Research and analysis

Executive summary of the impact of export promotion activities on firm outcomes

Published 4 October 2023

About Frontier Economics

Frontier Economics was commissioned by the Department for Business and Trade (DBT) to assess the economic impact of DBT’s export promotion activities on the firms which receive them.

Our study covers a range of export promotion activities provided to firms in the UK by DBT including:

• providing advice, information and support to firms looking to export through the Exporting is GREAT service
• leveraging overseas networks to create opportunities for British exporters
• finding overseas buyers for UK firms which are looking to export

These activities are intended to encourage new firms to export and current exporters to export more. In the longer term, the aim is to increase the output and success of UK firms by increasing the export intensity of their business models where possible.

We examine how receipt of DBT support provided between 2014 and 2016 impacted firm-level economic outcomes up to 3 years after the support was received.

The 5 key firm-level economic outcomes examined are:

• survival
• goods exports
• employment
• turnover
• productivity (using turnover per employee as a proxy)

Previous research, systematically reviewed in our “Lessons Learned” report (Frontier Economics, 2019), identified some evidence that receipt of export promotion services is associated with positive impacts in terms of goods exports, employment, turnover and productivity (Mion and Muuls, 2015; Rincón-Aznar and others, 2015).

Our report builds on these previous studies with more up-to-date and comprehensive data. This allows us to examine variation in the impact of DBT support over time, as well as variation by firm size and by the number of interactions between DBT and supported firms.

Data

The dataset we use is created from administrative sources, most importantly the Inter-departmental Business Register (IDBR), His Majesty’s Revenue and Customs (HMRC) overseas trade statistics and survey data provided by DBT.

The resulting dataset is a yearly panel that contains records of the key firm-level outcomes listed above, interactions with DBT and potential drivers of DBT support for around 1.5 million firms between 2014 and 2016. This is along with earlier data that helps us test whether the necessary assumptions for our analysis are valid. Approximately 1.5% of the firms in our dataset were supported by DBT in some way.

Within the timeline of this project, and given the data available on the economic outcomes of interest, we were able to estimate the impact of DBT support up to 3 years later for survival. For goods export status and goods export value, this was up to 2 years later for employment and it was up to 3 years later for turnover and turnover per employee.

Table 1 summarises the outcome data available.

Table 1: outcome data available for estimating the impact of DBT support, by support year and outcome

Support year	Employment	Turnover and turnover per employee	Survival, export value and export status
2014	t+1, t+2, t+3	t+1, t+2, t+3	t+1, t+2, t+3
2015	t+1, t+2	t+1, t+2	t+1, t+2
2016	t+1	t+1	t+1

Methodology

Our dataset tells us how firms which received DBT export promotion support performed in subsequent years in terms of our outcomes (survival, goods exports, employment, turnover and productivity).

We do not, however, know what those outcomes would have been in the absence of support from DBT. We therefore need an econometric model to establish the counterfactual (what outcomes would have occurred for each firm without export promotion support). This will allow us to estimate the impact of DBT export promotion support.

This estimation is complicated by the fact that supported firms may differ systematically from the wider population of UK firms. For example, a firm that seeks DBT support may be more likely than the average UK firm to already be an exporter and be more likely to operate in a trade-intensive industry such as manufacturing.

To account for these differences between supported and unsupported firms, we take a combined propensity score matching and difference-in-differences (PSM-DiD) approach. This allows us to estimate the impact of DBT export promotion support (the “treatment effect”).

Our approach can be summarised in 3 main steps.

1. We model the likelihood that a firm with a given set of characteristics will receive export promotion support from DBT (its “propensity score”).

2. Based on the propensity score, we match “treated firms” (those receiving DBT support), to “control” firms. Control firms are unsupported firms that look very similar to supported firms in terms of recent growth in key outcomes, industry, location, age and other observable characteristics.

3. We transform the outcomes of interest into difference form, by subtracting the baseline pre-treatment value for each firm, and estimate the average difference in outcomes between the matched treatment and control groups. This gives us a measure of the impact of DBT export promotion support on our outcomes.

For survival, there is no baseline difference as all firms “survive” before they are treated. By subtracting the baseline pre-treatment values, we ensure that our results are not driven by pre-existing differences in outcomes. Although the matching process should ensure that the treatment and control firms have similar baseline pre-treatment outcomes, it is possible that differences remain

This PSM-DiD approach is in line with previous scholarship in the area and is discussed at length in our Lessons Learned report (Frontier Economics, 2019).

Along with average effects across all supported firms, we also present our estimates separately for small firms, of fewer than 50 employees, and medium and large firms, of 50 or more employees.

This is to reflect the potential for variation in:

• the different drivers of treatment probability by firm size
• the impact of DBT support by firm size

In addition, our analysis is carried out separately for outcomes 1, 2 and 3 years after export promotion services are provided. This allows the analysis to give an indication of both the short- and medium-term impact of DBT support.

Results for the impact of DBT support

Table 2 summarises the estimated impact of DBT support on outcomes 1, 2 and 3 years after treatment.

These results are representative of the impact of DBT support for the support profile of the average supported firm (the results cover the mix of different profiles of export promotion support received by firms in practice). We consider impacts for more specific profiles of DBT support below.

Overall, our results suggest that DBT support has a statistically significantly positive impact on the survival, probability of exporting goods and employment of supported firms.

The impacts of support on turnover and value of export are also statistically significant if analysed in a logarithmic specification. A logarithmic specification uses the change in the log of the outcome as the dependent variable, allowing impacts to be measured in proportional terms so that, for example, treatment gives an 8% uplift in export value.

Given the range of firm sizes, modelling treatment effects with a proportional increase rather than using “linear” specification assumes that the same uplift in absolute terms is applied, regardless of the starting size. This is borne out by the logarithmic effects, which show positive impacts that increase over time, are statistically significant and are of broadly similar magnitude when comparing across different outcomes.

By contrast, there is much less congruence in the linear results for employment, turnover and export value. It should be noted that the linear £ outcomes have considerable volatility over time and are likely to be more prone to the effects of outliers.

Table 2: impact of DBT support for all treated firms, all firm sizes, all treatment years

Outcomes	t+1	t+2	t+3
Survival	2.2%*	3.4%*	4.4%*
Export status	7.2%*	8.0%*	7.7%*
Export value (£000s)	175	369	784
Employment	5.7*	10*	5.3*
Turnover (£000s)	4,197	1,717	6,013
Log export value	0.083*	0.087*	0.108*
Log employment	0.049*	0.062*	0.064*
Log turnover	0.038*	0.058*	0.067*

An asterisk indicates a statistically significant difference in the matched average treatment effect at the 95% confidence level.

It is worth noting that the impact of DBT support on outcomes 2 and 3 years later results from a combination of the initial treatment episode (at year t) and, for a majority of firms (around 60% of our sample), repeated support in following years.

Survival

As Table 3 shows, we find consistent evidence of positive, statistically significant survival impacts as a result of receiving support from DBT for both small firms and large firms. There are slightly larger impacts estimated for small firms.

Table 3: effect of DBT support on firm survival, by firm size, all treatment years

	t+1	t+2	t+3
All firms	2.2%*	3.4%*	4.4%*
Small firms	2.5%*	3.8%*	4.8%*
Medium and large firms	0.7%*	1.6%*	2.8%*

An asterisk indicates any year with a statistically significant difference in the matched average treatment effect at the 95% level or higher. For results by firm size, results are deemed statistically significant if they are statistically significant for at least one of the separately estimated years. For the “all firms” results, statistical significance is based on a weighted average of the results by firm size.

Survival effects grow over time. For small firms, survival effects associated with DBT support rise from 2.2 percentage points one year after treatment to 3.4 percentage points after 2 years and 4.4 percentage points after 3 years.

For medium and large firms, survival effects associated with DBT support rise from 0.7 percentage points one year after treatment to 1.6 percentage points after 2 years and 2.8 percentage points after 3 years.

The presence of a survival effect makes the interpretation of other estimates more challenging. Firms supported by DBT are more likely than control firms to be active 1 year, 2 years and 3 years from treatment.

This makes it difficult to understand to what extent the effect of DBT support on other outcomes (for example, employment) at t+1 (where t is the treatment year) is due to impacts on firms that would have survived anyway, regardless of support, or due to effects on which firms make it to t+1.

However, the estimated survival effects are relatively small, especially if compared to baseline survival rates. Therefore, the selection into survival is likely to have a relatively small influence on our estimated impact of DBT support on other outcomes.

Goods exports

We find that DBT support is associated with a higher likelihood of exporting goods. This effect is statistically significant for all firm sizes and broadly stable over time.

When measured in a logarithmic specification, the effect on value of exports is positive and significant, slightly increasing over time. Overall, the results are driven by small firms, which consistently see bigger and more significant impacts than medium and large firms.

Table 4: effect of DBT support on goods exports, by firm size, all treatment years

Size band	Outcome	t+1	t+2	t+3
All firms	Export status	7.2%*	8.0%*	7.7%*
Small firms	Export status	7.6%*	8.3%*	8.0%*
Medium and large firms	Export status	5.8%*	6.7%*	6.1%
All firms	Export value (£000s)	175	369	784
Small firms	Export value (£000s)	121	214	329
Medium and large firms	Export value (£000s)	402	1,010	2,718
All firms	Log export value	0.083*	0.087*	0.108*
Small firms	Log export value	0.093*	0.111*	0.115*
Medium and large firms	Log export value	0.06	0.037	0.094

An asterisk indicates any year with a statistically significant difference in the matched average treatment effect at the 95% level or higher. For results by firm size, results are deemed statistically significant if they are statistically significant for at least one of the separately estimated years. For the “all firms” results, statistical significance is based on a weighted average of the results by firm size.

The effect of treatment on export status is larger for small firms both in absolute terms (the size of the treatment effect) and relative to the baseline probability of exporting. Specifically, DBT support is associated with a 7.6 percentage point increase in the proportion of exporters among the small treated firms, from a baseline of around 30% of exporters.

Among large firms, treatment is associated with a 5.8 percentage point increase, from a baseline of around 70% of exporters. This is in line with international evidence on the impact of export promotion (for example, Broocks and Van Biesebroeck, 2017; Munch and Schaur, 2018).

These effect sizes suggest that export promotion services provided in 2014 and 2015 may have led around 2,900 additional small firms and around 600 additional medium and large firms to start exporting 2 years after first treatment.

Since our data focuses on goods exports, and there is no equivalent record of services exports at the firm level, this analysis may underestimate the impact of DBT support on total exports (including goods and services).

Employment and turnover

As Table 5 shows, our results suggest that DBT support is associated with an increase in firm size through employee count. The effect of support on employment is primarily driven by an impact on small firms, an effect that is roughly constant over time.

The effect for medium and large firms is rather unstable, non-significant one and 3 years after support but significant 2 years after support. The results for log turnover and log employment are also positive, significant and driven by small firms, with results for medium and large firms positive, unstable and insignificant.

Note that variation in persistence of impacts over time is also affected by the changing sample composition, with t+1 estimated from 2014, 2015 and 2016 cohorts, but t+3 estimated only from the 2014 cohort. The results for turnover are not statistically significant.

Table 5: effect of DBT support on firm employment and turnover, by firm size, all treatment years

Outcome	Size band	t+1	t+2	t+3
Employment	All firms	5.7*	10*	5.3*
Employment	Small firms	0.6*	0.8*	0.9*
Employment	Medium and large firms	26.7	47.8*	23.3
Turnover (£000s)	All firms	4,197	1,717	-6,013
Turnover (£000s)	Small firms	-215	-2944	32
Turnover (£000s)	Medium and large firms	22,848	20,819	-31,176
Log employment	All firms	0.049*	0.062*	0.064*
Log employment	Small firms	0.055*	0.069*	0.074*
Log employment	Medium and large firms	0.025*	0.035*	0.027
Log turnover	All firms	0.038*	0.058*	0.067*
Log turnover	Small firms	0.042*	0.061*	0.074*
Log turnover	Medium and large firms	0.02	0.048*	0.041

An asterisk indicates any year with a statistically significant difference in the matched average treatment effect. For results by firm size, results are deemed statistically significant if they are statistically significant for at least one of the separately estimated years. For the “all firms” results, statistical significance is based on a weighted average of the results by firm size.

Similar to Rincón-Aznar and others (2015), we find some evidence of positive effects of treatment on the employment of larger firms.

Small treated firms would have grown by 0.6 fewer employees on average after one year without treatment. Medium and large treated firms would have grown by 26.7 fewer employees after one year, around one-third of their actual growth.

Treatment duration

As noted above, the results in Table 1 to Table 5 combine results for firms that received DBT support in a single year (for example, 2014 only) and multiple years (for example, 2014 and 2015).

As an extension, we considered how these impacts break down if splitting by treatment duration. Overall, we might expect that a longer treatment duration is associated with larger impacts from DBT support.

There are a number of reasons for believing this:

• for a firm to be receiving support in a later year, by definition, it must have survived and, absent any effects, this will be correlated with higher growth than if it had not survived
• firms seeking more substantial support in the first place may have more ambitious growth plans than those that do not, or they may subsequently seek further support following on from successful experience of exporting from the first round. In other words, renewed support may follow ongoing export success
• absent any selection or survival effects, DBT support may bring direct benefits

There may also be differences in the composition of the groups, which could also affect performance. In fact, when comparing small firms in the 2014 cohort, we find the 2 groups have similar sector and region profiles.

However, points of difference are that those receiving treatment in multiple years are:

• larger (12 employees vs 9)
• more productive (turnover per employee 17% higher)
• more likely to already export (44% vs 29%)
• more innovative (twice as likely to have innovation support and significantly more likely to file patents)

Overall, treatment in multiple years is associated with larger effects that grow over time, across all the outcomes exhibiting positive impacts (probability of exporting and employment). For example, after 2 years, the impact on exporter status for the multiple-treatment years group relative to one year only is 8 percentage points higher for small firms and 3 percentage points higher for medium/large firms. Treatment in multiple years is associated with survival rates between 1 and 6 percentage points higher than if treated in just one year.

Treatment intensity within a year

The results in Table 1 to Table 5 also combine results for firms that received only one instance of support from DBT within a year with firms that received multiple instances of support from DBT within a year. Our data suggests that the latter is more common, with each firm receiving an average of around 3 service deliveries per year.

A more detailed breakdown of these results suggests that more intensive treatment is associated with larger impacts from DBT support on export status.

Among small firms, receiving multiple instances of support, rather than a single instance is linked with an increase in the probability of exporting in the following year of 2 to 5 percentage points, depending on treatment year. This is a sizeable effect compared to the impact of single treatment (5.2 percentage points for treatment in 2014 and 6.4 percentage points in 2015 respectively). However, the results for other outcomes are not statistically significant.

As in the case of treatment duration, the results from this comparison should be interpreted with caution. It is possible that different types of firms select into our single- and multiple-treatment groups.

Geographic variation in results

We also explored whether the results vary at a geographic level. This involved aggregating one-year impacts over the 3 treatment years in scope and dividing the UK into 5 aggregate regions, and exploring these separately. We continued to distinguish small firms from medium and large firms.

This analysis found little variation in impacts by region. While there were some differences in results, this should be expected in any exercise involving repeated cuts of the data. It was not obvious that any region outperformed others across multiple metrics.