In Jalal-Abad, in late January 2019, the focus group ran for almost three hours. We had budgeted ninety minutes. The reason it overran was a single answer to a single question — “how is academic IT teaching going at your universities?” — that one of the participants gave, and then everyone else in the room confirmed.
The answer was that there were no academic teaching staff in the city available to teach IT subjects at university level. Not a shortage. Not a few. None. The budget lines for IT lecturers existed at the relevant faculties. The student places existed. The buildings, in some cases, were new. What was missing was the people. The participants explained how the workaround functioned: a senior practitioner from a local firm would come in for a guest block; recent graduates of the universities themselves would teach the year below; someone’s brother who had returned briefly from Bishkek would do a semester. The substitution kept going. It produced graduates with certificates that, in the labour market the rest of the report was about to describe, did not function as evidence of skill.
That moment, more than any survey datum, set the frame for the cross-sector chapter of the report. We had gone to Jalal-Abad expecting to find a smaller, slower version of what we had seen in Bishkek. What we found was a structural absence — and it repeated, in different forms, in every other sub-sector we then went on to describe. The chapter we wrote up wasn’t a list of problems unique to each sub-sector. It was a list of eight problems that every sub-sector named, in the same words, in the same order of severity.
The Information and Communications sector, in 2017 numbers
Before the eight problems, the picture the chapter was set against.
The numbers in the table above are the National Statistics Committee’s “Information and Communications” sector — the broadest official cut of the data, which includes telecoms, broadcasting, and ICT alongside the software-development core. The doubling of revenue between 2016 and 2017 — from 241M KGS to 518M KGS — is the figure most people remembered from the report when we presented it. It is also the figure that needs the most context. The base was small. Three percent of GDP is a wedge, not a pillar. And the growth was concentrated, both geographically (88% of registered firms in Bishkek and Chui oblast) and within the High Technology Park, whose residents went from 7–8% of total sector revenue in 2014 to 29% in 2017.
The multiplier, and what it implied for jobs
The report’s most-quoted chart, the one that got pulled into other people’s slide decks more than anything else we produced, was the multiplier estimate.
The ratio underneath it was U.S. labour-market evidence — the Economic Policy Institute’s employment multiplier study — applied to a Kyrgyz baseline of 7,400 IT roles in IT firms. The point of the chart was not the ratio. The point was that the conversation about IT jobs in Kyrgyzstan, in 2019, was being conducted as though “IT jobs” meant programmers. The multiplier said: every developer hired creates roughly two non-technical roles around them, in their own firm, in adjacent firms, in the platforms their work makes possible. That is what the platform-driver job at Namba Taxi is, in employment-multiplier terms. That is what the call operator at a Lalafo-using merchant is. Counted that way, the sector’s footprint on the labour market was about 30,000 people in 2017, not 7,400.
If the country could create 5,000 to 7,500 entry-level IT jobs over three years, as the report estimated under the demand conditions observed in 2019, the multiplier suggested 15,000 to 22,500 additional non-technical jobs would land alongside them.
The graduate pipeline that wasn’t keeping up
The training side of the labour market was where the gap between projection and reality was most visible.
8,650 graduates per year from universities is a large number on paper. The number behind that number — the 47% drop in enrolment over the five years — was the structural signal. Students were already voting on the universities with their feet. They were enrolling in short-term private courses instead. KSSDA’s IT Academy was reporting 300–400 graduates a year. DevCIT at AUCA, similar. Attractor School, which had just opened a branch in Almaty, similar again. Together, the three best-known short-course providers were producing around 1,000 graduates per year by 2018 — and the next tier, around 60 smaller training providers, was adding several hundred more on top of that.
The Karakol focus-group participants in February 2019 — students who were planning to do six-month certificates rather than finish their degrees — had read the same signal we were reading from the numbers. The certificate, in their telling, plus a portfolio, plus an active Telegram presence, would get them a salary above the regional average. They had not even waited for someone to write a paper telling them that.
Where the vacancies actually were
The second piece of the labour-market puzzle the chapter laid out was the demand side, channel by channel.
Two things to read in that chart. First, the most active labour-market channel for the IT community in Kyrgyzstan, by 2019, was a Telegram channel — not a job board. Jobs Devkg was where the demand actually surfaced. The number of vacancies posted there grew 2.7 times between 2017 and 2018. Second, the conventional channels — job.kg, headhunter.kg, employment.kg — were not absent, but they were where larger and slower-moving employers posted. The faster the firm and the more technical the role, the more likely it was on Telegram first.
This had a downstream effect we surfaced repeatedly in the chapter: HR practice in the sector was running on chat, not on systems. The 4,424 resumes sitting on job.kg were where the applicants thought they should apply. The 824 vacancy ads on Jobs Devkg were where the jobs actually were. The matching function between the two was being done by people, manually, in messengers.
Languages: what people knew, what people needed
Closely related, but its own picture:
The one structural imbalance, on the language axis, was Java. Survey respondents reported it as a primary language at 15.1%; the vacancy data implied demand closer to 21%. Several of the in-depth interview respondents named this gap directly: banks and the larger enterprise customers were Java-heavy; the training pipeline (both formal and short-course) was JavaScript- and Python-heavy. The mismatch was small in percentage terms but it was the one that had the most direct effect on the question of whether Kyrgyz developers could move up the value chain into enterprise-scale work.
The low-level languages — Assembler, C, C++ — were almost absent from the survey responses. The report’s hardware chapter explains why: the sub-sectors that need low-level skills (3D printing controllers, IoT firmware, robotics) were under-developed enough that the labour market hadn’t yet signalled a clear demand. The lack of low-level competence was a downstream effect, not an upstream one.
The eight problems every sub-sector named
We came into the cross-sector chapter expecting different problems in different places. Instead we got eight problems in every place, ranked by frequency of mention across all 50 in-depth interviews and 106 survey respondents.
-
Shortage and turnover of trained personnel. Every sub-sector. Phrased the same way every time: we train people, they get good, they leave. One interviewee in the outsourcing chapter put it in the form that we kept hearing: “We have an expression: Belarusian vacuum cleaner, Ukrainian vacuum cleaner. This is when we cultivate cadres for ourselves, as a result, they are taken from here.” The Green Card was named explicitly in three different interviews, in three different sub-sectors. “Only a person begins to develop, he immediately wins the greencard and collects the suitcases. This year, two greencards left.”
-
IT education that didn’t fit market requirements. Universities were too slow to update curricula. Short courses were filling the gap but covered only the in-demand high-level languages and left the supporting craft — design thinking, testing discipline, communication — uncovered. The Jalal-Abad finding from the opening of this post was the most extreme version of the pattern, but the pattern was national.
-
Limited use of international payment systems. Freelancers were routing income through Payoneer because the local banking system charged $15 commission on a $1,000 transfer and delivered it a week late through correspondent banks. Software publishing companies couldn’t sign up to Stripe or PayPal because of geographic restrictions or low turnover thresholds. The animation studios were funded entirely through Google Ireland Limited as the YouTube counterparty. None of this was illegal; all of it was happening outside the formal Kyrgyz banking system.
-
Lack of investment. No venture funds. No formalised angel association — though the team counted about 20 visible private and institutional investors backing roughly 64 IT projects, with a combined commitment under $1 million. Banks didn’t lend against intangible assets. Founders self-funded. Founders also, very often, ran outsourcing alongside their product company to pay operating costs while the product matured.
-
Difficulties in legal registration. Both at company-formation stage and on the patent side. Several respondents could not find a tax category that fit what they did, and ended up paying under the “computer and photocopy services” voluntary patent — the same category as the man with the printer at the corner shop.
-
Document processing for freelancers. Upwork doesn’t issue a signed contract. The freelancer has nothing to show to a Kyrgyz bank or a Kyrgyz tax inspector to explain where the income came from. Income that can’t be documented cannot be officially recognised. Income that cannot be officially recognised cannot be the basis for a mortgage application, a business loan, or a residence permit for a Central Asian colleague who wants to come work alongside.
-
Low awareness of Kyrgyzstan as an IT destination. Kyrgyzstan did not appear on the A.T. Kearney Global Services Location Index. Did not appear in Software Magazine’s Software 500. Did not appear on Clutch.co’s main directories. One respondent reported losing a foreign contract because the client’s legal team refused to sign with a counterparty in a country ending in “-STAN”.
-
Undeveloped logistics. Strictly an e-commerce problem in terms of the immediate effect — small-scale sellers in Naryn or Talas could not reach buyers outside their oblast without losses on every order — but with knock-on effects in software publishing (no national fulfilment partner to plug a logistics API into) and animation (physical merchandise based on YouTube characters could not be produced and shipped at scale).
Two further problems were mentioned almost as often, and the report grouped them as 9 and 10 in the same list:
-
Limited number of IT academic staff. Both the cause and the consequence of problem 2.
-
Lack of English proficiency at the developer level — the constraint named most often when respondents were asked why Kyrgyz developers earned 500–1,000 USD per month while Belarusian developers with comparable technical skill earned 1,600–1,900.
What respondents asked for, ranked
When the survey asked what conditions were necessary for the IT sector to develop in Kyrgyzstan, the response rates fell out in a particular order:
- Up-to-date curriculum and practitioner-teachers — 75.6% of respondents.
- More employees in the field — 43.3%.
- Favourable environmental conditions for life — 40% (ecology, specifically Bishkek’s winter air, was named repeatedly as a reason senior specialists were leaving).
- Optimisation of the legal framework — 40%.
- Electronic payment system development — 36.5%.
- Local customers honouring contracts — 36.5%.
- Venture capital availability — 32.2%.
- More offices and workspaces, especially preferential ones for startups — 23.3%.
- Increased country recognition among foreign customers — under 10%.
- Simplified document circulation — under 10%.
The bottom two of that list are worth pausing on. The thing the report had spent thirty pages establishing — that Kyrgyzstan’s lack of recognition as an IT destination was a structural ceiling — was being ranked tenth out of ten by the practitioners we surveyed. The practitioners did not disagree that recognition mattered. They were saying that, in 2019, three more urgent things were eating their working week, and recognition could wait.
A year on
Writing this in September 2020, with the pandemic still re-shaping which constraints bite hardest and which can be lived with, two of the eight problems have moved.
Payment systems (problem 3) moved fastest. By April 2020 it was no longer possible to pretend that the inability of a Kyrgyz freelancer to receive a Stripe payment was a niche issue — the same constraint was now hitting everyone who lost office-based work and tried to go remote and freelance. Several of the recommendations the report made in this area, which sat without action through most of 2019, were being acted on by mid-2020.
Brain drain (problem 1) is the one that moved in the other direction, and not in the way anyone hoped. The pandemic did not bring Kyrgyz developers home from abroad. It accelerated departures of the ones who hadn’t yet left, because remote-from-anywhere meant the local salary discount was now bid up against San Francisco rates, not against Bishkek office norms. We will return to this in the recommendations post.
The rest of the list — education, investment, registration, freelancer documentation, recognition, logistics — sits roughly where the report found it in March 2019. The chapter that follows in this series, on software development outsourcing, is where the labour-market findings sharpen into a single sub-sector’s portrait.
Source: USAID Enterprise Competitiveness Project (2019). Analysis of the Information Technology Sector in the Kyrgyz Republic. Implemented by KG Labs Public Foundation; commissioned by USAID ECP / Nathan Associates / ACDI-VOCA, June 2019. Field research January–March 2019 across Bishkek, Osh, Jalal-Abad, and Karakol: 106-respondent survey, 18 in-depth interviews, 5 focus groups, plus desk research and primary-data scraping of Telegram, LinkedIn, and freelance platforms.