Models,　which　employ　panel　data　analysis　to　model　highway　crashes,　can　identify　fixed　differences　and　other　unobserved　factors　in　real　world.　This　paper　introduces　process　of　individual　fixed-effects　and　random-effects　models　and　related　tests.　These　models　are　applied　to　Jingjintang　highway.　Pool　data　regression,　models　of　fixed　effects　and　random　effects　are　established　respectively.　Hausman　results　show　that　fixed-effect　model　is　better　than　others　when　describing　the　relationship　between　accidents　and　other　factors.