Crop Sciences

Acceptability of Biofortified Orange Maize to Urban Consumers in Mozambique

Robyn Stevens, Torbert Rocheford & Alex Winter-Nelson
University of Illinois at Urbana-Champaign

Overview

Orange maize has been developed at the University of Illinois with increased levels of total carotenoids and vitamin A precursors, which may help in alleviating malnutrition in impoverished regions of the world.  The objective of this research was to determine the acceptability of orange maize to Mozambiquan consumers, who are accustomed to white maize and associate yellow maize with inferior food aid. We attempted to accomplish this through taste tests and trading exercises. Approximately 250 people of varying ages were surveyed over a four day period in two urban marketplaces in Maputo, Mozambique (August 15-18, 2006). Participants compared the eating characteristics of xhima, a local maize porridge, made with their local white maize, a line of orange maize grown at University of Illinois, and a white maize line grown at University of Illinois that is similar to the orange maize, but lacking carotenoid enhancement and the associated color and aroma. After tasting the xhimas, participants were given a quantity of white maize meal, 0.6kg, milled from local maize previously purchased in the marketplace. Participants were given the opportunity to trade the local meal for varying amounts of orange maize meal or a bag of tomatoes approximately equivalent in value to the white maize meal they had received.

We aim to determine what characteristics are important for maize consumption in this region in order to determine if orange maize can be a viable product in this marketplace. We also hope to estimate the market value of this product through the trading exercise. The results of this experiment will allow for further breeding of the maize to increase acceptability, and help to develop appropriate ways of marketing the product in this region.

Method

Orange and white near-isogenic lines of maize were shipped to Maputo, Mozambique from the University of Illinois Urbana-Champaign. Local white maize grain was purchased in the Xiquelene marketplace the day before starting the experiment. All three types of maize were milled without removing the pericarp for the first day; however for the following three days, local white maize meal was purchased in the marketplace that had been milled with the pericarp removed. Clear, plastic bags were filled with approximately 0.6 kg of local white maize meal.  This volume was chosen to approximate a local unit of measure in the marketplace and to provide an amount that could be practically used by the recipient. Bags were also filled with approximately 0.3 kg or 0.6 kg of orange maize meal. Xhima was prepared fresh the morning of each day of the experiment, along with a beef stew to be given upon completion of the taste testing.

Market stalls were set up in Xiquelene marketplace for the first two days, with a different location within the market used for each day. The following two days, stalls were set up in Xipamanine marketplace, again with different locations selected each day.  A market official was present to observe and permit the experiment to occur each day. Upon arrival, an announcement was made to surrounding people of the purpose and procedure of the experiment. A member of the research group participated in the experiment as an example to potential participants of the method that would be followed. Upon arrival at the table, participants were asked to provide some demographic and economic data. Each participant was given a small portion of each of the three types of xhima, asked to taste each separately, and then give a score for the taste, texture, appearance, and aroma of each type. Scores were given as 1-very bad, 2-bad, 3-okay, 4-good, 5-very good. Participants were also given an opportunity to make any comments on the xhimas. Each participant was then given a 0.6 kg bag of the local white maize meal and informed that it was his/hers to keep. The researcher then offered a randomly pre-assigned amount of orange maize meal, between 0.5 and 3 times the amount of white meal the participant had been given (0.5x, 1x, 1.5x, 2x, 3x). The participant was asked if he/she would want to trade the white meal for the amount of orange meal offered and the response was recorded on the survey. The participant was then asked if he/she would be willing to trade his/her white meal for a bag of tomatoes with a value equivalent to the white meal, and the response was recorded on the survey.  In order to obtain a large enough sample with the quantity of orange maize meal that was available, a coin was tossed to determine if a trade would be made if a participant indicated that he/she was willing to trade for the orange meal or tomatoes. Upon completion of the experiment, participants were given some meat and broth to eat with their remaining xhima.

Results – Descriptive Statistics

Descriptive statistics from the full sample are presented in Table 1.1. The median values of the independent variables indicate that a modal respondent is a 37 year old woman who lives in a household of six people, all of whom are over three years of age.  She does the household shopping but not the cooking.  She consumes xhima two days per week, green vegetables four days per week and animal products (eggs, chicken, meat or fish) four days per week. After tasting xhima made from orange maize (milled with the pericarp on) and local maize (milled with the pericarp removed) she expressed a slight preference for the flavor of local maize but considers both xhimas to have a good taste.
Based the estimation results, this modal respondent has a 57% likelihood of accepting a 1:1 trade of orange maize meal for local maize meal.  By extension, if orange maize meal were available on the market at the same price as local maize meal, there is a 57% probability that an “average shopper” will buy orange.  For a respondent characterized by the mean values of the variables and faced with a 1:1 exchange, the model suggests a 48% probability of accepting the orange maize meal.

In general, most participants seemed to prefer the taste, texture, and appearance of their local white maize over the Illinois varieties. The average taste score for the orange xhima was not much lower than that of the local counterpart (3.72 compared to 3.96), indicating that most participants were not strongly averse to it, however the difference was statistically significant (p=0.038). On average, participants preferred the aroma of the orange maize over both white varieties which may increase the likelihood of acceptance (mean score of 4.02 for orange maize versus 3.66 for local white (p=0.002) and 3.63 for Illinois white (p=0.000)). While most participants seemed open to orange xhima, there was a core of about 25% who rated the taste as bad or vary bad.  This compares to 13% who rated local xhima in this way.

Overall texture scores of the white and orange xhimas from Illinois were identical, indicating that this was an effective control. However, the lower texture ratings for the Illinois varieties overall in comparison to the local variety may be a result of the milling process, as the Illinois varieties were milled with the pericarp while the local was milled without pericarp days two through four. When the local variety was milled with pericarp on for day one, its texture score was lower than that of the Illinois varieties, indicating that the presence/absence of the pericarp does affect the texture of the xhima. Significance testing reveals that the taste, texture, and appearance scores of the local maize on day one were significantly lower than on days two through four.

Table 1.1 - Descriptive Statistical Analysis of Survey Data (n=249)

	Minimum	Maximum	Mean	Median	Mode	Std. Dev.
Gender(a)	0	1	0.64	1	1	0.48
Age	14	70	35.70	33	22	13.20
# people in house	1	20	6.56	6	6	3.14
# children <12	0	9	2.21	2	2	1.73
# children <3	0	8	0.69	0	0	1.06
Shops?(b)	0	1	0.60	1	1	0.49
Cooks?(b)	0	1	0.40	0	0	0.49
Month – Meat(c)	0	1	0.72	1	1	0.45
Month – Eggs(c)	0	1	0.43	0	0	0.50
Month – Xhima(c)	0	1	0.89	1	1	0.31
Month – Sweet potato(c)	0	1	0.46	0	0	0.50
Month – Vegetables(c)	0	1	0.97	1	1	0.17
Month – Fish(c)	0	1	0.73	1	1	0.44
Week – Meat(d)	0	7	1.11	1	0	1.44
Week – Eggs(d)	0	7	0.90	0	0	1.66
Week – Xhima(d)	0	7	2.86	2	1	2.51
Week - Sweet potato(d)	0	7	1.01	0	0	1.83
Week – Vegetables(d)	0	7	4.17	4	7	2.38
Week – Fish(d)	0	7	2.25	2	0	2.27
White Local – Taste(e)	1	5	3.96	4	5	1.19
White Local – Texture(e)	1	5	4.06	4	5	1.10
White Local – Appearance(e)	1	5	4.39	5	5	0.91
White Local – Aroma(e)	1	5	3.66	4	5	1.20
White IL – Taste(e)	1	5	3.59	4	4	1.14
White IL – Texture(e)	1	5	3.54	4	4	1.04
White IL – Appearance(e)	1	5	3.34	4	4	1.04
White IL – Aroma(e)	1	5	3.63	4	4	1.08
Orange IL – Taste(e)	1	5	3.72	4	5	1.34
Orange IL – Texture(e)	1	5	3.54	4	5	1.25
Orange IL – Appearance(e)	1	5	3.59	4	4	1.19
Orange IL – Aroma(e)	1	5	4.02	5	5	1.29
Orange Offered Relative to White	0.5	3	1.36	1.5	0.5	0.72
Trade for Orange?(b)	0	1	0.46	0	0	0.50
Trade for Tomato?(b)	0	1	0.02	0	0	0.14

(a): 0=male, 1=female
(b): 0=no, 1=yes
(c): 0=not consumed in previous month, 1=consumed in previous month
(d): # of days in the last week consumed
(e): 1=very bad, 2=bad, 3=okay, 4=good, 5=very good

The overall acceptance rate of orange maize meal at all trade ratios was just over 46%. The distribution of offers shown in Table 1.2 reveals that acceptance generally increased with an increase in the amount of orange offered – 41% at 1:2, 37% at 1:1, 51% at 3:2, 60% at 2:1, and 39% at 3:1. The difference in acceptance between rates of 1:1 and 2:1 was significant (p=0.009), as well as the difference in acceptance between rates of 1:1 and 3:2 (p=0.068). The decrease in acceptance at the highest ratio may be due to a belief that the orange must be inferior if we are willing to offer such a high amount, or possibly that the individuals given this offer strongly disliked the orange and were unwilling to take it at any ratio.

Table 1.3 presents some indication of the factors affecting acceptance of orange maize meal. Men were less willing to trade for the orange maize meal than were women, though this difference was not statistically significant. Those who did the shopping were slightly less likely to accept the trade for orange than the non-shoppers, although again, the difference is not significant.  Acceptance seems to correspond to the sensory perception data, as those who ranked the taste of orange at 1 or 2 (poor or very poor) only accepted the trade 19% of the time, while those who ranked it at 4 or 5 (good or very good) accepted the trade 64% of the time. The difference in these two acceptance rates is significant (p=6.23E-09).  Also, those who ranked the orange taste higher than the local white accepted the trade 71% of the time, while those who ranked the taste of the orange lower than the local white took the trade only 25% of the time. Again, this difference was significant (p=1.12E-12)

Table 1.3 - Acceptability of Orange Maize Flour by Consumer Characteristic

Characteristic	N	Acceptance Rate
Male	86	0.40
Female	148	0.50
Non-shopper	91	0.47
Shopper	131	0.44
Orange taste 1-2	48	0.19
Orange taste 4-5	146	0.64
Orange taste>Local taste	104	0.71
Orange taste<Local taste	128	0.25
Frequent meat/fish consumer (>4 times/week)	75	0.37
Infrequent meat/fish consumer (<4 times/week)	140	0.54

In order to get an idea of each participant’s income, data on their dietary diversity was collected. These data also indicate the likelihood that a respondent’s diet lacked adequate vitamin A. Figure 1.1 presents the share of participants who consumed each food a given number of days of the week. Most participants were not frequently consuming meat or eggs, high value products, indicating that the average income of the participants was low.  As seen in Table 5.3, the acceptance of the orange maize flour was significantly higher among infrequent consumers of meat and fish than among those who consumed meat and fish frequently (54% versus 37%, p=0.012). This result is encouraging, as those who do not frequently consume meat or fish products are more likely to suffer from vitamin A deficiency.

Figure 1.1 – Frequency of Participants’ Consumption of Selected Foods

When comparing sensory perception data across the four days of the study, we see some evidence that the use of local maize with pericarp may have affected the results on day one of the experiment (Table 1.4). For example, the orange maize was better-received (sensory perception and trading data) on day 1 in Xiquelene than on day 3 in Xipamanine. We see additional differences in average age and food consumption patterns that may provide some explanation of the difference in preferences upon further analysis. Most importantly, participants rated the taste, texture and appearance of the local maize significantly lower on day one, when the pericarp was included, than on the other days.

Table 1.4 - Comparison of Data Across Days and Markets

	DAY1	DAY2	DAY3	DAY4	Xiquelene	Xipamanine
White Local – Taste(a)	3.27**	4.00	4.27	4.12	3.70	4.18
White Local – Texture(a)	3.35**	4.14	4.18	4.36	3.81	4.27
White Local – Appearance(a)	3.69**	4.46	4.52	4.70	4.14	4.60
White Local – Aroma(a)	3.60	3.70	3.56	3.78	3.66	3.66
White IL – Taste(a)	3.88	3.76	3.19	3.59	3.81	3.41
White IL – Texture(a)	3.65	3.58	3.37	3.57	3.61	3.49
White IL – Appearance(a)	4.02**	3.18	3.03	3.30	3.53	3.18
White IL – Aroma(a)	4.04**	3.67	3.33	3.57	3.83	3.46
Orange IL – Taste(a)	4.21**	3.84	2.97	3.93	3.99	3.48
Orange IL – Texture(a)	4.17**	3.74	2.72	3.62	3.92	3.21
Orange IL – Appearance(a)	4.23**	3.79	2.86	3.61	3.97	3.26
Orange IL – Aroma(a)	4.31	4.15	3.52	4.15	4.22	3.86
Age	26.88**	33.53	40.23	39.41	30.73	39.81
# in house	5.90	7.05	6.23	6.77	6.57	6.77
Week – Meat(b)	0.83	0.87	1.55	1.16	0.85	1.34
Week – Eggs(b)	0.80	0.88	1.05	0.84	0.85	0.94
Week – Xhima(b)	2.68	3.16	2.38	3.14	2.97	2.78
Week – Sweet Potato(b)	0.91	0.85	0.80	1.44	0.88	1.13
Week – Vegetables(b)	3.63	3.94	4.16	4.83	3.81	4.51
Week – Fish(b)	-	2.33	3.24	2.99	2.33	3.11
Trade for Orange?(c)	0.54	0.51	0.27	0.55	0.52	0.40
Sample Size	48	67	64	70	115	134

NOTE: ** indicates significantly different from days 2-4 at α=0.01
(a): 0=very bad, 1=bad, 3=okay, 4=good, 5=very good
(b): # of days in last week consumed
(c): 0=no, 1=yes

Based on significance testing results, data from day one was dropped from the sample and the calculations reported above were repeated.  Use of the smaller sample did not greatly affect the descriptive statistics, so are not presented in this summary.  The average sensory perception results are somewhat higher for local maize and the average results for orange maize are somewhat lower.  All differences that were statistically significant using all data remained so after the exclusion of the day one data. Consequently, all subsequent data presented will consider only day two through four of the experiment.

Results – Multivariate Analysis

Logarithmic and quadratic functional forms were tested using a variety of dietary diversity calculations. Based on these specification tests, a quadratic functional form using dietary diversity relative to animal products was selected for the estimation.  Raw estimation results from both a full and reduced quadratic form are presented in Table 1.5.  As the table indicates, results showed that household size, the presence of small children, indicators or dietary diversity, perceived taste, and the exchange offered were all statistically significant determinants of acceptance.

Acceptance of the offer to exchange white maize meal for orange was significantly greater for people who reported lower consumption of animal products.  Consumption of such products can be understood as a proxy for income.  In Mozambique, a 1% decrease in dietary diversity has been associated with a 1% decrease in income (Hoddinott and Yohannes, 2002). The relationship between income and consumption of animal products is probably stronger.  Additionally, low consumption of animal products may suggest the existence of VAD.  Those who consume fewer animal products are more likely to suffer from VAD.  Calculating the elasticity at the mean values, results suggest that a 10% decline in the frequency of consumption of animal products is associated with a 2% increase in the probability of acceptance.  This suggests that orange maize could be a self-targeting nutritional intervention since those consumers who are more likely to be VAD are also more likely to accept the biofortified food.

Table 1.5 - Coefficients of Quadratic Probit Model Using Dietary Diversity of Animal Products

	Full Quadratic	Reduced Quad.
Dummy Day2	-0.4594551 (0.152)	-0.4019937 (0.202)
Dummy Day3	-1.197325*** (0.000)	-1.067751*** (0.001)
Gender	0.2955742 (0.393)	0.2738662 (0.409)
Age	-0.0245781 (0.645)	-0.0009924 (0.915)
Age2	0.0003046 (0.630)
Household Size	-0.2170999* (0.066)	-0.1841587 (0.117)
Household Size2	0.0095738 (0.109)	0.0073466 (0.222)
Kids Under Age 3	0.6944379** (0.018)	0.3704981*** (0.006)
Kids Under Age 3]2	-0.1225933 (0.148)
Dummy Shops	-0.4936097 (0.106)	-0.5135093* (0.081)
Dummy Cooks	-0.2834813 (0.337)	-0.2457134 (0.393)
Xhima/Week	-0.1697606 (0.370)	-0.0208551 (0.673)
Xhima/Week2	0.0210887 (0.414)
Sweet Potato/Week	0.0144962 (0.945)
Sweet Potato/Week2	-0.0072955 (0.823)	-0.0080863 (0.386)
Vegetables/Week	0.4158936* (0.076)	0.4046292* (0.071)
Vegetables/Week2	-0.0449299 (0.119)	-0.0436089 (0.110)
Dietary Diversity Animal(a)	-3.966624** (0.035)	-3.829556** (0.038)
Dietary Diversity Animal2	4.629112** (0.035)	4.461863** (0.048)
Mozambique Taste	0.3963107 (0.566)	0.0052091 (0.969)
Mozambique Taste2	-0.0597803 (0.557)
Orange Taste	-1.34237*** (0.005)	-1.248581*** (0.009)
Orange Taste2	0.2827445*** (0.000)	0.2746185*** (0.000)
Offer	0.9956328* (0.100)	0.8632118 (0.154)
Offer2	-0.2519504 (0.175)	-0.2092781 (0.254)
Constant	1.110676 (0.595)	0.9035353 (0.525)

(a): DDanimal = ∑(Meat/Wk, Eggs/Wk, Fish/Wk)/21
NOTE: * = Significant at p<0.1; ** = Significant at p<0.05; *** = Significant at p<0.01

Figure 1.2 presents the estimated probability of accepting orange maize by the frequency with which animal products are consumed.  The results based on mean values imply a price discount of 38% for the orange maize, while the results based on median values are calculated from a 50% discount on the orange price.  Results using the median values but removing the price discount are also presented. As the figure indicates, we find a probability of acceptance of over 70% for those who consume animal products one time or less per week. Those who consume animal products daily, and are therefore unlikely to be vitamin A deficient, are unlikely to accept orange maize.

Figure 1.2 - Probability of Acceptance by Frequency

VAD is especially problematic and prevalent among children.  Therefore, it is especially important to determine whether households with children are likely to accept orange maize.  Experience with the orange-fleshed sweet potato campaign suggests that mothers may be more inclined to use biofortified foods for their children than for themselves.  Again, our data suggests that acceptance is more likely when there are children in the household (Figure 1.3).  The probability of acceptance rises to well over 70% when there are 2 or more children under three years of age in the household.

Figure 1.3 - Probability of Acceptance by Number of Children Under Age 3 in Household

Figure 1.4 presents the impact of a price discount to orange maize meal on the probability of acceptance.  At the mean values, a 10% decrease in the price of orange maize relative to white results in an increase in the probability of acceptance of about 1.5 percentage points.  Based on mean values of control variables, a discount on the order of 25% may be required to make acceptance likely. We see modest sensitivity to further price discounts.  A price premium on orange maize would make acceptance unlikely.

Figure 1.4 - Probability of Acceptance by Discount on Orange Meal Relative to White Meal

Major Findings & Limitations of the Study

It appears that orange maize is an acceptable product to many consumers in urban Maputo, Mozambique, particularly at a modest price discount. The results also suggest that vitamin A-biofortified maize may be a self-targeting nutritional intervention, as those who are most likely to suffer from VAD were the most likely to accept the trade for orange maize meal. These groups include households with young children, and households that do not frequently consume animal products.

This experiment revealed some aspects of the research that could be improved in future studies. It would be more appropriate to have all maize for the study milled at the same location in the same way, particularly with respect to pericarp presence/absence.

With respect to the survey, the addition of a question regarding educational level or literacy and further questions regarding dietary diversity, particularly the inclusion of other starchy staples such as rice, would improve the data.

In addition, including a more varied set of locations may allow testing of a wider variety of consumers, particularly those who consume maize more frequently than the consumers encountered in the urban marketplaces in Maputo.

Special thanks to the following for their help with this research:

• Dr. Maria Andrade
• Dr. Jan Lowe
• Pedro Chauque
• Constantino Senete

Return to previous page.