marginal effect not consistent with odd ratio

[samaneh khaef]

Hi.
I am trying to examine the likelihood of migrants live in different regions. my reference category is small city. The issue is that when I run mlogit, rrr for refugee for living in metropolitan areas is greater than 0 but when I run margin it is negative. how come it is possible? if I interpret mlogit, I could say refugees are more likely to live in metropolitan areas than small cities but according to margin they are less likely. any comment?

result of mlogit	Rural	Larger cities	Metropolitan cities
Migrant category	rrr	rrr	rrr
Family	0.782***	1.51***	1.554***
Labor	0.463***	8.804***	5.452***
Refugees	0.402***	1.731***	1.135***

margin result

	dy/dx	std. err.	z	P>z	[95% conf.	interval]
Migrant category
Family	0.0580	0.00189	31	0	0.054901	0.062308
Labor market	0.116	0.002734	42.5	0	0.110729	0.121444
Refugees	0.0380	0.002784	-11.5	0	-0.037444	-0.026529

[Richard Williams]

In mlogit there are multiple coefficients for independent variables and multiple marginal effects. Coefficients for each variable can be both positive and negative, as can the marginal effects. I think if you showed the complete results for mlogit and margins it would be clear that there is no contradiction.