters: older firms tend to enhance regional specialization. In addition, the results, as shown by statistical significance of its regional dummy, suggest that most of the specialized industries in Java have better access to infrastructure.