This project is studying technology spillovers among soccer-ball producers in Sialkot, Pakistan, the world center for the production of hand-stitched soccer balls. Convincing estimates of technology spillovers among manufacturing firms have been elusive, for two main reasons: First, measurement: the standard approach is to measure the overall level of technology in a firm using total factor productivity, the residual from a regression of sales on input expenditures, which may be confounded by mark-ups or heterogeneity in input or output quality. Second, causality: technologies are generally not randomly assigned, and it is often not clear whether to interpret two network-linked firms adopting the same technology as a spillover or simply correlation among unobservable characteristics.
To address these issues, our research team invented a new technology that reduces production costs and randomly assigned the technology to a subset of football firms in Sialkot. The production process is similar across the 141 firms active in the sector. The majority of footballs are produced using the standard “buckyball” design of 20 hexagons and 12 pentagons. There are four main phases to production. First, layers of cloth are glued to the sheets of rexine (artificial leather) using a latex-based substance, a process known as lamination. The rexine, cloth and latex are the most expensive inputs and account for approximately 55% of the total cost of each soccer ball (and more if high-quality imported rexine is used in place of Pakistani rexine). Second, a skilled cutter then cuts hexagonal or pentagonal panels using a metal die and a hydraulic press, moving the die and positioning the press by hand to make each cut. Third, logos or other insignia are then printed on the panels. Fourth, the panels are stitched together, and an inflatable bladder is inserted.
Our research team invented a new technology that reduces production costs and randomly assigned the technology to a subset of football firms
The new technology that we invented is a new layout for cutting pentagons from the laminated rexine sheets and a metal die that implements the layout. A key challenge in the production process is to minimize waste of the rexine/cloth material. In the case of the hexagons, it is straightforward to minimize waste since hexagons tessellate (i.e. completely cover a plane). But pentagonal panels do not tessellate. Prior to our study, cutters in all firms in the cluster were imitating the hexagon cutting pattern when cutting pentagons. This was not optimal: approximately 24% of laminated rexine was being wasted when cutting pentagons, as opposed to about 8% waste for hexagons. Our new technology, which uses a finding from the field of computational geometry about the best-known packing of pentagons in a plane, allows firms to improve the yield of pentagons per rexine sheet from around 250 pentagons per sheet to 272 pentagons, an increase in yield of approximately 9%. Although firms are free to modify the die design (for instance, using a die that cuts two pentagons at a time rather than the four-pentagon die we originally designed), it is straightforward to observe whether firms are using our idea because on existing dies adjacent pentagons share a full edge whereas in our design the pentagons are offset, and share only half an edge.
We stratified the 141 firms in the sector into four strata by firm size (measured by output of balls) and divided them into three treatment groups: tech drop, cash drop, and control. To the tech drop firms, we gave a blueprint of the new layout and a die to implement the layout; since the size of hexagons and pentagons vary slightly across firms, we also offered to pay for a trade-in of the die we provided for a similar die made to their precise specifications. To the cash drop firm we gave cash equivalent to the amount that we paid for each of the dies we provided to the tech drop firms, approximately 30,000 Rupees (US$300) each. To the control group firms we gave nothing. We also treated the firms that did not respond to our baseline survey as a separate stratum and allocated them to the three treatment groups as well.
We began distributing the technology in May 2012 and are following the diffusion process closely.
We expect to conduct the endline survey and have final results by mid-2014.
Investigators: CDEP Affiliate Amit Khandelwal, CDEP Director Eric Verhoogen, David Atkin of Yale University and Azam Chaudhry and Shamyla Chaudry of the Lahore School of Economics.