Dynamics of EU food safety certification: a survival analysis of firm decisions

Agricultural and Food Economics

Table 5 Results of the means comparison, logit, and survival analyses to explain continued adoption (non-decertification)

Variable	Means comparison		Logit model^a	Survival analysis^b
Variable	Certified	Decertified	Logit model^a	Survival analysis^b
Incentive
Output price level (US$/kilogram [kg])	3.95	4.22	1.28 (0.87)	0.30* (0.12)
EU–US price difference (US$/kg)	−0.01	−0.94***	−0.94** (0.30)	0.01*** (0.00)
EU–others price difference (US$/kg)	1.35	1.09	−0.65** (0.21)	1.09 (0.89)
Price of labor (US$/day)	3.01	2.85	0.33 (0.22)	0.37*** (0.12)
Price of materials (US$/kg)^c	2.25	2.73
Price of capital (US$/kg)	0.20	0.22	−0.01 (0.02)	0.00 (0.00)
Number of markets (number)	3.00	3.00	0.02 (0.02)	0.47*** (0.13)
Number of product forms (number)	2.00	2.00	0.23 (0.21)	2.50*** (0.69)
Output (000 tons/year)	3.70	3.40	−0.25 (0.22)	0.97 (0.65)
Capacity
Difficulty in credit access (dummy = 0, 1)	0.35	0.50	0.42* (0.19)	3.04*** (1.01)
Years in business (number)	10.50	11.30	0.06 (0.05)	1.11 (1.02)
Active in associations (dummy = 0, 1)	0.80	0.60	−0.01 (0.01)	0.30*** (0.05)
Number of associations (number)^d	2.00	2.00
Age of plant facility (number)	9.00	11.20	−0.07 (0.06)	0.81 (0.67)
N			59	59
Pseudo R squared			0.29
% correctly predicted			55%
Schoenfeld residual				0.29

The test of difference in means was performed using a means comparison test (t test) of unequal variances (using both Satterthwaite’s and Welch’s formulas) in the Stata program. Figures in parentheses are standard errors
***Significant difference in averages at <0.01 level of significance; **at <0.05 level of significance; *at <0.10 level of significance
^aFigures presented here are the marginal effects
^bFigures presented here are the hazard ratios
^cDue to a multicollinearity problem from the high correlation between output price level and price of materials, we were able to use only one of these variables at a time in the models. However, the two models showed similar results. The one presented here is the one using output price level
^dDue to a multicollinearity problem from the high correlation between “number of association memberships” and “being active in these associations,” we were able to use only one of these variables at a time in the models. However, the two models showed similar results. The one presented here is the one using “being active in these associations”