Gender-based market constraints to informal fish retailing: Evidence from analysis of variance and linear regression

Over the last decade, Egypt’s aquaculture sector has expanded rapidly, which has contributed substantially to per capita fish supply, and the growth of domestic fish markets and employment across the aquaculture value chain. Despite the growing importance of aquaculture sector in Egyptian labour force, only a few studies have explored the livelihoods of Egypt’s women and men fish retailers. Even fewer studies have examined gender-based market constraints experienced by these informal fish retailers. This study uses sex-disaggregated data collected in 2013 in three governorates of Lower Egypt to examine the economic and social constraints to scale of enterprises between women (n = 162) and men informal fish retailers (n = 183). Specifically, we employ linear regression method to determine the correlates of enterprise performance. We found that both women and men retailers in the informal fish market earn low profits and face livelihood insecurities. However, women’s enterprise performance is significantly lower than that of men even after controlling for individual socio-economic and retailing characteristics. Specifically, the burden of unpaid household work and lack of support therein impedes women’s ability to generate higher revenues. These findings strengthen the argument for investing in understanding how gender norms and attitudes affect livelihood options and outcomes. This leads to recommendations on gender-responsive interventions that engage with both men and women and enhance the bargaining power and collective voice of fish retailers.

Aquaculture production in Egypt has grown steadily over recent years, which has contributed to a rise 49 in domestic fish supply from 803,211mt in 2000 to 1,947,087 in 20161,947,087 in (GAFRD, 2016Murphy et al, 50 under review). This has also stimulated substantial informal employment growth in local markets 51 (Hebisha and Faithi, 2014;Nasr-Allah et al., 2014;Soliman and Yacout, 2016;Wally, 2016). 52 According to recent review by Nasr-Allah et al (forthcominng), the aquaculture value chain provides 53 approximately employment for 19 full time jobs (FTE) for every 100 tonnes of fish produced to 54 formal actors and predominantly informal actors. Notably, highest employment rates were reported in 55 retail nodes [9.7 FTE/100mt] (ibid: p. 24), where substantial numbers of informal women traders 56 sustain livelihoods as unlicensed street vendors or traders in open markets [5]. 57 In addition, food market studies report that farmed fish currently offers cheapest animal-source 58 protein to Egyptian consumers and highlighted aquaculture as a strategic food sector to food and 59 nutrition security in the country [2,6]. Linking to the above on livelihoods, reports also indicate low-60 income consumers tend to buy more affordable fish products from informal women retailers, whose 61 enterprises were found operating in lower value markets and were characterised by smaller scale of 62 operations (Eltholth et al., 2015;Kantor and Kruijssen, 2014). 63 Global reviews have placed Egypt 11th lowest in terms of women's labour force participation, and 3 rd 64 lowest within the MENA region (Korotayev et al., 2015). In the MENA region, research has shed 65 light on the challenging working conditions faced by resource-poor women in the market that 66 contribute to lower levels of female labour force participation. Reports pointed to issues relating to 67 their limited access to market extension and financial services, the risk of personal security and their 68 limited freedom of mobility [7][8][9][10][11][12][13]. National surveys have also highlighted gender roles and norms at 69 the household level that may limit women's entry to the market. According to United Nations Women 70 and Promundo's 2018 study of household gender relations in Egypt, women's economic 71 disempowerment can stem from gendered division of household labour and related time burdens, 72 restrictions on women's mobility in public spaces at different times of the day, and their limited 73 access to and/or control over income, assets and credit (El Feki, Heilman, and Barker, 2017). The 74 report found only 31% of men believe women should have the same right to work outside the home as 75 their husband, while majority of both male and female respondents (80%) believe women should ask 76 permission before leaving the house. Meanwhile, greater majority of men (98%) and women (85%) 77 believe that household chores and childcare is sole responsibility of mothers. 78 There has been considerable research on women's empowerment and livelihood development 79 approaches in agricultural settings (Badstue et al., 2015;Hillenbrand et al., 2014;Kabeer, 1996;80 Malapit et al., 2014;Seymour et al., 2016;Sraboni et al., 2014). There is also substantial literature on 81 which is reflected in price differentials between morning and evening sales. This suggests that those 108 with greater access to marketplaces, service providers and basic resources like ice, sunshade, 109 weighing scales are at an advantage to those without. 110 In addition, traders without reliable supply or credit contracts may experience unfair price 111 fluctuations, poor quality of stock selection, restrictive lending terms or security risks when travelling 112 to wholesaler or accessing supply markets (Internaional Labour Organization, 2013;Kabeer, 2008;113 Kantor, 2005113 Kantor, , 2003. In Egypt, evidence has shown that there are significant differences between 114 market relations of women and men retailers, where more women retailers have reported paying 115 higher prices per kilogram to creditors and being given shorter repayment periods [20]. This suggests 116 that those who have greater access to or power in financial markets have an advantage over those who 117 do not, which in turn is likely to lead to more bargaining powers in upstream and downstream 118 transactions for the first group. In addition, there is pressure to avoid daily spoilages and clear stock 119 before the day's market closure, as there is generally a lack of cold chain facilities, which adds further 120 pressure on retailers' bargaining position. This is most visibly reflected in the daily reduction of fish 121 prices at the end of the sales day, most visible among women retailers [20]. 122 While acknowledging the backdrop of formal structural factors, such as policies and market 123 regulations that directly shape trading operations or locations, we focus our theoretical framework, on 124 the gendered social or market relations that are significant to retailers' scale of retailing enterprise. 125 Towards this, we apply perspectives of the research on informal trade markets and the work on gender 126 in value chain development. The sociology of informal trade markets has shed important light on the 127 economic significance of 'social relations' within which commodity exchanges are embedded [21-128 24]. Such approaches share much with the 'social relations approach' applied in the scholarship of 129 women's empowerment and gendered value chain development [25][26][27][28]. 130 Specifically, we consider here a 'social relations approach' that focuses on gender dynamics at 131 household and market levels where resources, labour and exchange relations, market access and 132 freedom of mobility are negotiated between women and men (Badstue et al., 2015;Kabeer et al., 133 1996;Kantor, 2005;Patil, 2013a). In particular, we focus on the elements of time use and dynamics of 134 household labour. We recognize the methodological concerns of these authors regarding the need to 135 consider intersectionality of sex with other sociodemographic and economic characteristics, by cross-136 referencing our data with social variables, namely age, education, retailing experience, marital status, 137 household size, time use data and location [29][30][31][32][33]. 138 included in the study. In Kafr el-Sheikh, there were few men fish retailers. Therefore, the total number 151 of interviewed men was only 49 instead of the desired 70. 152

Methods and methods
A structured questionnaire was used to collect data on retailers' household characteristics and 153 dynamics in retailing operations and market relations, as well as prices, margins and profits. The 154 following tables provide description of variables used in our descriptive and empirical analyses 155 (Tables 1 -3). 156 The questionnaire was translated from English into Arabic and field tested in two governorates before 165 being finalized for field implementation. Statistical analysis was performed using v.13.0 of STATA 166 software. Mean tests were conducted across sub-samples disaggregated by sex and location. Analysis 167 of variance was employed using one-way ANOVA t-tests. Linear regression models with robust 168 estimates were employed to determine the direction and intensity of effects of predictor variables on 169 and scale of enterprise (total value of daily sales). 170 171

Retailer's household characteristics and dynamics
173 Table 4 shows basic descriptive statistics of retailers surveyed. Mean household size was 4.18 ± 1.61. 174 In the overall sample, women retailers were older (37.0 ± 8.2) than men retailers (33.6 ± 10.0) [F(1, 175 343) = 3.055, p = 0.002]. Retailers educational years also differed significantly by sex [F(1, 3430 = 176 6.208, p = 0.000], with men retailers reporting a higher number of educational years (8.7 ± 7.0) than 177 women retailers (4.3 ± 6.1). Considering retailers and spouses together, both men and women in 178 households of male retailers were younger than men and women in households of female retailers. We 179 also found that women in the male retailers' households received higher education than women in the 180 female retailers' households, although this difference was not significant for all governorates. 181 Conversely, men in male retailers' households were more highly educated than men in the female 182 retailers' households and this difference was statistically significant in all governorates. 183 To examine wealth status of women and men retailers' households, we collected information on 14 184 durable household assets, including the possession of colour television, satellite dish, landline 185 telephone, mobile phone, electric fan, air conditioner, refrigerator, freezer, automatic washing 186 machine, manual washing machine, bicycle or motorcycle, animal drawn cart, larger vehicle 187 (car/van/truck/tuk-tuk) and agricultural land. Weighting these evenly, we scored households 188 accordingly along an 'asset index'. Across the total sample, we observed a significant effect of sex 189 [F(1, 343) = 3.965, p = 0.000], with higher asset index reported by male retailer households (5.84 ± 190 1.18) compared to female retailer households (5.38 ± 0.93). This differences was also statistically 191 significant in Beheira [F(1, 132) = 3.638, p = 0.000]. 192 To assess the retailer's time burdens, questions were asked regarding family labour roles and hours 193 spent in different paid and unpaid activities. Respondents were first asked to provide estimates of 194 hours they spent in retailing activities, and unpaid household work and childcare. In Table 4 below, 195 results indicated that women retailers in Sharkhia spent significantly less time in retail activities (7.8 ± 196 1.7) than men did (8.5 ± 1. reporting such conflicts. This difference was also statistically significant in all governorates. 205 Regarding household labour roles, retailers were asked if they had family help during household 206 tasks. Across the sample, a significantly higher share of men reported having family help in the home 207 (85.8%) than women did (34.0%) [F(1, 343) = 11.623, p = 0.000]. This difference was also 208 statistically significant in all governorates. 209

210
To assess retailers' agency over household finance, a questions was asked about intra-household 211 decision-making regarding the use of income and own earnings. Regarding intra-household decisions 212 over use of own earnings, retailers were asked whose decisions prevails during disagreement. A 213 significantly higher percentage of men reported 'self' as opposed to their spouse as the prevailing

218
In Table 5 below, results are presented of percentages of retailers who were operating in different 219 market areas, between trading in government registered formal marketplaces, trading illegally on the 220 street, from their homestead doorstep. Across the aggregate sample, a higher share of men retailers 221 reported access to formal 'marketplaces' (91.3%) than women retailers did (85.2%). In contrast, 222 significantly more women reported selling fish on their 'doorstep' (5.6%) than men did (

266
In Table 6, data on previous day of retailers' fish purchases, sales, and variable and fixed costs are 267 presented, in addition to gross margins for each species and daily gross profits for total stock. In 268 Sharkhia, Beheira, and for the overall sample, gross margin of tilapia per kilogram (i.e. price sold 269 minus price bought) was significantly higher for women than for men. Across locations, highest mean 270 margin for a kilogram of tilapia sold was in Beheira (E£1.65 ± 2.88) and lowest in Sharkhia (E£0.

Empirical analysis of social and market-based factors to retailing
283 performance 284 The current study investigated whether or not there were differences between male and female fish 285 retailers' enterprise performance after controlling for individual, household, regional, business and 286 entrepreneurial characteristics. We also examined the effects of gender-based constraints on retail 287 performance, namely variables informing gender dynamics within retailers' households and those 288 considered significant to market operations and exchange relations. 289 290 In our model, we examined how household gender dynamics, proxied by time-use in fish retail and 291 unpaid care work, and family help with household work affected men's and women's scale of fish 292 retailing differently, if at all. In Egypt, which is a strong patriarchal society, we expect gender-based 293 constraints and the social norms that underpin them to affect men and women differently. The model 294 also evaluated the effects of market-based factors on retailing performance, by examining the 295 significance of trade relations, access to supply markets and enterprise characteristics including 296 variety of species sales, fixed and variable costs, all which transactions is considered strongly shaped 297 by gendered power hierarchies and the bargaining power that lays out between them [34]. 298 The regression model, based on the ordinary least squares (OLS) method, was specified as follows: 299 where Yi is the performance indicator, (scale of operation) of retailer i, Si is the sex of the retailer; 301 GBC is the vector of gender-based constraints like time spent in fish retail, household care work, 302 availability of help with household chores, interaction of household help with the sex variable; RC is 303 the vector of retailers' characteristics like age, education and years of experience in fish retail; TC is 304 the vector of retail related variables like fixed and variable costs, number of fish species traded, and 305 certainty of supply relations (buying from the same supplier and location of supply transactions); CO 306 is the vector of control variables including regional dummies (R1 and R2); and µi is the error term. 307 The model specifies interaction terms between the proxies for gendered constraints and sex of the 308 retailer primarily because we hypothesize that the effects of these constraints are very different for 309 men and women due to patriarchal norms. We analyze this model for the dependent variable: scale of 310 operation. We measured the scale of enterprise as a natural logarithm of total value of all fish sales. 311 The OLS estimates of the model are presented in Table 7. Results revealed positive and highly significant relationships with between male gender, higher 322 socioeconomic status of respondent, time spent in retailing and household help with preferences for 323 tilapia larger sizes at the 1% levels (Table 7). We observed a highly significant effect of sex on scale 324 of operation, with model predicting women retailers (1) compared to men retailers (0)  The study's descriptive analysis can shed further light on these results (Table 4). Across the aggregate 337 sample, 62% of retailers had one or more household members who are able to shoulder the 338 responsibilities of household chores and care provision. However, between men and women retailers, 339 the difference was stark: 85% vs. 34%, respectively. This result implies that women, who rely on 340 household members to help in household work will likely enjoy larger scales of operation. The 341 interactions of these phenomenon with different aspects of retailing performance require further 342 exploration beyond the scope of the current study. However, in the final Discussion section, literature 343 discussing these dynamics in other economic spheres are discussed in relation to these finding 344 invitation to future research. 345 346 Not surprisingly, both independent variables 'number of species sold' and 'operational costs', 347 expressed significantly positive associations with the dependent (p < 0.001). If we refer back to 348 descriptive results on retailer's operations (Table 5), we find statistically significant differences 349 between 'fixed and variable costs' of women and men retailers. Across the aggregate sample and 350 within two subsample governorates, men reported higher operational costs. In addition, proportionally 351 more men than women dealt in higher value species, such as mullet, while men also reported greater 352 volumes of daily sales across the sample.  Significant p-values are reported at levels of <0.10, <0.05, <0.01, <0.001 as *, **, ***, ****.

366
Feminist economics and informal market literature suggest two different categories of factors that 367 affect the performance of informal retail businesses -retailers' individual and enterprise 368 characteristics [21][22][23][24]31,35,36]. Evidence from the literature also shows that after controlling for 369 these factors, retailer's sex is highly significant in determining business performance and market 370 transactions [18,19,[25][26][27][28][37][38][39][40][41]. Our findings confirm that the sex of retailer is significant to 371 outcomes of retailing performance in terms scale of enterprise (Table 7), and signal that this relates to 372 gender dynamics in the households and public market places or urban Egypt. 373 374 At the household level, we observed several such factors, in particular gender labour relations and 375 time burdens. As expected, results indicated that as time spent in retailing increased, retailers' scale of 376 enterprise increased. Conversely, more time spent in household care and unpaid activities increased, 377 retailers' scale of enterprise decreased, Given that gendered household labour relations mean that 378 domestic time burdens fall disproportionately on women, these gendered relations also 379 disproportionately negatively impact on the business performance of women compared to men. Our 380 results confirm similar findings to those reported by Arora's study of peasant households in 381 Mozambique (Arora, 2015;Arora and Rada, 2016). That is, labour allocation of unpaid activities such 382 as child care and domestic chores produce greater time burdens on women. As well as affecting 383 individual economic activities, this likely influences overall household productivity in terms of 384 livelihood security and gross profitability. 385 386 At the market scale, our findings also indicated that the availability of supply agreements were 387 significant to trade performance (Table 7). These findings echo similar discussions by Sabarwal and 388 Terrell (2008), in which they highlighted capital constraints and limited access to financial markets as 389 key barriers to women's enterprise growth. It also reflects on findings above which point to the 390 market-based dynamics affected gendered bargaining power and the gender-differences identified in 391 scale of daily sales by retailers. This reflects also on the differences observed between reported 392 operational costs of women and men. Across the sample, men reported higher costs, while results on 393 species also indicated men were predominantly involved in the sale of higher value products (Table  394 5). Considering the above studies and current findings on gendered bargaining power, including 395 differential access to supply and financial markets, we find important spaces of gender inequality 396 worthy of investigation that evidently affect gender differences in enterprise outcomes. 397 398 Considering the significance of supply relations, results also indicated that retailers relying point of 399 sale wholesalers suffered poorer scale of enterprises. This raises some important questions for future 400 research, which calls for more detailed analysis of supply markets and the transactions between 401 retailers and wholesalers, both within their local markets and within regional wholesale depots. 402 Evidently, there are important differences between such spaces regarding women's bargaining power 403 in these male dominated marketplaces. In Egypt, several studies have highlighted varied constraints 404 facing women's retailing enterprises, namely the lack of supply contracts, restrictive lending terms 405 and unfair price and trade negotiations, all of which stem in part from vulnerable credit relations or 406 trade agreements (Farnworth et al., 2015;Kruijssen et al., 2018). These combined factors add pressure 407 to the risk of daily spoilages and the need to clear daily stock before market close, therefore placing 408 further pressure on retailers' bargaining power, all of which were made evident by changing sales 409 prices and differences in profit margins throughout the day (Table 6). 410 411 Further progress in understanding gender-based market constraints should be considered in future 412 research through more detailed investigation of additional time-use variables and intra-household 413 decision-making between dual respondents between dual-adult and female-headed households. 414 Literature on the gendered dynamics of intra-household decision-making have examined their 415 significance to household economics, agriculture productivity as well as off-farm paid market 416 activities [44][45][46][47]. While our survey collected data on household structure among both single and 417 dual-adult households, the percentage of single-adult households and the sampling only of lead 418 respondents did not allow for intra-household analysis at the depth required. Therefore, we 419 recommend future research to adopt more purposive sampling of dual-adult respondents that allows 420 for robust investigation of household decision-making processes. 421 422 Finally, the dynamics identified as significant in this study imply the need for more detailed analysis 423 of how social institutions and gender norms in both the marketplace and within the household affect 424 livelihood options and outcomes of women retailers. Ignoring such gender dynamics may impinge on 425 enterprise performance, scale and profitability of retailing operations among women and thus 426 contribute to perpetuating the cycle of poverty amongst urban poor. Moreover, these dynamics may 427 limit the ability of women retailers to affordably supply the fish needed to meet growing demand in 428 these critical markets for poor, food insecure, consumers. To address such market constraints and 429 improve the agency and bargaining power of female economic actors, value chain interventions 430 require approaches that engage with both men and women retailers from within household and across 431 the community. We offer two main recommendations: 1) interventions of gender-responsive value 432 chain interventions retail should consider the findings of such current studies to propose policy 433 strategies that specifically address enabling factors such as sharing of household labour in paid and 434 unpaid labour and the strengthening of bargaining power of women retailers in key spaces of 435 transactions and access to physical supply and financial markets. 3) key to effective implementation 436 of gender-responsive market intervention, we recommend that initiatives need to conducted with both 437 women and men retailers and community stakeholders that ensure women 438 ensure women maintain a voice in mixed sex groups, 2) , 439 440